Datasets:
wanshenl
/
pgsql-hackers

Dataset card Data Studio Files
xet
Community
threads
listlengths
1
2.99k
[ { "msg_contents": "Hi\n\nappended below is a simple database schema (which may not\nbe a candidate for the next Nobel Prize for SQL Database\nDesign, but represents enough of a production\ndatabase to demonstrate the following problem).\n\nAnd that is:\n\nunder 7.3 this statement:\n SELECT foo_id, thingy_name, bar_name\n FROM foo_view, bar\n WHERE bar_id=foo_bar_id\n\nproduces the desired results.\n\nThis however:\n SELECT foo_id, thingy_name, bar_name\n FROM foo_view\n INNER JOIN bar ON bar_id=foo_bar_id\n\nproduces\n\n ERROR: ExecEvalExpr: unknown expression type 108\n\nThe latter statement does however work in 7.1.3 with no\napparent problems.\n\nQuestion: what does \"unknown expression type 108\" mean and\nwhy should it suddenly occur in 7.3? A bit of Googling\nreveals the same message occurs when using subselects\nin constraints, but that doesn't seem related to this case.\n\n\nIan Barwick\nbarwick@gmx.net\n\n\n-- sample DB for \"unknown expression type 108\" error\n\nCREATE TABLE a_thingy (\n a_id INT,\n a_firstname VARCHAR(64),\n a_lastname VARCHAR(64),\n PRIMARY KEY (a_id)\n);\n\nCREATE TABLE b_thingy (\n b_id INT,\n b_name VARCHAR(64),\n PRIMARY KEY (b_id)\n);\n\nCREATE TABLE bar (\n bar_id INT,\n bar_name varchar(64),\n PRIMARY KEY (bar_id)\n);\n\nCREATE TABLE foo (\n foo_id INT,\n foo_a_id INT REFERENCES a_thingy NULL, \n foo_b_id INT REFERENCES b_thingy NULL,\n foo_bar_id INT REFERENCES bar NOT NULL,\n PRIMARY KEY (foo_id),\n CHECK((foo_a_id IS NOT NULL AND foo_b_id IS NULL) OR\n (foo_b_id IS NOT NULL AND foo_a_id IS NULL))\n);\n\nCREATE VIEW foo_view AS\n SELECT *,\n CASE\n WHEN foo_a_id IS NOT NULL THEN \n (SELECT a_lastname || ', ' || a_firstname\n FROM a_thingy\n WHERE a_id=foo_a_id\n )\n WHEN foo_b_id IS NOT NULL THEN \n (SELECT b_name\n FROM b_thingy\n WHERE b_id=foo_b_id\n )\n END \n AS thingy_name\n FROM foo;\n\nINSERT INTO a_thingy VALUES\n (1, 'John', 'Doe');\n\nINSERT INTO b_thingy VALUES\n (1, 'Megacorp');\n\nINSERT INTO bar VALUES(1, 'squid');\nINSERT INTO bar VALUES(2, 'octopus');\n\nINSERT INTO foo VALUES (1,1,NULL,1);\nINSERT INTO foo VALUES (2,NULL,1,2);\n\n-- END\n\n", "msg_date": "Fri, 6 Dec 2002 00:06:12 +0100", "msg_from": "Ian Barwick <barwick@gmx.net>", "msg_from_op": true, "msg_subject": "Q: \"unknown expression type 108\" ?" }, { "msg_contents": "Ian Barwick <barwick@gmx.net> writes:\n> ERROR: ExecEvalExpr: unknown expression type 108\n\nThis is fixed as of yesterday --- see thread on (I think) pgsql-general\nfor the patch.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 06 Dec 2002 14:50:20 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Q: \"unknown expression type 108\" ? " }, { "msg_contents": "On Friday 06 December 2002 20:50, Tom Lane wrote:\n> Ian Barwick <barwick@gmx.net> writes:\n> > ERROR: ExecEvalExpr: unknown expression type 108\n>\n> This is fixed as of yesterday --- see thread on (I think) pgsql-general\n> for the patch.\n\nMany thanks. Applied and appears to work :-)\n\nFor reference:\nhttp://groups.google.com/groups?dq=&hl=en&lr=&ie=UTF-8&selm=2630.1039125066%40sss.pgh.pa.us\n\n\nIan Barwick\nbarwick@gmx.net\n\n\n", "msg_date": "Fri, 6 Dec 2002 23:29:33 +0100", "msg_from": "Ian Barwick <barwick@gmx.net>", "msg_from_op": true, "msg_subject": "Re: Q: \"unknown expression type 108\" ?" } ]
[ { "msg_contents": "Folks,\n\nWe have a marketing group: PGSQL-ADVOCACY. Our problem is that we\ndon't have enough volunteers.\n\nFor example, last week Robert and Justin had job crises, and I left for\nthe mountains for Thanksgiving. As a result Marc had to pitch in at\nthe last minute to try to get some kind of release out. Thus the lack\nof coordinated media splash for the 7.3 release.\n\nWe need more people!!! We have right now about 7 active volunteers and\n6-8 translators for Advocacy. That's not nearly enough. If the people\non this thread care about marketing Postgresql, then please join the\npgsql-advocacy mailing list.\n\nBTW, we do coordinate with the Website development group, and for that\nmatter TechDocs.\n\n-Josh Berkus\n", "msg_date": "Thu, 05 Dec 2002 15:37:28 -0800", "msg_from": "\"Josh Berkus\" <josh@agliodbs.com>", "msg_from_op": true, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" }, { "msg_contents": "Hi Josh,\n\nI'm in. My time is limited, and I do like coding for Postgres as well. I'd\nlike to contribute articles, etc. once we have some sort of portal up and\nrunning...\n\nBTW, is there any way of making the subject line say [ADVOCACY] instead of\n[pgsql-advocacy]?\n\nChris\n\n----- Original Message -----\nFrom: \"Josh Berkus\" <josh@agliodbs.com>\nTo: <pgsql-hackers@postgresql.org>\nSent: Thursday, December 05, 2002 3:37 PM\nSubject: Re: [HACKERS] [GENERAL] PostgreSQL Global Development Group\n\n\n> Folks,\n>\n> We have a marketing group: PGSQL-ADVOCACY. Our problem is that we\n> don't have enough volunteers.\n>\n> For example, last week Robert and Justin had job crises, and I left for\n> the mountains for Thanksgiving. As a result Marc had to pitch in at\n> the last minute to try to get some kind of release out. Thus the lack\n> of coordinated media splash for the 7.3 release.\n>\n> We need more people!!! We have right now about 7 active volunteers and\n> 6-8 translators for Advocacy. That's not nearly enough. If the people\n> on this thread care about marketing Postgresql, then please join the\n> pgsql-advocacy mailing list.\n>\n> BTW, we do coordinate with the Website development group, and for that\n> matter TechDocs.\n>\n> -Josh Berkus\n>\n> ---------------------------(end of broadcast)---------------------------\n> TIP 3: if posting/reading through Usenet, please send an appropriate\n> subscribe-nomail command to majordomo@postgresql.org so that your\n> message can get through to the mailing list cleanly\n>\n\n", "msg_date": "Fri, 6 Dec 2002 13:06:59 -0800", "msg_from": "\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] [GENERAL] PostgreSQL Global Development Group" } ]
[ { "msg_contents": "I just sync'd up with cvs and tried to make clean then configure, and I'm \ngetting this:\n\nconfig.status: linking ./src/backend/libpq/v6util.c to \nsrc/interfaces/libpq/v6util.c\nconfig.status: error: ./src/backend/libpq/v6util.c: file not found\n\nIs this a missing file from the ipv6 stuff that just got committed?\n\nThanks,\n\nJoe\n\n", "msg_date": "Thu, 05 Dec 2002 20:07:40 -0800", "msg_from": "Joe Conway <mail@joeconway.com>", "msg_from_op": true, "msg_subject": "configure error on cvs tip" }, { "msg_contents": "\nFixing now. This just isn't my night --- another patch with a missing\nfile.\n\n\n---------------------------------------------------------------------------\n\nJoe Conway wrote:\n> I just sync'd up with cvs and tried to make clean then configure, and I'm \n> getting this:\n> \n> config.status: linking ./src/backend/libpq/v6util.c to \n> src/interfaces/libpq/v6util.c\n> config.status: error: ./src/backend/libpq/v6util.c: file not found\n> \n> Is this a missing file from the ipv6 stuff that just got committed?\n> \n> Thanks,\n> \n> Joe\n> \n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 2: you can get off all lists at once with the unregister command\n> (send \"unregister YourEmailAddressHere\" to majordomo@postgresql.org)\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Thu, 5 Dec 2002 23:11:08 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: configure error on cvs tip" }, { "msg_contents": "Bruce Momjian wrote:\n> Fixing now. This just isn't my night --- another patch with a missing\n> file.\n> \n\nOK - I can run configure and make now, but I'm getting these warnings:\n\nIn file included from ../../../../src/include/libpq/libpq.h:22,\n from printtup.c:20:\n../../../../src/include/libpq/v6util.h:3: warning: `struct addrinfo' declared \ninside parameter list\n../../../../src/include/libpq/v6util.h:3: warning: its scope is only this \ndefinition or declaration, which is probably not what you want.\n../../../../src/include/libpq/v6util.h:5: warning: `struct addrinfo' declared \ninside parameter list\n\nlots of similar warnings to the above -- and:\n\nauth.c: In function `ClientAuthentication':\nauth.c:414: warning: passing arg 1 of `isAF_INETx' from incompatible pointer type\ngcc -O2 -g -Wall -Wmissing-prototypes -Wmissing-declarations \n-I../../../src/include -c -o crypt.o crypt.c -MMD\nIn file included from ../../../src/include/libpq/libpq.h:22,\n from crypt.c:24:\n\nJoe\n\n", "msg_date": "Thu, 05 Dec 2002 20:27:24 -0800", "msg_from": "Joe Conway <mail@joeconway.com>", "msg_from_op": true, "msg_subject": "Re: configure error on cvs tip" }, { "msg_contents": "\nYep, I am about to yank out the whole patch. I am seeing on postmaster\nstartup:\n\n\tLOG: FATAL: StreamServerPort: getaddrinfo2() failed: hostname\n\tnor servname provided, or not known\n\t\nWhat is strange is that initdb worked. I will just throw it back to the\nauthor.\n\nDone. Code is returned to author for review.\n\n---------------------------------------------------------------------------\n\nJoe Conway wrote:\n> Bruce Momjian wrote:\n> > Fixing now. This just isn't my night --- another patch with a missing\n> > file.\n> > \n> \n> OK - I can run configure and make now, but I'm getting these warnings:\n> \n> In file included from ../../../../src/include/libpq/libpq.h:22,\n> from printtup.c:20:\n> ../../../../src/include/libpq/v6util.h:3: warning: `struct addrinfo' declared \n> inside parameter list\n> ../../../../src/include/libpq/v6util.h:3: warning: its scope is only this \n> definition or declaration, which is probably not what you want.\n> ../../../../src/include/libpq/v6util.h:5: warning: `struct addrinfo' declared \n> inside parameter list\n> \n> lots of similar warnings to the above -- and:\n> \n> auth.c: In function `ClientAuthentication':\n> auth.c:414: warning: passing arg 1 of `isAF_INETx' from incompatible pointer type\n> gcc -O2 -g -Wall -Wmissing-prototypes -Wmissing-declarations \n> -I../../../src/include -c -o crypt.o crypt.c -MMD\n> In file included from ../../../src/include/libpq/libpq.h:22,\n> from crypt.c:24:\n> \n> Joe\n> \n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Thu, 5 Dec 2002 23:36:54 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: configure error on cvs tip" }, { "msg_contents": "Bruce Momjian wrote:\n> Yep, I am about to yank out the whole patch. I am seeing on postmaster\n> startup:\n> \n> \tLOG: FATAL: StreamServerPort: getaddrinfo2() failed: hostname\n> \tnor servname provided, or not known\n> \t\n> What is strange is that initdb worked. I will just throw it back to the\n> author.\n> \n> Done. Code is returned to author for review.\n> \n\nhmmm -- now I'm back to :\n\nconfig.status: linking ./src/backend/libpq/v6util.c to \nsrc/interfaces/libpq/v6util.c\nconfig.status: error: ./src/backend/libpq/v6util.c: file not found\n\nJoe\n\n\n", "msg_date": "Thu, 05 Dec 2002 20:39:50 -0800", "msg_from": "Joe Conway <mail@joeconway.com>", "msg_from_op": true, "msg_subject": "Re: configure error on cvs tip" }, { "msg_contents": "BJoe Conway wrote:\n> Bruce Momjian wrote:\n> > Yep, I am about to yank out the whole patch. I am seeing on postmaster\n> > startup:\n> > \n> > \tLOG: FATAL: StreamServerPort: getaddrinfo2() failed: hostname\n> > \tnor servname provided, or not known\n> > \t\n> > What is strange is that initdb worked. I will just throw it back to the\n> > author.\n> > \n> > Done. Code is returned to author for review.\n> > \n> \n> hmmm -- now I'm back to :\n> \n> config.status: linking ./src/backend/libpq/v6util.c to \n> src/interfaces/libpq/v6util.c\n> config.status: error: ./src/backend/libpq/v6util.c: file not found\n\nSorry. Run autoconf.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Thu, 5 Dec 2002 23:44:16 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: configure error on cvs tip" }, { "msg_contents": "Bruce Momjian wrote:\n> Sorry. Run autoconf.\n> \n\nOK -- works now. I've never needed to do that before.\n\nThanks!\n\nJoe\n\n", "msg_date": "Thu, 05 Dec 2002 20:47:45 -0800", "msg_from": "Joe Conway <mail@joeconway.com>", "msg_from_op": true, "msg_subject": "Re: configure error on cvs tip" }, { "msg_contents": "\nThe INETv6 patch was rejected because of this report, and an error on\npostmaster startup from BSD/OS:\n\n\tLOG: FATAL: StreamServerPort: getaddrinfo2() failed: hostname nor\n\tservname provided, or not known\n\nPlease submit a new patch that addresses these issues. I can work with\nyou to do testing.\n\n---------------------------------------------------------------------------\n\nJoe Conway wrote:\n> Bruce Momjian wrote:\n> > Fixing now. This just isn't my night --- another patch with a missing\n> > file.\n> > \n> \n> OK - I can run configure and make now, but I'm getting these warnings:\n> \n> In file included from ../../../../src/include/libpq/libpq.h:22,\n> from printtup.c:20:\n> ../../../../src/include/libpq/v6util.h:3: warning: `struct addrinfo' declared \n> inside parameter list\n> ../../../../src/include/libpq/v6util.h:3: warning: its scope is only this \n> definition or declaration, which is probably not what you want.\n> ../../../../src/include/libpq/v6util.h:5: warning: `struct addrinfo' declared \n> inside parameter list\n> \n> lots of similar warnings to the above -- and:\n> \n> auth.c: In function `ClientAuthentication':\n> auth.c:414: warning: passing arg 1 of `isAF_INETx' from incompatible pointer type\n> gcc -O2 -g -Wall -Wmissing-prototypes -Wmissing-declarations \n> -I../../../src/include -c -o crypt.o crypt.c -MMD\n> In file included from ../../../src/include/libpq/libpq.h:22,\n> from crypt.c:24:\n> \n> Joe\n> \n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 6 Dec 2002 00:13:09 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "IPv6 patch rejected" } ]
[ { "msg_contents": "\n\n> -----Original Message-----\n> From: Josh Berkus [mailto:josh@agliodbs.com] \n> Sent: 05 December 2002 23:37\n> To: pgsql-hackers@postgresql.org\n> Subject: Re: [HACKERS] [GENERAL] PostgreSQL Global Development Group\n> \n> \n> BTW, we do coordinate with the Website development group\n\nWhen did that happen then? I think I must have blinked :-)\n\nRegards, Dave.\n", "msg_date": "Fri, 6 Dec 2002 08:31:56 -0000", "msg_from": "\"Dave Page\" <dpage@vale-housing.co.uk>", "msg_from_op": true, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" }, { "msg_contents": "Dave,\n\n> > \n> > BTW, we do coordinate with the Website development group\n> \n> When did that happen then? I think I must have blinked :-)\n\nMarc and Justin are periodically keeping the Advocacy group informed\nof progress on wwwdevel, and we were asked to test it before. Vince\nasked us for suggestions, too. \n\nIt's not like Advocacy has so much time to mess around with the\nAdvocacy site that we need weekly updates from WWW as well ...\n\n==========================================\n\nPostges People:\n\nWhat really troubles me is that I'm seeing the *implication* on this\nlist that one or more people offered to help the WWW team and were\nrejected. If this is true, I'd like to see that person say so\nexplicitly and we can find out from Vince and Marc what happened; if\nnot, I think the insinuations about \"exclusiveness\" by the WWW team are\ncompletely uncalled for. \n\n-Josh Berkus\n\n\n", "msg_date": "Fri, 06 Dec 2002 09:44:47 -0800", "msg_from": "\"Josh Berkus\" <josh@agliodbs.com>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" }, { "msg_contents": "On Fri, 6 Dec 2002, Josh Berkus wrote:\n\n> Dave,\n>\n> > >\n> > > BTW, we do coordinate with the Website development group\n> >\n> > When did that happen then? I think I must have blinked :-)\n>\n> Marc and Justin are periodically keeping the Advocacy group informed\n> of progress on wwwdevel, and we were asked to test it before. Vince\n> asked us for suggestions, too.\n\nI did what? When?\n\nVince.\n-- \n Fast, inexpensive internet service 56k and beyond! http://www.pop4.net/\n http://www.meanstreamradio.com http://www.unknown-artists.com\n Internet radio: It's not file sharing, it's just radio.\n\n", "msg_date": "Sun, 8 Dec 2002 00:41:51 -0500 (EST)", "msg_from": "Vince Vielhaber <vev@michvhf.com>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" } ]
[ { "msg_contents": "I'm working on a new interface to PostgreSQL and having a difficult\ntime. Perhaps someone here can shed some light on the subject.\n\n1.  Am I correct in thinking that the OID that I receive with the\nRowDescription is the column type?  I have several of those defined from\nthe 7.3 source code.  Have those changed in the past and are they likely\nto change in the future?  I have heard people talking about querying the\nserver for type names or type oid values but I can't believe it is that\ndifficult.\n\n2.  How do I receive the column modifers such as not null, primary key,\nindexed, etc.  I know RowDescription sends down type modifiers but on an\nint4 column that field is always -1.  I use another client app to change\nthe column from allowing nulls to not allowing nulls (same value in\ncolumn). The client app as well as psql on Linux sees the change but my\ninterface still receives just -1 for the type modifier.  What am I\nmissing?\n\nI have read through all the docs/faqs I can find and still have no good\nidea about this.\n\nIf there is someplace that has the oids documented or the low-level\nconstants documented, I would love to see that.\n\nThanks\nReggie\n\n\n", "msg_date": "Fri, 6 Dec 2002 08:55:24 -0600", "msg_from": "\"Reggie Burnett\" <rykr@bellsouth.net>", "msg_from_op": true, "msg_subject": "new interface" }, { "msg_contents": "On Fri, 2002-12-06 at 09:55, Reggie Burnett wrote:\n> I'm working on a new interface to PostgreSQL and having a difficult\n> time. Perhaps someone here can shed some light on the subject.\n> \n> 1. Am I correct in thinking that the OID that I receive with the\n> RowDescription is the column type? I have several of those defined from\n\nThe OID is just a relatively unique database wide row marker -- which\nyou cannot count on existing in user tables. Nothing more or less. In\nsome of the system tables it is used as a foreign key to records in\nother system tables, but Oid by itself has no such meaning.\n\n> \n> I have read through all the docs/faqs I can find and still have no good\n> idea about this.\n\nTake a look at the 'System Catalog' documentation, and in psql the -E\noption will show you the 'system' queries used to retrieve the\ninformation for display in psql.\n\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "06 Dec 2002 10:19:04 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": false, "msg_subject": "Re: new interface" }, { "msg_contents": "\n\n> -----Original Message-----\n> From: Rod Taylor [mailto:rbt@rbt.ca]\n> Sent: Friday, December 06, 2002 9:19 AM\n> To: Reggie Burnett\n> Cc: PostgreSQL-development\n> Subject: Re: [HACKERS] new interface\n> \n> On Fri, 2002-12-06 at 09:55, Reggie Burnett wrote:\n> > I'm working on a new interface to PostgreSQL and having a difficult\n> > time. Perhaps someone here can shed some light on the subject.\n> >\n> > 1. Am I correct in thinking that the OID that I receive with the\n> > RowDescription is the column type? I have several of those defined\nfrom\n> \n> The OID is just a relatively unique database wide row marker -- which\n> you cannot count on existing in user tables. Nothing more or less.\nIn\n> some of the system tables it is used as a foreign key to records in\n> other system tables, but Oid by itself has no such meaning.\n\nSo if OID can't be counted on to be there 100% of the time, what do I\nuse to index the catalog to determine type, not null-ness, uniqueness,\netc?\n\n> \n> >\n> > I have read through all the docs/faqs I can find and still have no\ngood\n> > idea about this.\n> \n> Take a look at the 'System Catalog' documentation, and in psql the -E\n> option will show you the 'system' queries used to retrieve the\n> information for display in psql.\n\n\n> \n> --\n> Rod Taylor <rbt@rbt.ca>\n> \n> PGP Key: http://www.rbt.ca/rbtpub.asc\n\n", "msg_date": "Fri, 6 Dec 2002 09:20:27 -0600", "msg_from": "\"Reggie Burnett\" <rykr@bellsouth.net>", "msg_from_op": true, "msg_subject": "Re: new interface" }, { "msg_contents": "Let me see if I can add more information here. PSQL appears to only\nretrieve table metadata when \\d is given. For my interface, I will need\nthe ability to present every column as a proper system object (int, byte\narray, currency, etc) depending on the db type. I will not be able to\nwait for the user of my interface to execute a method before retrieving\nthis data so I will need to automatically retrieve it every time. \n\nHaving looked at the query issued by \\d in PSQL, it would seem that I\nwould need to issue that same query for every column in my select. I\nsay that because it contains a join of pg_class and pg_attribute.\n\nAm I understanding this right?\n\n\n> -----Original Message-----\n> From: Rod Taylor [mailto:rbt@rbt.ca]\n> Sent: Friday, December 06, 2002 9:19 AM\n> To: Reggie Burnett\n> Cc: PostgreSQL-development\n> Subject: Re: [HACKERS] new interface\n> \n> On Fri, 2002-12-06 at 09:55, Reggie Burnett wrote:\n> > I'm working on a new interface to PostgreSQL and having a difficult\n> > time. Perhaps someone here can shed some light on the subject.\n> >\n> > 1. Am I correct in thinking that the OID that I receive with the\n> > RowDescription is the column type? I have several of those defined\nfrom\n> \n> The OID is just a relatively unique database wide row marker -- which\n> you cannot count on existing in user tables. Nothing more or less.\nIn\n> some of the system tables it is used as a foreign key to records in\n> other system tables, but Oid by itself has no such meaning.\n> \n> >\n> > I have read through all the docs/faqs I can find and still have no\ngood\n> > idea about this.\n> \n> Take a look at the 'System Catalog' documentation, and in psql the -E\n> option will show you the 'system' queries used to retrieve the\n> information for display in psql.\n> \n> --\n> Rod Taylor <rbt@rbt.ca>\n> \n> PGP Key: http://www.rbt.ca/rbtpub.asc\n\n", "msg_date": "Fri, 6 Dec 2002 09:30:10 -0600", "msg_from": "\"Reggie Burnett\" <rykr@bellsouth.net>", "msg_from_op": true, "msg_subject": "Re: new interface" }, { "msg_contents": "On Fri, 2002-12-06 at 10:30, Reggie Burnett wrote:\n> Let me see if I can add more information here. PSQL appears to only\n> retrieve table metadata when \\d is given. For my interface, I will need\n> the ability to present every column as a proper system object (int, byte\n> array, currency, etc) depending on the db type. I will not be able to\n> wait for the user of my interface to execute a method before retrieving\n> this data so I will need to automatically retrieve it every time. \n\n> Having looked at the query issued by \\d in PSQL, it would seem that I\n> would need to issue that same query for every column in my select. I\n> say that because it contains a join of pg_class and pg_attribute.\n\nSomething like that I would suspect.\n\nBut if they're arbitrary selects, how are you going to handle:\n\nSELECT CAST(32 AS text);\n\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "06 Dec 2002 10:41:19 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": false, "msg_subject": "Re: new interface" } ]
[ { "msg_contents": "----- Original Message ----- \nFrom: Igor Georgiev \nTo: mlw \nSent: Friday, December 06, 2002 4:57 PM\nSubject: Re: Postmaster windows shell\n\n\n\n----- Original Message ----- \nFrom: \"mlw\" <pgsql@mohawksoft.com>\nTo: \"Igor Georgiev\" <gory@alphasoft-bg.com>\nSent: Friday, December 06, 2002 3:43 PM\nSubject: Re: Postmaster windows shell\n\n\n> Yea, send it.\n> \n> Is it GPL?\n\nI just receive approval from gborg.\nThis project is released under the Mozilla Public License (MPL). \nhttp://gborg.postgresql.org/project/winmaster/projdisplay.php \n\nbut with my zero expirience with cvs i dont upload nothing yet :(((\nany help be welcome :)\n\nhere is raw source with Dev-C++ project\n\n> Does it compile under cygwin or MSC?\n> (I have both)\n> \nmingw, but i thin it will run wtoh -mno-cygwin\n\n\n\n\n\n\n\n\n \n----- Original Message ----- \nFrom: Igor \nGeorgiev \nTo: mlw \nSent: Friday, December 06, 2002 4:57 PM\nSubject: Re: Postmaster windows shell\n\n \n----- Original Message ----- \nFrom: \"mlw\" <pgsql@mohawksoft.com>\nTo: \"Igor Georgiev\" <gory@alphasoft-bg.com>\nSent: Friday, December 06, 2002 3:43 \nPM\nSubject: Re: Postmaster windows \nshell\n\n> Yea, send it.> > Is it \nGPL?\n \nI just receive approval from \ngborg.\nThis project is released under the Mozilla Public \nLicense (MPL). \nhttp://gborg.postgresql.org/project/winmaster/projdisplay.php \n \nbut with my zero expirience with cvs i dont upload \nnothing yet :(((\nany help be welcome :)\n \nhere  is raw \nsource with Dev-C++ \nproject\n> Does it compile under cygwin or \nMSC?> (I have both)> mingw, but i thin it will run wtoh \n-mno-cygwin", "msg_date": "Fri, 6 Dec 2002 16:59:31 +0100", "msg_from": "\"Igor Georgiev\" <gory@alphasoft-bg.com>", "msg_from_op": true, "msg_subject": "Re: Postmaster windows shell" } ]
[ { "msg_contents": "Joe, have you heard of a standard called SQL/MED? I came across a\ndescription of it the other day. You might think it's got some medical\nconnotation, but actually the acronym is Management of External Data,\nand what it is is a standard spec for shipping chunks of SQL queries to\nremote servers. For instance, given\n\n\tSELECT * FROM a.foo, b.bar WHERE ...\n\nwhere a.foo is on a remote machine, the spec lays down how the local and\nremote servers can cooperate to execute this query intelligently ---\nincluding deciding where to execute various WHERE clauses to minimize\nthe amount of data shipped. (The article I found was actually about\nhow the new draft version of SQL/MED improves the protocol to let this\nsort of thing be done better; it seems the original spec only allowed\nretrieval of a whole table's contents.)\n\nThis looks like it might be a great long-term replacement for dblink,\nand if it is standard, so much the better.\n\nI imagine the draft version of the new SQL/MED spec may be available on\nthe web, but haven't gone looking.\n\nJust a heads-up in case you are interested...\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 06 Dec 2002 11:04:58 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": true, "msg_subject": "SQL/MED spec for cross-database linkages" }, { "msg_contents": "On Fri, 2002-12-06 at 11:04, Tom Lane wrote:\n> Joe, have you heard of a standard called SQL/MED? I came across a\n\nIt's that one of the later parts of the SQL 99 spec? Section 9 or\nthereabouts?\n\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "06 Dec 2002 11:35:25 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": false, "msg_subject": "Re: SQL/MED spec for cross-database linkages" }, { "msg_contents": "On Fri, 2002-12-06 at 11:35, Rod Taylor wrote:\n> On Fri, 2002-12-06 at 11:04, Tom Lane wrote:\n> > Joe, have you heard of a standard called SQL/MED? I came across a\n> \n> It's that one of the later parts of the SQL 99 spec? Section 9 or\n> thereabouts?\n\nNote to self, don't reply to emails and have a separate conversation at\nthe same time.\n\nFound SQL/MED, Section 21 of ISO 9075-9.\n\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "06 Dec 2002 11:49:03 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": false, "msg_subject": "Re: SQL/MED spec for cross-database linkages" }, { "msg_contents": "Tom Lane wrote:\n> Joe, have you heard of a standard called SQL/MED? I came across a\n> description of it the other day. You might think it's got some medical\n> connotation, but actually the acronym is Management of External Data,\n> and what it is is a standard spec for shipping chunks of SQL queries to\n> remote servers. For instance, given\n> \n> \tSELECT * FROM a.foo, b.bar WHERE ...\n> \n> where a.foo is on a remote machine, the spec lays down how the local and\n> remote servers can cooperate to execute this query intelligently ---\n> including deciding where to execute various WHERE clauses to minimize\n> the amount of data shipped. (The article I found was actually about\n> how the new draft version of SQL/MED improves the protocol to let this\n> sort of thing be done better; it seems the original spec only allowed\n> retrieval of a whole table's contents.)\n> \n> This looks like it might be a great long-term replacement for dblink,\n> and if it is standard, so much the better.\n\nGreat! Thanks for the heads up. I see that Rod provided the reference in his \npost -- I'll go find it.\n\nThe idea of expanding dblink to other RDBMSs is picking up steam. I've been \nconversing off list with someone who has a semi-working hacked version of \ndblink that uses JDBC in place of libpq.\n\nDo you think a proposal based on the SQL/MED spec would be entertained for \n7.4, or would the release after be a safer bet? I'm not sure (since I haven't \nseen it yet) what I'm getting myself into ;-), but I might like to take it on \nsince there seems to be a lot of interest.\n\nJoe\n\n\nJoe\n\n", "msg_date": "Fri, 06 Dec 2002 10:02:47 -0800", "msg_from": "Joe Conway <mail@joeconway.com>", "msg_from_op": false, "msg_subject": "Re: SQL/MED spec for cross-database linkages" }, { "msg_contents": "Rod Taylor <rbt@rbt.ca> writes:\n> Found SQL/MED, Section 21 of ISO 9075-9.\n\nThat's the old version, though. The new draft is at (digs out article)\nhttp://sqlstandards.org/SC32/WG3/Progression_Documents/FCD/4FCD1-14-XML-2002-03.pdf\naccording to this article, but I'm not having any luck accessing that\nURL. Let me email the authors and see if I can get a copy.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 06 Dec 2002 13:52:01 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": true, "msg_subject": "Re: SQL/MED spec for cross-database linkages " }, { "msg_contents": "Joe Conway <mail@joeconway.com> writes:\n> Do you think a proposal based on the SQL/MED spec would be entertained for \n> 7.4, or would the release after be a safer bet?\n\nThe impression I get from this article is that SQL/MED is nontrivial.\nIf you think you can get something useful going for 7.4, step right up\n--- but I think you should plan on being at it for awhile. In any\ncase, a multi-step implementation plan would be a good idea.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 06 Dec 2002 14:38:50 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": true, "msg_subject": "Re: SQL/MED spec for cross-database linkages " }, { "msg_contents": "On Fri, Dec 06, 2002 at 01:52:01PM -0500, Tom Lane wrote:\n> Rod Taylor <rbt@rbt.ca> writes:\n> > Found SQL/MED, Section 21 of ISO 9075-9.\n> \n> That's the old version, though. The new draft is at (digs out article)\n> http://sqlstandards.org/SC32/WG3/Progression_Documents/FCD/4FCD1-14-XML-2002-03.pdf\n> according to this article, but I'm not having any luck accessing that\n> URL. Let me email the authors and see if I can get a copy.\n\nHmm, seems you need to change protocols:\n\nftp://sqlstandards.org/SC32/WG3/Progression_Documents/FCD/4FCD1-14-XML-2002-03.pdf\n\nRoss\n", "msg_date": "Fri, 6 Dec 2002 13:53:50 -0600", "msg_from": "\"Ross J. Reedstrom\" <reedstrm@rice.edu>", "msg_from_op": false, "msg_subject": "Re: SQL/MED spec for cross-database linkages" }, { "msg_contents": "On Fri, Dec 06, 2002 at 01:52:01PM -0500, Tom Lane wrote:\n> Rod Taylor <rbt@rbt.ca> writes:\n> > Found SQL/MED, Section 21 of ISO 9075-9.\n> \n> That's the old version, though. The new draft is at (digs out article)\n> http://sqlstandards.org/SC32/WG3/Progression_Documents/FCD/4FCD1-14-XML-2002-03.pdf\n> according to this article, but I'm not having any luck accessing that\n> URL. Let me email the authors and see if I can get a copy.\n> \n> \t\t\tregards, tom lane\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 6: Have you searched our list archives?\n\n\nHaving pulled that one, I discover it's actually the XML related spec. The \nMED one is at:\n\nftp://sqlstandards.org/SC32/WG3/Progression_Documents/FCD/4FCD1-09-MED-2002-01.pdf\n\nThere's a text version, as well (always nice for grepping):\n\nftp://sqlstandards.org/SC32/WG3/Progression_Documents/FCD/4FCD1-09-MED-2002-01.txt\n\nRoss\n", "msg_date": "Fri, 6 Dec 2002 13:58:20 -0600", "msg_from": "\"Ross J. Reedstrom\" <reedstrm@rice.edu>", "msg_from_op": false, "msg_subject": "Re: SQL/MED spec for cross-database linkages" }, { "msg_contents": "\"Ross J. Reedstrom\" <reedstrm@rice.edu> writes:\n> Having pulled that one, I discover it's actually the XML related spec. The \n> MED one is at:\n\n> ftp://sqlstandards.org/SC32/WG3/Progression_Documents/FCD/4FCD1-09-MED-2002-01.pdf\n\nAh-hah, so the URL in the SIGMOD Record article is just wrong :-(\n\nMeanwhile, one of the authors sent me a copy privately (it helps to have\nshared an office at grad school ;-)) which I was about to offer around,\nbut I guess we can just use the FTP link instead.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 06 Dec 2002 15:56:57 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": true, "msg_subject": "Re: SQL/MED spec for cross-database linkages " } ]
[ { "msg_contents": "I have a program which worked merrily under 1.3b1, but with 1.4devel\nI get the first query working (so connection OK), and the second query,\nwhich may well be wrong, getting me:\n\nDEBUG: reaping dead processes\nDEBUG: child process (pid 13025) was terminated by signal 11\nLOG: server process (pid 13025) was terminated by signal 11\nLOG: terminating any other active server processes\nDEBUG: CleanupProc: sending SIGQUIT to process 13022\nWARNING: Message from PostgreSQL backend:\n The Postmaster has informed me that some other backend\n died abnormally and possibly corrupted shared memory.\n I have rolled back the current transaction and am\n going to terminate your database system connection and exit.\n Please reconnect to the database system and repeat your query.\nDEBUG: reaping dead processes\nDEBUG: child process (pid 13022) exited with exit code 1\nLOG: all server processes terminated; reinitializing shared memory and\nsemaphores\n\nShould I be able to cause this with a duff query?\n\nI'm now digging myself in further by trying to recompile the program which\nuses libpq++ - I don't think libpq++ works terribly well anymore but\nunfortunately I have to leave now before figuring out what pg_config.h\nincludes may be needed to define eg DLLIMPORT etc.\n\nMore later no doubt...\n\nCheers,\n\nPatrick\n", "msg_date": "Fri, 6 Dec 2002 17:34:37 +0000", "msg_from": "Patrick Welche <prlw1@newn.cam.ac.uk>", "msg_from_op": true, "msg_subject": "possible libpq++ prob" }, { "msg_contents": "Patrick Welche <prlw1@newn.cam.ac.uk> writes:\n> WARNING: Message from PostgreSQL backend:\n> The Postmaster has informed me that some other backend\n> died abnormally and possibly corrupted shared memory.\n\n> Should I be able to cause this with a duff query?\n\nNo. That's a backend bug, not a libpq++ bug. Let's see the query ...\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 06 Dec 2002 15:35:44 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: possible libpq++ prob " } ]
[ { "msg_contents": "We have the following query:\n\nSELECT certificate_id\n INTO TEMP TABLE x_certs\n FROM certificate\n WHERE cert_status = 0 AND\n certificate_id BETWEEN 1111 AND 2222 AND\n client_id IN (1, 2, 3)\n ORDER BY certificate_id\n FOR UPDATE;\n\nIs there any reason that this query should lock the entire certificate table? \n Is there something strange because of the IN clause or because it is going \ninto a temporary table? This is a production server running 7.2.2 so perhaps \nit is fixed in 7.3.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.\n", "msg_date": "Fri, 6 Dec 2002 14:03:02 -0500", "msg_from": "\"D'Arcy J.M. Cain\" <darcy@druid.net>", "msg_from_op": true, "msg_subject": "SELECT FOR UPDATE locks whole table" }, { "msg_contents": "\nIt should lock only the rows you retrieved, but I have no idea how FOR\nUPDATE and INTO TEMP behave. My guess is that it should work fine, but\nI have never seen those two used together before.\n\n---------------------------------------------------------------------------\n\nD'Arcy J.M. Cain wrote:\n> We have the following query:\n> \n> SELECT certificate_id\n> INTO TEMP TABLE x_certs\n> FROM certificate\n> WHERE cert_status = 0 AND\n> certificate_id BETWEEN 1111 AND 2222 AND\n> client_id IN (1, 2, 3)\n> ORDER BY certificate_id\n> FOR UPDATE;\n> \n> Is there any reason that this query should lock the entire certificate table? \n> Is there something strange because of the IN clause or because it is going \n> into a temporary table? This is a production server running 7.2.2 so perhaps \n> it is fixed in 7.3.\n> \n> -- \n> D'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\n> http://www.druid.net/darcy/ | and a sheep voting on\n> +1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 6 Dec 2002 14:10:46 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: SELECT FOR UPDATE locks whole table" }, { "msg_contents": "\"D'Arcy J.M. Cain\" <darcy@druid.net> writes:\n> We have the following query:\n> SELECT certificate_id\n> INTO TEMP TABLE x_certs\n> FROM certificate\n> WHERE cert_status = 0 AND\n> certificate_id BETWEEN 1111 AND 2222 AND\n> client_id IN (1, 2, 3)\n> ORDER BY certificate_id\n> FOR UPDATE;\n\n> Is there any reason that this query should lock the entire certificate\n> table?\n\nIt should only lock the selected rows ... and does, in a quick test\nhere. Would you provide the test case that makes you think it's doing\notherwise?\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 06 Dec 2002 18:00:24 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: SELECT FOR UPDATE locks whole table " }, { "msg_contents": "On December 6, 2002 02:10 pm, Bruce Momjian wrote:\n> It should lock only the rows you retrieved, but I have no idea how FOR\n> UPDATE and INTO TEMP behave. My guess is that it should work fine, but\n> I have never seen those two used together before.\n\nTurns out that it wasn't the SELECT ... FOR UPDATE that was causing the \nproblem. I did a test like this.\n\n> > SELECT certificate_id\n> > INTO TEMP TABLE x_certs\n> > FROM certificate\n> > WHERE cert_status = 0 AND\n> > certificate_id BETWEEN 1111 AND 2222 AND\n> > client_id IN (1, 2, 3)\n> > ORDER BY certificate_id\n> > FOR UPDATE;\n\nBasically this query after a BEGIN TRANSACTION except without the INTO part. \nThen I went to another window and tried to update two certificates, one \ninside and one outside the range. The first failed and the second succeeded \nas expected. I then updated one of the certs in the range. After that I \ncould not update any certificates until I closed the transaction. Very weird.\n\nI then built a new database and repeated the experiment with fresh, simple \ntables and was able to confirm that normally PostgreSQL does NOT have this \nbehaviour so then I started thinking about differences between the simple \nsetup and our real production setup. One thing that I thought of was that \nthe real database has this trigger on certificate.\n\nCREATE TRIGGER mk_cardnum\n BEFORE INSERT OR UPDATE ON certificate\n FOR EACH ROW\n EXECUTE PROCEDURE mk_cardnum (cardnum, certificate_id, validation);\n\nmk_cardnum is a C function that reads certificate_id and validation and \nwrites something into cardnum. My understanding is that this only affects \nthe row(s) being updated. It must since this table has over seven million \nrecords and we would notice if it took minutes to do a simple update.\n\nI couldn't find anything in the docs or web specifically about this. Does \nanyone have any ideas?\n\nAdding hackers as this may be an internal issue.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.\n", "msg_date": "Sat, 7 Dec 2002 08:17:48 -0500", "msg_from": "\"D'Arcy J.M. Cain\" <darcy@druid.net>", "msg_from_op": true, "msg_subject": "Re: [SQL] SELECT FOR UPDATE locks whole table" } ]
[ { "msg_contents": "Hi all,\n\nI didn't find any other list compatible with this post, I hope it is the right\nplace.\nI got a sparccenter 2000 (sun4d) few weeks ago and I wondered if the\npostgresql team was interested in testing postgres on it. It has only 2 CPUs\nat this time but I may get more cpu and system boards in the next month.\nI thought it would have been interesting to test postgres on uch an\narchitecture.\n\nAre you interested ? Would you be interested ?\n\nIt runs solaris 8 and linux, but linux doesn't support smp on sun4d :/\n... and since the kernel team doesn't work on sparc32 anymore, I thought it\ncould be used by user-land software projects.\n\nThe machine is located on an dsl lins 128/512kbps.\n\nseeya\n\na+\n\n-- \nEric GENTILINI\n", "msg_date": "Fri, 6 Dec 2002 21:10:24 +0100", "msg_from": "Eric Gentilini <egentili@nerim.net>", "msg_from_op": true, "msg_subject": "hardware needed ?" }, { "msg_contents": "Hi Eric,\n\nWe've already done a fair amount of testing of PostgreSQL on Sparc \nhardware and Solaris, so it's probably not all that interesting... :-/\n\nHowever, about an hour after you sent through your message, we received \nthis one from Myk Melez. He's asking us if there is a publically \navailable PostgreSQL server, and there isn't yet, so I'm wondering if \nyou'd be willing to setup a PostgreSQL server on your Sparccenter 2000 \nand let anyone anywhere connect to it, for a while at least.\n\nWe wouldn't want it to be extremely long term, as you never know what \nun-cool things people could decide to store in there, but it might be \nuseful for a week or two after Myk's article becomes available for readers.\n\nWould you be interested in this?\n\n:-)\n\nRegards and best wishes,\n\nJustin Clift\n\n\n-------- Original Message --------\nSubject: [GENERAL] publicly available PostGreSQL server?\nDate: Fri, 06 Dec 2002 12:33:10 -0800\nFrom: Myk Melez <myk@mozilla.org>\nOrganization: mozilla.org\nTo: pgsql-general@postgresql.org\n\nIs there a publicly available PostGreSQL server? I'm co-writing an\narticle for an online journal that includes information about\nMozilla's upcoming database support, and I'd like to point readers to\na PostGreSQL installation where they can try out an example app\nwithout having to install their own server.\n\nMore info about database support in Mozilla:\n\nhttp://bugzilla.mozilla.org/show_bug.cgi?id=81653\n\n-myk\n\n\nEric Gentilini wrote:\n> Hi all,\n> \n> I didn't find any other list compatible with this post, I hope it is the right\n> place.\n> I got a sparccenter 2000 (sun4d) few weeks ago and I wondered if the\n> postgresql team was interested in testing postgres on it. It has only 2 CPUs\n> at this time but I may get more cpu and system boards in the next month.\n> I thought it would have been interesting to test postgres on uch an\n> architecture.\n> \n> Are you interested ? Would you be interested ?\n> \n> It runs solaris 8 and linux, but linux doesn't support smp on sun4d :/\n> ... and since the kernel team doesn't work on sparc32 anymore, I thought it\n> could be used by user-land software projects.\n> \n> The machine is located on an dsl lins 128/512kbps.\n> \n> seeya\n> \n> a+\n\n\n-- \n\"My grandfather once told me that there are two kinds of people: those\nwho work and those who take the credit. He told me to try to be in the\nfirst group; there was less competition there.\"\n- Indira Gandhi\n\n", "msg_date": "Sat, 07 Dec 2002 14:23:08 +1100", "msg_from": "Justin Clift <justin@postgresql.org>", "msg_from_op": false, "msg_subject": "Re: hardware needed ?" } ]
[ { "msg_contents": "\n\n> -----Original Message-----\n> From: Josh Berkus [mailto:josh@agliodbs.com] \n> Sent: 06 December 2002 17:45\n> To: pgsql-hackers@postgresql.org\n> Subject: Re: [HACKERS] [GENERAL] PostgreSQL Global Development Group\n> \n> \n> Dave,\n> \n> > > \n> > > BTW, we do coordinate with the Website development group\n> > \n> > When did that happen then? I think I must have blinked :-)\n> \n> Marc and Justin are periodically keeping the Advocacy group informed\n> of progress on wwwdevel, and we were asked to test it before. Vince\n> asked us for suggestions, too. \n> \n> It's not like Advocacy has so much time to mess around with \n> the Advocacy site that we need weekly updates from WWW as well ...\n\nAhh, it's the other way round we don't see (advocacy -> www).\n\n> ==========================================\n> \n> Postges People:\n> \n> What really troubles me is that I'm seeing the *implication* \n> on this list that one or more people offered to help the WWW \n> team and were\n> rejected. If this is true, I'd like to see that person say so\n> explicitly and we can find out from Vince and Marc what \n> happened; if not, I think the insinuations about \n> \"exclusiveness\" by the WWW team are completely uncalled for. \n\nYes, this worries me. We have very recently had a couple of new\nvolunteers join the team who have been actively helping me with the new\nportal - I've certainly not heard of anyone being turned away. It was\nalso not long ago that Vince solicited new design ideas from the\ncommunity on -general.\n\nWrt the closed status of the list, I think (please correct me if I'm\nwrong Marc/Vince) this is mainly because we try to keep it very focused.\nIt is very much a developers meeting place where we *all* contribute\nheavily to the work. It is not a support list, or general discussion\nlist.\n\nRegards, Dave.\n", "msg_date": "Fri, 6 Dec 2002 21:43:08 -0000", "msg_from": "\"Dave Page\" <dpage@vale-housing.co.uk>", "msg_from_op": true, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" } ]
[ { "msg_contents": "Hi,\n\nWhat do the columns conforencoding and contoencoding refer to in\npg_conversion?\n\nHow would I convert those numbers to a string encoding name, just using SQL?\n\nChris\n\n", "msg_date": "Fri, 6 Dec 2002 17:33:32 -0800", "msg_from": "\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au>", "msg_from_op": true, "msg_subject": "pg_conversion" }, { "msg_contents": "> What do the columns conforencoding and contoencoding refer to in\n> pg_conversion?\n> \n> How would I convert those numbers to a string encoding name, just using SQL?\n> \n> Chris\n\nUse pg_encoding_to_char().\n--\nTatsuo Ishii\n", "msg_date": "Sat, 07 Dec 2002 19:11:13 +0900 (JST)", "msg_from": "Tatsuo Ishii <t-ishii@sra.co.jp>", "msg_from_op": false, "msg_subject": "Re: pg_conversion" } ]
[ { "msg_contents": "\nRod, can you comment on these warnings Joe Conway's compiler is showing?\nI don't see the warnings with gcc, but clearly they look like problems.\n\nI can just assign a NULL on definition, but I thought you should take a\nlook.\n\n---------------------------------------------------------------------------\n\nJoe Conway wrote:\n> Hi Bruce,\n> \n> I just sync'd up/make clean/make all and get this:\n> \n> gcc -O2 -g -Wall -Wmissing-prototypes -Wmissing-declarations \n> -I../../../src/include -c -o typecmds.o typecmds.c -MMD\n> typecmds.c: In function `AlterDomainAddConstraint':\n> typecmds.c:1237: warning: `ccbin' might be used uninitialized in this function\n> typecmds.c: In function `get_rels_with_domain':\n> typecmds.c:1450: warning: `rtc' might be used uninitialized in this function\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 6 Dec 2002 22:57:56 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: more compile warnings" } ]
[ { "msg_contents": "\nI now have the INETv6 patch working using IPv4 on my machine, and I\ndon't have IPv6 enabled in my kernel. Tomorrow, I will review the\nentire patch for portability issues, then post it so others can test it.\nI am pretty sure it is going to fail if your machine isn't INET6 aware,\nwhich may be many.\n\n---------------------------------------------------------------------------\n\n> > The INETv6 patch was rejected because of this report, and an error on\n> > postmaster startup from BSD/OS:\n> > \n> > \tLOG: FATAL: StreamServerPort: getaddrinfo2() failed: hostname nor\n> > \tservname provided, or not known\n> > \n> > Please submit a new patch that addresses these issues. I can work with\n> > you to do testing.\n> > \n> > ---------------------------------------------------------------------------\n> > \n> > Joe Conway wrote:\n> > > Bruce Momjian wrote:\n> > > > Fixing now. This just isn't my night --- another patch with a missing\n> > > > file.\n> > > > \n> > > \n> > > OK - I can run configure and make now, but I'm getting these warnings:\n> > > \n> > > In file included from ../../../../src/include/libpq/libpq.h:22,\n> > > from printtup.c:20:\n> > > ../../../../src/include/libpq/v6util.h:3: warning: `struct addrinfo' declared \n> > > inside parameter list\n> > > ../../../../src/include/libpq/v6util.h:3: warning: its scope is only this \n> > > definition or declaration, which is probably not what you want.\n> > > ../../../../src/include/libpq/v6util.h:5: warning: `struct addrinfo' declared \n> > > inside parameter list\n> > > \n> > > lots of similar warnings to the above -- and:\n> > > \n> > > auth.c: In function `ClientAuthentication':\n> > > auth.c:414: warning: passing arg 1 of `isAF_INETx' from incompatible pointer type\n> > > gcc -O2 -g -Wall -Wmissing-prototypes -Wmissing-declarations \n> > > -I../../../src/include -c -o crypt.o crypt.c -MMD\n> > > In file included from ../../../src/include/libpq/libpq.h:22,\n> > > from crypt.c:24:\n> > > \n> > > Joe\n> > > \n> > > \n> > > ---------------------------(end of broadcast)---------------------------\n> > > TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org\n> > > \n> > \n> > \n> \n> -- \n> \n> \n> Nigel Kukard (Chief Executive Officer)\n> Lando Technologies Africa (Pty) Ltd\n> nigel@lando.co.za www.lando.co.za\n> Tel: 083 399 5822 Fax: 086 1100036\n> Hoheisen Park Bellville, Cape Town\n> National Internet Service Provider\n> \n> \n> The best language to use is the language that was designed for\n> what you want to use it for - 1997\n> \n> \n> =====================================================================\n> \n> Disclaimer\n> ----------\n> The contents of this message and any attachments are intended \n> solely for the addressee's use and may be legally privileged and/or \n> confidential information. This message may not be retained, \n> distributed, copied or used if you are not he addressee of this \n> message. If this message was sent to you in error, please notify \n> the sender immediately by reply e-mail and then destroy the message \n> and any copies thereof.\n> \n> Opinions, conclusions and other information in this message may be \n> personal to the sender and is not that of Lando Technologies Africa \n> or any of it's subsideries, associated companies or principals and \n> is therefore not endorsed by any of the Lando groups of companies. \n> Due to e-maill communication being insecure, Lando groups of \n> companies do not guarantee confidentiality, security, accuracy or \n> performance of the e-mail. Any liability for viruses is excluded \n> to the fullest extent.\n> \n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 6 Dec 2002 23:06:43 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: IPv6 patch rejected" }, { "msg_contents": "Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> I am pretty sure it is going to fail if your machine isn't INET6 aware,\n> which may be many.\n\nThat's definitely not gonna do :-(\n\n\t\t\tregards, tom lane\n", "msg_date": "Sat, 07 Dec 2002 16:11:36 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: IPv6 patch rejected " } ]
[ { "msg_contents": "> -----Original Message-----\n> From: Bruce Momjian [mailto:pgman@candle.pha.pa.us] \n> Sent: Friday, December 06, 2002 7:58 PM\n> To: Joe Conway\n> Cc: rbt@zort.ca; PostgreSQL-development\n> Subject: Re: [HACKERS] more compile warnings\n> \n> \n> Rod, can you comment on these warnings Joe Conway's compiler \n> is showing? I don't see the warnings with gcc, but clearly \n> they look like problems.\n> \n> I can just assign a NULL on definition, but I thought you \n> should take a look.\n> \n> --------------------------------------------------------------\n> -------------\n> \n> Joe Conway wrote:\n> > Hi Bruce,\n> > \n> > I just sync'd up/make clean/make all and get this:\n> > \n> > gcc -O2 -g -Wall -Wmissing-prototypes -Wmissing-declarations \n> > -I../../../src/include -c -o typecmds.o typecmds.c -MMD\n> > typecmds.c: In function `AlterDomainAddConstraint':\n> > typecmds.c:1237: warning: `ccbin' might be used \n> uninitialized in this \n> > function\n> > typecmds.c: In function `get_rels_with_domain':\n> > typecmds.c:1450: warning: `rtc' might be used uninitialized \n> in this function\n\nInstead of just assigning a value, it means it is conceivable that a\npath allows undefined behavior. Example:\n\n...\n int y;\n if (x < 5) \n y = 3;\n if (x > 5)\n y = 1;\n\nWhat happens if x == 5? Then y is indeterminate.\n\nSometimes, the flow analysis just gets confused and it really will be\ninitialized along every path. But at least it bears checking. Hence\nthe warning.\n\nI like to use -W -Wall -ansi -pedantic -O3\n", "msg_date": "Fri, 6 Dec 2002 20:10:20 -0800", "msg_from": "\"Dann Corbit\" <DCorbit@connx.com>", "msg_from_op": true, "msg_subject": "Re: more compile warnings" }, { "msg_contents": "Dann Corbit wrote:\n> Instead of just assigning a value, it means it is conceivable that a\n> path allows undefined behavior. Example:\n> \n> ...\n> int y;\n> if (x < 5) \n> y = 3;\n> if (x > 5)\n> y = 1;\n> \n> What happens if x == 5? Then y is indeterminate.\n\n> Sometimes, the flow analysis just gets confused and it really will be\n> initialized along every path. But at least it bears checking. Hence\n> the warning.\n\nYes, I looked at the code, and they are legitimate warnings.\n\n> I like to use -W -Wall -ansi -pedantic -O3\n\nI use:\n\n -Wall -Wmissing-prototypes -Wmissing-declarations -Wpointer-arith -Wcast-align\n\nYou would think that would catch it. My problem is that I am compiling\nwith -O0, because I compile all day and I don't care about optimization.\nIn this case, the -O3 is doing some optimization that catches the\nproblem, while -O0 does not. Interesting. Even -O catches it.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 6 Dec 2002 23:16:30 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: more compile warnings" }, { "msg_contents": "On Fri, Dec 06, 2002 at 11:16:30PM -0500, Bruce Momjian wrote:\n> \n> I use:\n> \n> -Wall -Wmissing-prototypes -Wmissing-declarations -Wpointer-arith -Wcast-align\n \nSome other useful ones are -pedantic -W -Wfloat-equal -Wshadow\n-Wcast-qual -Wwrite-strings -Wconversion -Wsign-compare -Wsign-promo.\n\n\n> You would think that would catch it. My problem is that I am compiling\n> with -O0, because I compile all day and I don't care about optimization.\n> In this case, the -O3 is doing some optimization that catches the\n> problem, while -O0 does not. Interesting. Even -O catches it.\n\nLast time I checked (which was admittedly some time ago) all the\ninteresting analysis that could give you new warnings was done by -O;\n-O2 mostly involves the back-end, and -O3 adds pretty much nothing\nexcept aggressive inlining. Which was more likely to trigger compiler\nbugs at the time than to find anything in your code.\n\n\nJeroen\n\n", "msg_date": "Sat, 7 Dec 2002 05:35:11 +0100", "msg_from": "\"Jeroen T. Vermeulen\" <jtv@xs4all.nl>", "msg_from_op": false, "msg_subject": "Re: more compile warnings" }, { "msg_contents": "Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> You would think that would catch it. My problem is that I am compiling\n> with -O0, because I compile all day and I don't care about optimization.\n\nYou should reconsider that. At -O0 gcc doesn't do any flow analysis,\nand thus you lose many important warnings. I'd recommend -O1 at least.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sat, 07 Dec 2002 16:14:37 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: more compile warnings " }, { "msg_contents": "Tom Lane wrote:\n> Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> > You would think that would catch it. My problem is that I am compiling\n> > with -O0, because I compile all day and I don't care about optimization.\n> \n> You should reconsider that. At -O0 gcc doesn't do any flow analysis,\n> and thus you lose many important warnings. I'd recommend -O1 at least.\n\nYes, I will re-add -O to my flags. When I did it I forgot it would\naffect warnings. In fact, I am now seeing a similar warning in python\nthat I hadn't seen before, and others probably don't see it because they\ndon't compile python.\n\nFYI, -O2 adds only 2 minutes to my 13 minute test script (but increases\nthe cpu usage from 4 to 6 minutes).\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Mon, 9 Dec 2002 13:42:23 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: more compile warnings" } ]
[ { "msg_contents": "I ask me, the store of large file like 2M in one field take to postgres\na big work, you do that ? any of you do that before ? the sistem work\nfine or not ? Thanks for all\n\n--\n Saludos Horacio Miranda.\n hmiranda@yale.cl\n------------------------------------------------------------\nPostgreSQL. Because life's too short to learn Oracle. :)\n Billy O'Connor\n\n IBM --> Immer Backup Machen\n\n\n", "msg_date": "Sat, 07 Dec 2002 00:53:58 -0400", "msg_from": "Horacio Miranda <hmiranda@yale.cl>", "msg_from_op": true, "msg_subject": "about store Large file." } ]
[ { "msg_contents": "All,\n\nI've simplified the Darwin/Mac OS X startup script I submitted earlier \nin the year. This version has only the two files required by the Darwin \nstartup bundle design. Plus the sh script now uses Darwin-standard \nfunctions to start up PostgreSQL, and it checks for the presence of a \nvariable in /etc/hostconfig, as do other Darwin startup scripts.\n\nI suggest that a new directory be created, \ncontrib/start-scripts/darwin, and that these two files be put into it. \nFolks who want to use the script can read the comments inside it to \nfigure out how to use it.\n\nEnjoy,\n\nDavid\n\n-- \nDavid Wheeler AIM: dwTheory\ndavid@wheeler.net ICQ: 15726394\nhttp://david.wheeler.net/ Yahoo!: dew7e\n Jabber: Theory@jabber.org", "msg_date": "Sat, 7 Dec 2002 18:24:18 -0800", "msg_from": "David Wheeler <david@wheeler.net>", "msg_from_op": true, "msg_subject": "Darwin/Mac OS X Startup Script" }, { "msg_contents": "\nAdded to /contrib/start-scripts as:\n\n\n\tPostgreSQL.darwin StartupParameters.plist.darwin\n\nThanks.\n\n---------------------------------------------------------------------------\n\nDavid Wheeler wrote:\n> All,\n> \n> I've simplified the Darwin/Mac OS X startup script I submitted earlier \n> in the year. This version has only the two files required by the Darwin \n> startup bundle design. Plus the sh script now uses Darwin-standard \n> functions to start up PostgreSQL, and it checks for the presence of a \n> variable in /etc/hostconfig, as do other Darwin startup scripts.\n> \n> I suggest that a new directory be created, \n> contrib/start-scripts/darwin, and that these two files be put into it. \n> Folks who want to use the script can read the comments inside it to \n> figure out how to use it.\n> \n> Enjoy,\n> \n> David\n> \n> -- \n> David Wheeler AIM: dwTheory\n> david@wheeler.net ICQ: 15726394\n> http://david.wheeler.net/ Yahoo!: dew7e\n> Jabber: Theory@jabber.org\n\n[ Attachment, skipping... ]\n\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 3: if posting/reading through Usenet, please send an appropriate\n> subscribe-nomail command to majordomo@postgresql.org so that your\n> message can get through to the mailing list cleanly\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Mon, 9 Dec 2002 16:26:54 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Darwin/Mac OS X Startup Script" } ]
[ { "msg_contents": "Hi folks,\n\nLet's create a release team. This strategy is one well established \nin other projects and in industry. For lack of a better starting \nreference, let me suggest http://www.freebsd.org/releng/charter.html \nas a starting point for consideration. See also \nhttp://www.freebsd.org/releng/index.html. \n\nThis will also lighten the load on the core team allowing them to \nfocus on development and such. \n\ncheers\n\n\n\n-- \nDan Langille : http://www.langille.org/\n\n", "msg_date": "Sat, 07 Dec 2002 22:03:48 -0500", "msg_from": "\"Dan Langille\" <dan@langille.org>", "msg_from_op": true, "msg_subject": "Let's create a release team" }, { "msg_contents": "\"Dan Langille\" <dan@langille.org> writes:\n> Let's create a release team. This strategy is one well established \n> in other projects and in industry. For lack of a better starting \n> reference, let me suggest http://www.freebsd.org/releng/charter.html \n> as a starting point for consideration. See also \n> http://www.freebsd.org/releng/index.html. \n\n> This will also lighten the load on the core team allowing them to \n> focus on development and such. \n\nI don't really see any value-added here. The core committee's only\nroutinely-exercised function is to organize releases; separating that\nout would leave core with nothing to do. Also, to the extent that\ncore has any real or perceived authority in the project, I think it\ncomes from having control of the release process --- there's surely\nno other reason for people to defer to the core team as a group (as\nopposed to whatever respect might be accorded to individual people\nas a result of their individual contributions). So ISTM such a\nreorganization would leave the core committee as a figurehead and make\nthe release team into the effective new core.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 11:38:47 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Let's create a release team " }, { "msg_contents": "Tom Lane wrote:\n> as a result of their individual contributions). So ISTM such a\n> reorganization would leave the core committee as a figurehead and make\n> the release team into the effective new core.\n\nI thought we were already only figureheads? ;-)\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Mon, 9 Dec 2002 16:28:12 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Let's create a release team" }, { "msg_contents": "On 9 Dec 2002 at 11:38, Tom Lane wrote:\n\n> \"Dan Langille\" <dan@langille.org> writes:\n> > Let's create a release team. This strategy is one well established\n> > in other projects and in industry. For lack of a better starting\n> > reference, let me suggest http://www.freebsd.org/releng/charter.html\n> > as a starting point for consideration. See also\n> > http://www.freebsd.org/releng/index.html. \n> \n> > This will also lighten the load on the core team allowing them to\n> > focus on development and such. \n> \n> I don't really see any value-added here. The core committee's only\n> routinely-exercised function is to organize releases; separating that\n> out would leave core with nothing to do. \n\nSo we already have a release team, but not titled as such.\n\n> Also, to the extent that\n> core has any real or perceived authority in the project, I think it\n> comes from having control of the release process --- there's surely no\n> other reason for people to defer to the core team as a group (as\n> opposed to whatever respect might be accorded to individual people as\n> a result of their individual contributions).\n\nIs the process documented? Any set procedure? Who knows how to do \nit?\n\n> So ISTM such a\n> reorganization would leave the core committee as a figurehead and make\n> the release team into the effective new core.\n\nIs 'core' the same as 'steering'? I couldn't find any reference to \n\"core committe\" or \"core team\" via google. At \nhttp://developer.postgresql.org/bios.php I see the group of people \nreferred to as \"Steering\". Is their function defined anywhere?\n\nIf these things are not documented, they should be. Where do I \nstart?\n-- \nDan Langille : http://www.langille.org/\n\n", "msg_date": "Mon, 09 Dec 2002 23:51:07 -0500", "msg_from": "\"Dan Langille\" <dan@langille.org>", "msg_from_op": true, "msg_subject": "Re: Let's create a release team " }, { "msg_contents": "Hi Dan,\n\nIt's been mentioned a few times on the Advocacy and Marketing list that \nwe should put together a process for ensuring that all the parts \nnecessary for a release occur properly and smoothly.\n\n***********\n\nSource code\n\n - Initial packaging of the new releases' source code\n\n\nDocs\n\n - Confirm with Peter that the Docs are 100% correct in the new source \narchive\n\n\nRPM's & SRPM's\n\n - Co-ordinate with Lamar to have these ready before the general \nannouncement?\n\n\nPress Releases for the General Public (multiple languages)\n\n - Advocacy and Marketing guys should put together a Press Release \nintended for the General Public, and have it reviewed/confirmed by the \nHackers before getting it ready\n\n - Robert (?) should arrange translation of this \"confirmed good\" Press \nRelease into multiple languages\n\n\nPress Release for the Technically Minded (?)\n\n - Advocacy and Marketing guys (?) should put together a Press Release \nintended for the Hackers and other Technically Minded folk. Should \ndefinitely be reviewed for accuracy by the Hackers before releasing it\n\n\nWebsites\n\n - Ensure all of the required documentation mentions, links, release \ninfo, etc is put in place on the website\n\n\nMailout\n\n - Email the appropriate Press Releases to the General Public, and to \nthe Technically Minded groups\n\n\nFeedback\n\n - Find out what could have been done better, and figure out how to \nmake it so for the next one if appropriate\n\n***********\n\nThat was just what came to mind and there's probably more. Each part \nshould probably be something that can be broken down into the necessary \nparts so that everyone can take care of the bits they're into. I \nsuppose it would be good to have this listed somewhere so that people \ncan make suggestions.\n\nJust whipped up a page listing these main points here, and everyone has \nthe ability to make suggestions/edits directly onto that page:\n\nhttp://advocacy.postgresql.org/documents/ReleaseProcess\n\nHopefully that's helpful.\n\n:-)\n\nRegards and best wishes,\n\nJustin Clift\n\n-- \n\"My grandfather once told me that there are two kinds of people: those\nwho work and those who take the credit. He told me to try to be in the\nfirst group; there was less competition there.\"\n- Indira Gandhi\n\n", "msg_date": "Tue, 10 Dec 2002 16:24:13 +1100", "msg_from": "Justin Clift <justin@postgresql.org>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Let's create a release team" }, { "msg_contents": "\"Dan Langille\" <dan@langille.org> writes:\n> Is the process documented? Any set procedure? Who knows how to do \n> it?\n\nEr ... nope, nope, the core bunch ...\n\n> Is 'core' the same as 'steering'?\n\nYes, the webpage takes some license here. 'core' is the most common\nterminology for the-usual-suspects. I'm not sure where 'steering'\ncame from, but it's the same suspects...\n\n> If these things are not documented, they should be.\n\nMost of the undocumented details of the release process are in the heads\nof Marc Fournier and Bruce Momjian. If either of them falls off the end\nof the earth, we have worse troubles than whether we remember how to do\na release --- for example: Marc owns, runs, and pays for the\npostgresql.org servers. (Me, I just hack code, so I'm replaceable.)\nBut if you want to try to document the process better, there are some\ndetails written down already (eg, src/tools/RELEASE_CHANGES) and I'm\nsure Marc and Bruce would cooperate in writing down more.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Dec 2002 00:56:22 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Let's create a release team " }, { "msg_contents": "On 10 Dec 2002 at 0:56, Tom Lane wrote:\n\n> \"Dan Langille\" <dan@langille.org> writes:\n> > Is the process documented? Any set procedure? Who knows how to do\n> > it?\n> \n> Er ... nope, nope, the core bunch ...\n\nSounds like we need to do a brain dump then. I just happen to have \nsome equipment left over from \"The Matrix\"....\n\n> > If these things are not documented, they should be.\n> \n> Most of the undocumented details of the release process are in the\n> heads of Marc Fournier and Bruce Momjian. If either of them falls off\n> the end of the earth, we have worse troubles than whether we remember\n> how to do a release\n\nOn a project, anyone is replaceable. And anyone might leave for any \nnumber of reasons. If they do, the affect upon the project will be \nminimized by having the major processes documented.\n\n> --- for example: Marc owns, runs, and pays for the\n> postgresql.org servers.\n\nIs the cvs repo mirrored?\n\n> (Me, I just hack code, so I'm replaceable.)\n\nYeah, yeah, stop being humble... ;)\n\n> But if you want to try to document the process better, there are some\n> details written down already (eg, src/tools/RELEASE_CHANGES) and I'm\n> sure Marc and Bruce would cooperate in writing down more.\n\nThat's a good start. It looks like a list of things easily forgotten \nbut if forgotten, make us look bad.\n-- \nDan Langille : http://www.langille.org/\n\n", "msg_date": "Tue, 10 Dec 2002 09:04:49 -0500", "msg_from": "\"Dan Langille\" <dan@langille.org>", "msg_from_op": true, "msg_subject": "Re: Let's create a release team " }, { "msg_contents": "\"Dan Langille\" <dan@langille.org> writes:\n>> --- for example: Marc owns, runs, and pays for the\n>> postgresql.org servers.\n\n> Is the cvs repo mirrored?\n\nAnyone running cvsup would have a complete copy of the source CVS,\nI believe. It would be more troubling to reconstruct the mailing list\narchives; I'm not sure that those are mirrored anywhere. (Marc?)\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Dec 2002 09:34:40 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Let's create a release team " }, { "msg_contents": "On 10 Dec 2002 at 9:34, Tom Lane wrote:\n\n> \"Dan Langille\" <dan@langille.org> writes:\n> >> --- for example: Marc owns, runs, and pays for the\n> >> postgresql.org servers.\n> \n> > Is the cvs repo mirrored?\n> \n> Anyone running cvsup would have a complete copy of the source CVS, I\n> believe. It would be more troubling to reconstruct the mailing list\n> archives; I'm not sure that those are mirrored anywhere\n\nDo you mean the repository, or the source. The repository is the ,v \nfiles.... The source isn't. Most developers would have the source, \nbut not necessarily the repo.\n-- \nDan Langille : http://www.langille.org/\n\n", "msg_date": "Tue, 10 Dec 2002 09:37:20 -0500", "msg_from": "\"Dan Langille\" <dan@langille.org>", "msg_from_op": true, "msg_subject": "Re: Let's create a release team " }, { "msg_contents": "Dan Langille writes:\n > On 10 Dec 2002 at 9:34, Tom Lane wrote:\n > > Anyone running cvsup would have a complete copy of the source CVS, I\n > > believe. It would be more troubling to reconstruct the mailing list\n > > archives; I'm not sure that those are mirrored anywhere\n > Do you mean the repository, or the source. The repository is the ,v \n > files.... The source isn't. Most developers would have the source, \n > but not necessarily the repo.\n\nSee:\n\n http://www.cvsup.org/\n\nIt mirrors the repository and some of the PostgreSQL developers use\nthis...\n\nLee.\n", "msg_date": "Tue, 10 Dec 2002 14:53:53 +0000", "msg_from": "Lee Kindness <lkindness@csl.co.uk>", "msg_from_op": false, "msg_subject": "Re: Let's create a release team " }, { "msg_contents": "On Tue, 10 Dec 2002, Tom Lane wrote:\n\n> \"Dan Langille\" <dan@langille.org> writes:\n> >> --- for example: Marc owns, runs, and pays for the\n> >> postgresql.org servers.\n>\n> > Is the cvs repo mirrored?\n>\n> Anyone running cvsup would have a complete copy of the source CVS,\n> I believe. It would be more troubling to reconstruct the mailing list\n> archives; I'm not sure that those are mirrored anywhere. (Marc?)\n\nArchives are mirrored at a number of sites. There was a time when all\nweb mirrors also mirrored them but that was split off about a year ago.\n\nVince.\n-- \n Fast, inexpensive internet service 56k and beyond! http://www.pop4.net/\n http://www.meanstreamradio.com http://www.unknown-artists.com\n Internet radio: It's not file sharing, it's just radio.\n\n", "msg_date": "Tue, 10 Dec 2002 10:09:18 -0500 (EST)", "msg_from": "Vince Vielhaber <vev@michvhf.com>", "msg_from_op": false, "msg_subject": "Re: Let's create a release team " }, { "msg_contents": "On Tuesday 10 December 2002 00:24, Justin Clift wrote:\n> RPM's & SRPM's\n\n> - Co-ordinate with Lamar to have these ready before the general\n> announcement?\n\nAs I am merely a volunteer in this, the availability of RPMs is directly \nimpacted by my workload. There are several times during the year that my \nworkload goes from being just difficult to absolutely swamping. These times \nare typically during mid February through early March; late August through \nlate September; and November through January.\n\nSee, not only am I the 'Chief Engineer' for several radio stations, but I am \nalso the 'IT Director' for WGCR, and the 'Network Administrator' for PARI. \nThe Chief Engineer duties include generator work, transmitter work, and \nstudio work -- and in winter there is alot of the generator/transmitter work \nin the mix. The 'IT Director' hat includes eradicating virus infections, \nunlicensed software, etc. This is currently my busiest area, as we try to \nput our entire fundraising system on our intranet (backed by PostgreSQL, of \ncourse). While I say 'we,' I really should say 'I,' as I am the entirety of \nthe programming team in this project. Fortunately I have access to an \ninterface design consultant and a good web designer.\n\nI was able to get the RPMs out when I did almost entirely due to the ice storm \nthat paralyzed the Carolinas last week -- our particular area did not get hit \nhard with ice, but got mostly snow, which then changed to mostly rain later \nin the day. So we didn't lose power -- and so I was able to get them done, \nsince I was unable to travel to work.\n\nTypically, I would try to track the betas and release candidates (like I did \nwith previous releases, to varying degrees), and with the 24 hour notice we \nall get on this list I can have a general release RPM ready. During this \ncycle I found myself excessively swamped by work -- so I was unable to \ngenerate RPM's until the general release. For that I apologize. I cannot \nguarantee that it won't happen again; but I will try to prevent its \nrecurrence.\n\nFor the 7.0 cycle, during the maintenance releases, I was retained by Great \nBridge to produce RPMs -- that ensured that I spent time on them for that \ncycle.\n-- \nLamar Owen\nWGCR Internet Radio\n1 Peter 4:11\n", "msg_date": "Tue, 10 Dec 2002 11:06:55 -0500", "msg_from": "Lamar Owen <lamar.owen@wgcr.org>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Let's create a release team" }, { "msg_contents": "Dan Langille wrote:\n> > But if you want to try to document the process better, there are some\n> > details written down already (eg, src/tools/RELEASE_CHANGES) and I'm\n> > sure Marc and Bruce would cooperate in writing down more.\n> \n> That's a good start. It looks like a list of things easily forgotten \n> but if forgotten, make us look bad.\n\nThere's not much I can add to that list. It is everything I normally\ncheck. Of course, Marc does a whole bunch of other things, but I am not\ninvolved in that.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 10 Dec 2002 13:58:02 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Let's create a release team" } ]
[ { "msg_contents": "\nThe notes below are the results of various tuning issues experienced \nrecently on a large database (several GB) that has many tables and a high \ntransient data flow (ie. thousands of records added, updated, and deleted \nevery hour) on a few tables. This kind of data flow is not at all well \nhandled by the default postgresql settings. Experiments have also been \nconducted using a much smaller test database with a text field written to a \nTOAST relation (which is what the large table contains).\n\nI think this example is useful because it encapsulates in several hours the \nlevel of updates that most of us see in several weeks, so the rules below \nshould apply equally well but in different time frames, with provisos as \nnoted. The database in question is subject to periodic bulk deletes where \nup to 50% of the rows in the large table are deleted. It is also has to run \n24x7.\n\nAny comments or suggestions would be welcome.\n\n\nTuning\n======\n\n1. max_fsm_relations\n--------------------\n\nFirst of all, the free space manager is useless at managing free space if \nit can not map all relations (including system relations and toast \nrelations). The following query should give the correct ballpark:\n\n select count(*) from pg_class where not relkind in ('i','v');\n\nSet max_fsm_relations to a number greater than this. Add extra to deal with \nany tables you will create etc. It costs 40 bytes per table, so be \ngenerous - if it is set too low, you will get bizarre space usage.\n\n[Note: the FSM is so bad at reclaiming space when this value is too low \nthat I believe it should be overridden at startup if it is not at least \nequal to the result of the above query. Similarly, I think a warning should \nbe given at startup and/or runtime when it is exceeded, or work should be \ndone to make it dynamic - and it should then not be a config item].\n\n\n2. VACUUM Frequency\n-------------------\n\nIdeally VACUUM should run constantly; a future version will support \nsomething like it. But for now, vacuum should be run when a significant \namount of data has been inserted, updated or deleted. The definition of \n'significant' is not immediately obvious.\n\nMost tables will *not* be updated frequently in most databases; such tables \ncan be vacuumed irregularly, or vacuumed when the more frequently updated \ntables are vacuumed.\n\nIn our specific case we have one table that has a few rows (< 1000), but it \nis updated as many as 3 times per second. In this case, we chose a 5 minute \ninterval, which results in at worst 1000 'dead' rows in the table as a \nresult of the updates. Since it was such a small table, we saw no reason to \nvacuum every minute, or even constantly.\n\nFor larger or more complex tables, the output of VACUUM ANALYZE must be used.\n\nThe following is an extract of the output from a VACUUM VERBOSE of a simple \ntest database - the table is the TOAST table of a large text column, where \nthe table has been constructed to be 75% empty. The output is after \ndeleting some rows.\n\n1 INFO: --Relation pg_toast.pg_toast_16979--\n2 INFO: Index pg_toast_16979_index: Pages 575; Tuples 16384: Deleted 25984.\n3 CPU 0.05s/0.16u sec elapsed 7.41 sec.\n4 INFO: Removed 25984 tuples in 6496 pages.\n5 CPU 0.75s/0.79u sec elapsed 14.17 sec.\n6 INFO: Pages 22480: Changed 6496, Empty 0; Tup 16384: Vac 25984, Keep 0, \nUnUsed 47552.\n7 Total CPU 1.98s/1.05u sec elapsed 23.30 sec.\n\nLine 6 shows that there are 22480 pages, and 6496 (roughly 25%) were \nchanged since the last vacuum. Line 4 indicates that these were all \nremoved. Note that when tuples are updated, a new copy of the record is \nwritten and the old one deleted, so updates will also result in tuples \nbeing 'removed'.\n\nA more complex example follows; this was after deleting 512 rows and adding \n256:\n\n1 INFO: --Relation pg_toast.pg_toast_16979--\n2 INFO: Index pg_toast_16979_index: Pages 667; Tuples 24576: Deleted 16384.\n3 CPU 0.02s/0.10u sec elapsed 4.73 sec.\n4 INFO: Removed 16384 tuples in 4096 pages.\n5 CPU 0.52s/0.48u sec elapsed 9.38 sec.\n6 INFO: Pages 20528: Changed 6144, Empty 0; Tup 24576: Vac 16384, Keep 0, \nUnUsed 41152.\n7 Total CPU 1.81s/0.64u sec elapsed 22.51 sec.\n\nnote that line 6 has a 'changed' value, and line 4 has a 'removed' value. \nThis gives some indication of the pages consumed and released in any period.\n\nThe final example is for 512 inserts, 512 updates (of different records) \nand 512 deletes.\n\n1 INFO: --Relation pg_toast.pg_toast_16979--\n2 INFO: Index pg_toast_16979_index: Pages 854; Tuples 32768: Deleted 32768.\n3 CPU 0.05s/0.20u sec elapsed 8.41 sec.\n4 INFO: Removed 32768 tuples in 8192 pages.\n5 CPU 1.01s/0.91u sec elapsed 13.52 sec.\n6 INFO: Pages 26672: Changed 12288, Empty 0; Tup 32768: Vac 32768, Keep 0, \nUnUsed 41152.\n7 Total CPU 2.92s/1.25u sec elapsed 30.01 sec.\n\nagain it shows the effects of UPDATE/DELETE vs. INSERT.\n\nIn each case the 'Changed' value indicates the maximum number of pages \nrequired between vaccuums; and the 'removed' values indicates that some \npages will be added to the FSM when vacuum is run. The high 'unused' value \nshows the results of an earleir bulk delete.\n\nNote that 'Changed' seems to be 'updates+deletes+inserts' whereas 'removed' \nis 'deletes+updates', so it is not possible to determine 'updates+inserts', \nwhich would be the best indicator of the required number of new pages. If \nnecessary, it *could* be derived by looking at tuple counts across vacuums, \nbut using the 'changed' figure will give a good upper limit, since in most \nmeaningful cases deletes will be lower than inserts -- hence it will be out \nby at worst a factor of two.\n\nWe have chosen (arbitrarily) to keep the number of pages 'changed' below \n25% of the total number of pages; in the above case that would involve \nrunning VACUUM twice as often.\n\n\n3. max_fsm_pages\n----------------\n\nContrary to other advice, FOR TABLES THAT ARE SUBJECT PERIODIC TO BULK \nDELETES, this figure should not be set based on how many pages you delete \nbetween vacuums, but how many pages you will consume between vacuums (the \n'changed' figure, above). This difference is important for sites that purge \ndata periodically: it does not need to be set high enough to hold all \ndeleted pages for one bulk delete -- it should be set high enough to have \nsufficient pages to manage the consumption between vacuums. If the pool is \nexhausted, then vacuum will find more 'unused' pages the next time it is \nrun. UPDATE & INSERT will consume free pages, DELETE will create free pages.\n\nFor tables that are more typical (insert-intensive, or a general mix of \nupdates, deletes and inserts), the 'removed' value should be used.\n\n====\nNOTE: max_fsm_pages is a cumulative value. It should be set to the sum of \nthese values for all tables.\n====\n\nBased on the example above, VACUUM should probably be run more frequently. \nHowever, if the above was typical of the chosen VACUUM frequency, then \nmax_fsm_pages should be set to at least 12288 for this table, despite the \nfact that only 8192 pages were released (since we assume there may be bulk \ndeletes freeing up many pages).\n\nIf VACUUM were run twice as often, the max_fsm_pages should be set to at \nleast 6144 for this table.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Mon, 09 Dec 2002 01:41:31 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "DB Tuning Notes for comment..." }, { "msg_contents": "On Sun, 2002-12-08 at 09:41, Philip Warner wrote:\n> Any comments or suggestions would be welcome.\n> \n\nfirst and foremost, this is really excellent work! We need to look into\ngetting this info into the standard documentation and/or Bruce's tuning\nguide.\n\n> \n> Tuning\n> ======\n> \n> 1. max_fsm_relations\n> --------------------\n> \n> First of all, the free space manager is useless at managing free space if \n> it can not map all relations (including system relations and toast \n> relations). The following query should give the correct ballpark:\n> \n> select count(*) from pg_class where not relkind in ('i','v');\n> \n\nIt should be noted that if you have multiple databases, you'll need to\nget the number of relations minus the system tables for each database,\nthen add in the number of system tables.\n\n> Set max_fsm_relations to a number greater than this. Add extra to deal with \n> any tables you will create etc. It costs 40 bytes per table, so be \n> generous - if it is set too low, you will get bizarre space usage.\n> \n> \n> 2. VACUUM Frequency\n> -------------------\n> \n> Ideally VACUUM should run constantly; a future version will support \n> something like it. But for now, vacuum should be run when a significant \n> amount of data has been inserted, updated or deleted. The definition of \n> 'significant' is not immediately obvious.\n>\n\nI don't think this is entirely true. On tables that have large numbers\nof inserts, but no updates or deletes, you do not need to run vacuum. It\nmight be helpful to run analyze on these tables if your inserting enough\ndata to change the statistical relationships, but vacuum itself is not\nneeded. \n \n> Most tables will *not* be updated frequently in most databases; such tables \n> can be vacuumed irregularly, or vacuumed when the more frequently updated \n> tables are vacuumed.\n> \n> In our specific case we have one table that has a few rows (< 1000), but it \n> is updated as many as 3 times per second. In this case, we chose a 5 minute \n> interval, which results in at worst 1000 'dead' rows in the table as a \n> result of the updates. Since it was such a small table, we saw no reason to \n> vacuum every minute, or even constantly.\n\nI have some similar tables in my system, with between 250 and 3500 rows.\nThese tables turn over at least every 15 minutes, so I have decided on a\n10 minute vacuum interval. As with Phillip's, since they are small\ntables, more frequent vacuuming seemed excessive. \n\n> \n> For larger or more complex tables, the output of VACUUM ANALYZE must be used.\n> \n<snip>\n\nagain, great work Philip.\n\nRobert Treat\n\n", "msg_date": "09 Dec 2002 14:46:30 -0500", "msg_from": "Robert Treat <rtreat@webmd.net>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes for comment..." }, { "msg_contents": "Robert Treat wrote:\n> On Sun, 2002-12-08 at 09:41, Philip Warner wrote:\n> \n>>Any comments or suggestions would be welcome.\n>>\n> \n> \n> first and foremost, this is really excellent work! We need to look into\n> getting this info into the standard documentation and/or Bruce's tuning\n> guide.\n> \n\nSeconded!\n\n> \n>>Tuning\n>>======\n>>\n>>1. max_fsm_relations\n>>--------------------\n>>\n>>First of all, the free space manager is useless at managing free space if \n>>it can not map all relations (including system relations and toast \n>>relations). The following query should give the correct ballpark:\n>>\n>> select count(*) from pg_class where not relkind in ('i','v');\n>>\n> \n> \n> It should be noted that if you have multiple databases, you'll need to\n> get the number of relations minus the system tables for each database,\n> then add in the number of system tables.\n> \n> \n>>Set max_fsm_relations to a number greater than this. Add extra to deal with \n>>any tables you will create etc. It costs 40 bytes per table, so be \n>>generous - if it is set too low, you will get bizarre space usage.\n>>\n>>\n>>2. VACUUM Frequency\n>>-------------------\n>>\n>>Ideally VACUUM should run constantly; a future version will support \n>>something like it. But for now, vacuum should be run when a significant \n>>amount of data has been inserted, updated or deleted. The definition of \n>>'significant' is not immediately obvious.\n>>\n> \n> \n> I don't think this is entirely true. On tables that have large numbers\n> of inserts, but no updates or deletes, you do not need to run vacuum. It\n> might be helpful to run analyze on these tables if your inserting enough\n> data to change the statistical relationships, but vacuum itself is not\n> needed. \n> \n\nIn my experience I've seen tables with numerous indexes continue to \nbenefit greatly from vacuum/vacuum full operations when large volumes of \ninserts are performed. This is true even when the update/delete activity \non the base table itself is manageable. While dropping and recreating \nthe index after loading is possible in some cases, my general comment is \nthat index maintenance is an issue you should keep in mind when \ndesigning your vacuum strategy.\n\n> \n>>Most tables will *not* be updated frequently in most databases; such tables \n>>can be vacuumed irregularly, or vacuumed when the more frequently updated \n>>tables are vacuumed.\n>>\n>>In our specific case we have one table that has a few rows (< 1000), but it \n>>is updated as many as 3 times per second. In this case, we chose a 5 minute \n>>interval, which results in at worst 1000 'dead' rows in the table as a \n>>result of the updates. Since it was such a small table, we saw no reason to \n>>vacuum every minute, or even constantly.\n> \n> \n> I have some similar tables in my system, with between 250 and 3500 rows.\n> These tables turn over at least every 15 minutes, so I have decided on a\n> 10 minute vacuum interval. As with Phillip's, since they are small\n> tables, more frequent vacuuming seemed excessive. \n> \n> \n>>For larger or more complex tables, the output of VACUUM ANALYZE must be used.\n>>\n> \n> <snip>\n> \n> again, great work Philip.\n> \n> Robert Treat\n> \n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 5: Have you checked our extensive FAQ?\n> \n> http://www.postgresql.org/users-lounge/docs/faq.html\n> \n\n\nss\n\n\nScott Shattuck\nTechnical Pursuit Inc.\n\n\n\n", "msg_date": "Mon, 09 Dec 2002 13:18:47 -0700", "msg_from": "Scott Shattuck <ss@technicalpursuit.com>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes for comment..." }, { "msg_contents": "Robert Treat <rtreat@webmd.net> writes:\n> On Sun, 2002-12-08 at 09:41, Philip Warner wrote:\n>> First of all, the free space manager is useless at managing free space if \n>> it can not map all relations (including system relations and toast \n>> relations). The following query should give the correct ballpark:\n>> \n>> select count(*) from pg_class where not relkind in ('i','v');\n\nFSM entries aren't needed for sequences either, so more correct is\n\n\tselect count(*) from pg_class where relkind in ('r', 't');\n\n> It should be noted that if you have multiple databases, you'll need to\n> get the number of relations minus the system tables for each database,\n> then add in the number of system tables.\n\nYou're assuming that system tables are shared, which they mostly aren't.\nSumming the pg_class count over all databases (or all that get vacuumed,\nanyway; you can exclude template0) will be close enough.\n\nBTW, this neglects what seems possibly an important factor: you don't\nneed FSM entries for tables that are effectively read-only or insert-only\n(no deletes or updates). At least in some database designs, that's a\nsignificant number of tables.\n\nHowever, I suspect that the present FSM code is not very effective at\ndeciding *which* tables to track if it has too few slots, so Philip's\nadvice of \"make sure there's one for every table\" may be the best in the\nnear term. But we need to work at improving that logic.\n\nI have some uncommitted patches concerning the FSM management heuristics\nfrom Stephen Marshall, which I deemed too late/risky for 7.3, but we\nshould get something done for 7.4. Anyone interested in playing around\nin this area?\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 15:54:20 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes for comment... " }, { "msg_contents": "Scott Shattuck <ss@technicalpursuit.com> writes:\n> Robert Treat wrote:\n>> I don't think this is entirely true. On tables that have large numbers\n>> of inserts, but no updates or deletes, you do not need to run vacuum.\n\n> In my experience I've seen tables with numerous indexes continue to \n> benefit greatly from vacuum/vacuum full operations when large volumes of \n> inserts are performed. This is true even when the update/delete activity \n> on the base table itself is manageable.\n\nThis is hard to believe, as VACUUM does not even touch the indexes\nunless it has found deletable tuples --- and I am quite certain that\nbtree indexes, at least, do not do any VACUUM-time reorganization beyond\ndeleting deletable entries. (I wouldn't swear to it one way or the\nother for GiST though.) Robert's opinion coincides with what I know of\nthe code.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 16:10:27 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes for comment... " }, { "msg_contents": "Tom Lane wrote:\n> Scott Shattuck <ss@technicalpursuit.com> writes:\n> \n>>Robert Treat wrote:\n>>\n>>>I don't think this is entirely true. On tables that have large numbers\n>>>of inserts, but no updates or deletes, you do not need to run vacuum.\n>>\n> \n>>In my experience I've seen tables with numerous indexes continue to \n>>benefit greatly from vacuum/vacuum full operations when large volumes of \n>>inserts are performed. This is true even when the update/delete activity \n>>on the base table itself is manageable.\n> \n> \n> This is hard to believe, as VACUUM does not even touch the indexes\n> unless it has found deletable tuples --- and I am quite certain that\n> btree indexes, at least, do not do any VACUUM-time reorganization beyond\n> deleting deletable entries. (I wouldn't swear to it one way or the\n> other for GiST though.) Robert's opinion coincides with what I know of\n> the code.\n> \n\nWilling to learn here but skipping a vacuum full has caused some issues \nfor us. Here's some data from a recent 3 day test run that was done with \nregular vacuums but not vacuum fulls. When running with vacuum full the \nindexes remain in line:\n\nnsuite-10=# select relname, relpages, reltuples from pg_class where \nrelname not like 'pg_%' order by reltuples desc;\n-[ RECORD 1 ]------------------------------\nrelname | directory_fullpath_ix\nrelpages | 96012\nreltuples | 1.38114e+06\n-[ RECORD 2 ]------------------------------\nrelname | directory_pkey\nrelpages | 16618\nreltuples | 1.38114e+06\n-[ RECORD 3 ]------------------------------\nrelname | directory\nrelpages | 23924\nreltuples | 59578\n<snip>\n\nNeedless to say, the system performance was pathetic but the test did \nserve to highlight this index issue.\n\nAnyone want to give a quick summary of index maintenance or give me a \npointer into the codebase where someone who's not a C expert might still \nget a sense of what's being done? I'd really like to understand how an \nindex can get so completely out of whack after a weekend of testing.\n\nIt seems you're telling me that the data here \"proves\" there's an update \nor delete going on somewhere in the system, even though this test is of \na database initialization driven by a stored procedure with no update or \ndelete operations targeting the directory table. There may be some \noperations being done external to that process that I've not been made \naware of but I'm still curious to learn more about indexing behavior so \nI know why something like this happens in the first place.\n\n\nss\n\n\n\n\n", "msg_date": "Mon, 09 Dec 2002 14:56:17 -0700", "msg_from": "Scott Shattuck <ss@technicalpursuit.com>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes for comment..." }, { "msg_contents": "Scott Shattuck <ss@technicalpursuit.com> writes:\n> Willing to learn here but skipping a vacuum full has caused some issues \n> for us. Here's some data from a recent 3 day test run that was done with \n> regular vacuums but not vacuum fulls. When running with vacuum full the \n> indexes remain in line:\n\n> nsuite-10=# select relname, relpages, reltuples from pg_class where \n> relname not like 'pg_%' order by reltuples desc;\n> -[ RECORD 1 ]------------------------------\n> relname | directory_fullpath_ix\n> relpages | 96012\n> reltuples | 1.38114e+06\n> -[ RECORD 2 ]------------------------------\n> relname | directory_pkey\n> relpages | 16618\n> reltuples | 1.38114e+06\n> -[ RECORD 3 ]------------------------------\n> relname | directory\n> relpages | 23924\n> reltuples | 59578\n> <snip>\n\n<<blink>> There's no way that the index and table tuple counts should\nget that far out of line; in the absence of any concurrent updates,\nthey should be *equal* (or index < table, if you have a partial index,\nwhich I assume these are not). I would credit the recorded index count\nexceeding the recorded table count by the number of tuples inserted/\nupdated while a (plain) VACUUM is in process on that table --- but this\ndoesn't look like it meets that situation.\n\nThere was a bug a long time ago wherein vacuum would forget to update\npg_class.reltuples for indexes in some cases, but according to the CVS\nlogs that was fixed before 7.2 release. What version are you running\nexactly?\n\nIn any case, you seem to be dealing with some kind of bug here. It\nmight be helpful to look at the output of \"vacuum verbose directory\"\nif you still have it available.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 18:19:57 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes for comment... " }, { "msg_contents": "At 03:54 PM 9/12/2002 -0500, Tom Lane wrote:\n>However, I suspect that the present FSM code is not very effective at\n>deciding *which* tables to track if it has too few slots,\n\nYou are definitely right there.\n\nI think it would be worth looking at removing max_fsm_tables as a tuning \noption, and adding a 'relhasfsm' flag to pg_class for those tables that \nshould not be mapped. Default to 't'. Then, make the table grow dynamically \nas tables are added, or when a VACUUM occurs...\n\nAFAICT, the only justification for a smaller list of relations is for those \nthat are *almost never* subject to deletes or updates. They are certainly \ncommon in DB design, but I'd let the DBA designate them.\n\nDoes this sound reasonable?\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Tue, 10 Dec 2002 10:40:19 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes for comment... " }, { "msg_contents": "At 02:46 PM 9/12/2002 -0500, Robert Treat wrote:\n>getting this info into the standard documentation and/or Bruce's tuning\n>guide.\n\nI'd vote for the standard docs since it is sufficiently basic as to be \nneeded by most users. We either need a tuning chapter or a new section in \nruntime configuration.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Tue, 10 Dec 2002 10:47:28 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes for comment..." }, { "msg_contents": "Philip Warner <pjw@rhyme.com.au> writes:\n> I think it would be worth looking at removing max_fsm_tables as a tuning \n> option, and adding a 'relhasfsm' flag to pg_class for those tables that \n> should not be mapped. Default to 't'. Then, make the table grow dynamically \n> as tables are added, or when a VACUUM occurs...\n\nIf we could \"make the table grow dynamically\" then there'd not be much\nneed for the config parameters at all. The real problem is to fit into\na shmem segment whose size has to be frozen at postmaster start (which,\nnot incidentally, is before we've ever looked at the database...). We\ncould make the constraint be on total space for relation entries + page\nentries rather than either individually, but I think that'd mostly make\nit harder to interpret the config setting rather than offer any real\nease of administration.\n\n> AFAICT, the only justification for a smaller list of relations is for those \n> that are *almost never* subject to deletes or updates. They are certainly \n> common in DB design, but I'd let the DBA designate them.\n\nIt doesn't seem to me to be that hard for the system to recognize them\nautomatically. Basically, if there are no holes of useful size in the\ntable, there's no need to create an FSM entry for it. The trick is\n\"useful size\" here --- but VACUUM already does the work needed to\nestimate an average tuple size, so I'd think it could do a reasonably\ngood job of realizing that all the available holes are just leftover\nspace. (The relation's very last page is also a special case that's\nlikely not special-cased correctly at the moment: perhaps it should never\nbe entered in FSM at all, certainly not if it's the only page that would\nbe entered.)\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 19:01:28 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes for comment... " }, { "msg_contents": "At 03:54 PM 9/12/2002 -0500, Tom Lane wrote:\n>FSM entries aren't needed for sequences either, so more correct is\n>\n> select count(*) from pg_class where relkind in ('r', 't');\n\npresumably:\n\n select count(*) from pg_class where relkind in ('r', 't', 'i');\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Tue, 10 Dec 2002 12:00:14 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes for comment... " }, { "msg_contents": "Philip Warner <pjw@rhyme.com.au> writes:\n> At 03:54 PM 9/12/2002 -0500, Tom Lane wrote:\n>> FSM entries aren't needed for sequences either, so more correct is\n>> \n>> select count(*) from pg_class where relkind in ('r', 't');\n\n> presumably:\n\n> select count(*) from pg_class where relkind in ('r', 't', 'i');\n\nNo, I meant what I said. Indexes don't use the FSM. (The premise of\nthe FSM is that one bit of free space in a table is as good as any other\nbit; a premise quite incorrect for indexes.)\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 20:15:30 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes for comment... " }, { "msg_contents": "At 07:01 PM 9/12/2002 -0500, Tom Lane wrote:\n>We\n>could make the constraint be on total space for relation entries + page\n>entries rather than either individually, but I think that'd mostly make\n>it harder to interpret the config setting rather than offer any real\n>ease of administration.\n\nPerhaps doing both? And issue a warning to the logs when max_fsm_relations \nis borrowing from max_fsm_pages.\n\nIt might be that the outstanding patches address the problem, but at the \nmoment the choice of which relations to include is not well made when \nmax_fsm_relations of much too small. We should at least issue a warning; \nbut allowing max_fsm_relations to borrow from max_fsm_pages seems like a \ngood idea, since having the number too low (with 161 relations setting it \nto the default of 100) is useless.\n\nSecondly, an empty database contains 98 tables, so the default setting of \nmax_fsm_pages to 100 is way too low.\n\nThe tradeoff of losing 7 pages from the map to include another relation is \nworth it, especially if the logs contain a warning.\n\nBut perhaps the test itself is flawed and there is another problem \nresulting in this behaviour (doing vacuums twice in a row seems to make it \nuse the free space, but I'd guess this is just edge behaviour of the FSM \nheuristics):\n\nCreate Table t(i serial, t text);\ninsert into t(t) .... 47K of UUEncoded jpeg file -> ~47K of toast.\ninsert into t(t) select t from t;\n...repeat 9 times...\ncreate table t1 as select * from t limit 1;\n...\ncreate table t19 as select * from t limit 1;\ncreate table t20(i serial, t text);\ninsert into t20(t) select t from t;\n\nie. build a lot of tables, with two big ones separated by OID (not sure if \nthe last part is relevant).\n\nselect count(*) from pg_class where relkind in ('t','r');\n\nin my case this resulted in 161, so I set max_fsm_relations to 100 (ie. not \na borderline case, but the default setting).\n\nI also left max_fsm_pages at 10000 so that we should have space for several \nthousand rows.\n\nStop & start postmaster, then vacuum full to be comfortable no other \nproblems occur, an look at file sizes of relation file and toast file.\n\nNow:\n\ndelete from t where i <= 128;\ndelete from t20 where i <= 128;\n\nvacuum;\n\ncheck file sizes - no surprises, they should be unchanged.\n\nTue Dec 10 12:03:53 EST 2002\n-rw------- 1 pjw users 65536 2002-12-10 12:03 16979\n-rw------- 1 pjw users 65536 2002-12-10 12:03 33432\n-rw------- 1 pjw users 67108864 2002-12-10 12:03 16982\n-rw------- 1 pjw users 67108864 2002-12-10 12:03 33435\n\nthen do:\n\ninsert into t(t) select t from t20 limit 10;\ninsert into t20(t) select t from t limit 10;\n\nand both files have grown:\n\nTue Dec 10 12:08:20 EST 2002\n-rw------- 1 pjw users 65536 2002-12-10 12:08 33432\n-rw------- 1 pjw users 67764224 2002-12-10 12:08 33435\n-rw------- 1 pjw users 67764224 2002-12-10 12:08 16982\n-rw------- 1 pjw users 65536 2002-12-10 12:08 16979\n\noddly (bug? edge behaviour?) doing two vacuums in a row results in the free \nspace being used.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Tue, 10 Dec 2002 12:17:48 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes for comment... " }, { "msg_contents": "At 12:17 PM 10/12/2002 +1100, Philip Warner wrote:\n>Secondly, an empty database contains 98 tables,\n\nCorrected based on Tom's later mail; from the FSM PoV, it contains 37 \n(indices don't count). So it is exhausted when more than two DBs are created.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Tue, 10 Dec 2002 12:29:35 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes for comment... " }, { "msg_contents": "Tom Lane wrote:\n> Scott Shattuck <ss@technicalpursuit.com> writes:\n> \n>>Willing to learn here but skipping a vacuum full has caused some issues \n>>for us. Here's some data from a recent 3 day test run that was done with \n>>regular vacuums but not vacuum fulls. When running with vacuum full the \n>>indexes remain in line:\n> \n> \n>>nsuite-10=# select relname, relpages, reltuples from pg_class where \n>>relname not like 'pg_%' order by reltuples desc;\n>>-[ RECORD 1 ]------------------------------\n>>relname | directory_fullpath_ix\n>>relpages | 96012\n>>reltuples | 1.38114e+06\n>>-[ RECORD 2 ]------------------------------\n>>relname | directory_pkey\n>>relpages | 16618\n>>reltuples | 1.38114e+06\n>>-[ RECORD 3 ]------------------------------\n>>relname | directory\n>>relpages | 23924\n>>reltuples | 59578\n>><snip>\n> \n> \n> <<blink>> There's no way that the index and table tuple counts should\n> get that far out of line; in the absence of any concurrent updates,\n> they should be *equal* (or index < table, if you have a partial index,\n> which I assume these are not). I would credit the recorded index count\n> exceeding the recorded table count by the number of tuples inserted/\n> updated while a (plain) VACUUM is in process on that table --- but this\n> doesn't look like it meets that situation.\n> \n> There was a bug a long time ago wherein vacuum would forget to update\n> pg_class.reltuples for indexes in some cases, but according to the CVS\n> logs that was fixed before 7.2 release. What version are you running\n> exactly?\n\ntest=# select version();\n version\n-------------------------------------------------------------\n PostgreSQL 7.2.2 on i686-pc-linux-gnu, compiled by GCC 2.96\n(1 row)\n\ntest=#\n\n> \n> In any case, you seem to be dealing with some kind of bug here. It\n> might be helpful to look at the output of \"vacuum verbose directory\"\n> if you still have it available.\n> \n\n\nNOTICE: --Relation directory--\nNOTICE: Index directory_pkey: Pages 15628; Tuples 4988848: Deleted 35407.\n CPU 0.73s/3.00u sec elapsed 40.53 sec.\n\nNOTICE: Index directory_fullpath_ix: Pages 80808; Tuples 4989317: \nDeleted 35407.\n CPU 4.84s/3.91u sec elapsed 275.66 sec.\nNOTICE: Removed 35407 tuples in 786 pages.\n CPU 0.13s/0.11u sec elapsed 1.80 sec.\nNOTICE: Pages 80156: Changed 18, Empty 0; Tup 4988787: Vac 35407, Keep \n4977704, UnUsed 348422.\n Total CPU 7.85s/7.58u sec elapsed 343.84 sec.\n\n\n> \t\t\tregards, tom lane\n> \n\nThanks for any insight you can offer here.\n\n\nss\n\n\n", "msg_date": "Mon, 09 Dec 2002 18:34:01 -0700", "msg_from": "Scott Shattuck <ss@technicalpursuit.com>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes for comment..." }, { "msg_contents": "Philip Warner <pjw@rhyme.com.au> writes:\n> Secondly, an empty database contains 98 tables, so the default setting of \n> max_fsm_pages to 100 is way too low.\n\nOnly 37 of them need FSM entries, but still a good point; we should\nprobably bump it up to 1000 to be more realistic.\n\n> oddly (bug? edge behaviour?) doing two vacuums in a row results in the free \n> space being used.\n\nI'm on my way out the door, so no time to think about what's actually\nhappening in the current code, but ideally I would think that when the\nFSM doesn't have enough space, it should prefer to remember info about\nrels with heavy update activity (which might be approximated by \"rels\nwith lots of free space\", but isn't really the same thing). A VACUUM\ndone just after startup does not have any historical info to base this\ndecision on. So it's not unreasonable for the system to make better\nchoices after it's been running awhile than when it's freshly booted.\nI'm not sure that this is actually what's happening today, just pointing\nout that I don't consider it a bug per se if the code behaves that way.\n(The existing code does have some LRU effects, IIRC, but not sure\nif they account for what you see.)\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 20:39:28 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes for comment... " }, { "msg_contents": "At 08:39 PM 9/12/2002 -0500, Tom Lane wrote:\n>A VACUUM\n>done just after startup does not have any historical info to base this\n>decision on.\n\nThe actual order is:\n\nstart\ndelete\nvacuum;\ninsert <- does not use free space\nvacuum;\ninsert <- does not use free space\nvacuum;\nvacuum;\ninsert <- uses free space\n\nmy guess is the fact that the second vacuum has no useful statistics means \nthat (somehow) it is choosing the tables with free space -- perhaps because \nthey are the largest?\n\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Tue, 10 Dec 2002 12:47:03 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes for comment... " }, { "msg_contents": "On Mon, 09 Dec 2002 19:10:23 -0500, Tom Lane wrote:\n\n> Philip Warner <pjw@rhyme.com.au> writes:\n>> I think it would be worth looking at removing max_fsm_tables as a\n>> tuning option, and adding a 'relhasfsm' flag to pg_class for those\n>> tables that should not be mapped. Default to 't'. Then, make the table\n>> grow dynamically as tables are added, or when a VACUUM occurs...\n> \n> If we could \"make the table grow dynamically\" then there'd not be much\n> need for the config parameters at all. The real problem is to fit into\n> a shmem segment whose size has to be frozen at postmaster start (which,\n> not incidentally, is before we've ever looked at the database...). We\n> could make the constraint be on total space for relation entries + page\n> entries rather than either individually, but I think that'd mostly make\n> it harder to interpret the config setting rather than offer any real\n> ease of administration.\n> \n\nCan we not just have vacuum of a database return a total # of pages\nmodified and relations modified, and then report suggested free space map\nsettings? Even this little bit would be a step in the right direction.\n\nRobert Treat\n", "msg_date": "Mon, 09 Dec 2002 21:10:38 -0500", "msg_from": "Robert Treat <xzilla@users.sourceforge.net>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes for comment..." }, { "msg_contents": "At 09:10 PM 9/12/2002 -0500, Robert Treat wrote:\n>Even this little bit would be a step in the right direction.\n\nWhat I would find really useful is a 'VACUUM...WITH HISORY' which wrote the \nunderlying details of VACUUM VERBOSE to a 'pg_vacuum_history' table.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Tue, 10 Dec 2002 14:10:58 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes for comment..." }, { "msg_contents": "At 03:54 PM 9/12/2002 -0500, Tom Lane wrote:\n>I have some uncommitted patches concerning the FSM management heuristics\n>from Stephen Marshall, which I deemed too late/risky for 7.3, but we\n>should get something done for 7.4. Anyone interested in playing around\n>in this area?\n\nI'd be interested in seeing the patches, but can't commit to doing anything \nwith them at this point. I would like to get to the bottom of the weird \nbehaviour, however.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Wed, 11 Dec 2002 16:17:11 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes for comment... " } ]
[ { "msg_contents": "Hi,\n\nI have a problem with contrib/tsearch module. \n\nSimple select txt2txtidx('2-3') causes psql to crash with error:\n\nserver closed the connection unexpectedly\n This probably means the server terminated abnormally\n before or while processing the request.\n\n\nAs I found out, this happens allways when in txt2txtidx input parameter occurs string with pattern: \"number hyphen\" or \"hyphen number\" !!!\n\npg 7.3 \nredhat 7.3\n\nThanks, any idea ?\n\n\n\n-----------------------------------------\nHot Mobiil - helinad, logod ja pilts�numid!\nhttp://www.hot.ee\n", "msg_date": "Sun, 8 Dec 2002 17:05:19 +0200", "msg_from": "Argo Priivits <listdb@hot.ee>", "msg_from_op": true, "msg_subject": "7.3 txt2txtidx -> crash" }, { "msg_contents": "Yes i found this bug earlier.\nThere is a patch for it in the mail archives.\n\nMagnus\n\nArgo Priivits <listdb@hot.ee> wrote:\n> Hi,\n> \n> I have a problem with contrib/tsearch module.\n> \n> Simple select txt2txtidx('2-3') causes psql to crash with error:\n> \n> server closed the connection unexpectedly\n> This probably means the server terminated abnormally\n> before or while processing the request.\n> \n> \n> As I found out, this happens allways when in txt2txtidx input\n> parameter occurs string with pattern: \"number hyphen\" or \"hyphen\n> number\" !!! \n> \n> pg 7.3\n> redhat 7.3\n> \n> Thanks, any idea ?\n> \n\n", "msg_date": "Sun, 8 Dec 2002 16:44:02 +0100", "msg_from": "\"Magnus Naeslund(f)\" <mag@fbab.net>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] 7.3 txt2txtidx -> crash" }, { "msg_contents": "Argo,\n\nthis is known bug, please find patch from\nhttp://www.sai.msu.su/~megera/postgres/gist/\n\n Regards,\n\n Oleg\nOn Sun, 8 Dec 2002, Argo Priivits wrote:\n\n> Hi,\n>\n> I have a problem with contrib/tsearch module.\n>\n> Simple select txt2txtidx('2-3') causes psql to crash with error:\n>\n> server closed the connection unexpectedly\n> This probably means the server terminated abnormally\n> before or while processing the request.\n>\n>\n> As I found out, this happens allways when in txt2txtidx input parameter occurs string with pattern: \"number hyphen\" or \"hyphen number\" !!!\n>\n> pg 7.3\n> redhat 7.3\n>\n> Thanks, any idea ?\n>\n>\n>\n> -----------------------------------------\n> Hot Mobiil - helinad, logod ja piltsО©╫numid!\n> http://www.hot.ee\n>\n> ---------------------------(end of broadcast)---------------------------\n> TIP 5: Have you checked our extensive FAQ?\n>\n> http://www.postgresql.org/users-lounge/docs/faq.html\n>\n\n\tRegards,\n\t\tOleg\n_____________________________________________________________\nOleg Bartunov, sci.researcher, hostmaster of AstroNet,\nSternberg Astronomical Institute, Moscow University (Russia)\nInternet: oleg@sai.msu.su, http://www.sai.msu.su/~megera/\nphone: +007(095)939-16-83, +007(095)939-23-83\n\n", "msg_date": "Mon, 9 Dec 2002 00:09:39 +0300 (MSK)", "msg_from": "Oleg Bartunov <oleg@sai.msu.su>", "msg_from_op": false, "msg_subject": "Re: 7.3 txt2txtidx -> crash" } ]
[ { "msg_contents": "I'm working on the TODO item \"Allow easy display of usernames in a group\" in \nthe context of a slightly larger effort to improve usability of arrays. I'm \nfar enough down the road to have a better idea of where I want to go with \nthis, but I'd like to vet those ideas with the list so I don't waste too much \neffort if everyone hates them ;-)\n\nThe first function borrows from an idea Nigel Andrews had -- i.e. expand an \narray into rows (and possibly columns). It currently works like this:\n\n-- 1D array\ntest=# select * from array_values('{101,102,103,104}'::int[]) as (a int, b1 int);\n a | b1\n---+-----\n 1 | 101\n 2 | 102\n 3 | 103\n 4 | 104\n(4 rows)\n\nCREATE TABLE arr_text(f1 int, f2 text[]);\nINSERT INTO arr_text VALUES (1, '{\"a\",\"b\",\"c\"}');\nUPDATE arr_text SET f2[-2:0] = '{\"x\",\"y\",\"z\"}' WHERE f1 = 1;\nCREATE OR REPLACE FUNCTION get_arr_text(int) RETURNS text[] AS 'SELECT f2 FROM \narr_text WHERE f1 = $1' LANGUAGE 'sql';\n\ntest=# select * from array_values(get_arr_text(1)) as (a int, b1 text);\n a | b1\n----+----\n -2 | x\n -1 | y\n 0 | z\n 1 | a\n 2 | b\n 3 | c\n(6 rows)\n\n-- 2D array\ntest=# select * from array_values('{{1,2,3,4},{5,6,7,8}}'::int[]) as (a int, \nb1 int, b2 int, b3 int, b4 int);\n a | b1 | b2 | b3 | b4\n---+----+----+----+----\n 1 | 1 | 2 | 3 | 4\n 2 | 5 | 6 | 7 | 8\n(2 rows)\n\nIt accepts type anyarray, and returns record. The first column preserves the \narray subscript for the 1st dimension.\n\nOne question I have is this: what, if anything, should be done with 3 (and \nhigher) dimension arrays? I was considering returning 2 columns -- the 1st \ndimension array subscript, and a 2nd column containing the sub-array left \nover. E.g.:\n\narray_values('{{{111,112},{121,122}},{{211,212},{221,222}}}'::int[]) would become:\n\n a | b1\n----+-----------------------\n 1 | {{111,112},{121,122}}\n 2 | {{211,212},{221,222}}\n\nDoes this make sense, or is something else better, or would it be better not \nto support 3 dim arrays and up?\n\n\nNow on to the TODO item. Given the array_values() function, here's what I was \nthinking of to implement listing members of a group:\n\nCREATE OR REPLACE FUNCTION pg_get_grolist(text) RETURNS INT[] AS 'SELECT \ngrolist FROM pg_group WHERE groname = $1' LANGUAGE 'sql';\n\nCREATE TYPE pg_grolist_rec AS (array_index int, member_name text);\n\nCREATE OR REPLACE FUNCTION group_list(text) RETURNS SETOF pg_grolist_rec AS \n'SELECT g.id, pg_get_userbyid(g.usesysid)::text AS member_name FROM \narray_values(pg_get_grolist($1)) AS g(id int, usesysid int)' LANGUAGE 'sql';\n\ntest=# select * from pg_group;\n groname | grosysid | grolist\n---------+----------+---------------\n g1 | 100 | {100,101}\n g2 | 101 | {100,101,102}\n(2 rows)\n\ntest=# select * from group_list('g2');\n array_index | member_name\n-------------+-------------\n 1 | user1\n 2 | user2\n 3 | user3\n\n\npg_get_grolist(text) is intended for internal use, as is the pg_grolist_rec \ncomposite type. group_list() is intended as the user facing table function. I \nwould implement this by running the three statements above during initdb.\n\nAny comments or objections WRT object names or the method of implementation? I \ndon't think this is a very speed critical application, but even using the sql \nfunctions it is very fast:\ntest=# explain analyze select * from group_list('g2');\n QUERY PLAN\n------------------------------------------------------------------------------------------------------------\n Function Scan on group_list (cost=0.00..12.50 rows=1000 width=36) (actual \ntime=1.49..1.50 rows=3 loops=1)\n Total runtime: 1.55 msec\n(2 rows)\n\n\nI have more planned beyond the above as outlined in an earlier post (see \nhttp://archives.postgresql.org/pgsql-hackers/2002-11/msg01213.php).\n\nNext on my list will be a split() function (as discussed in early September) \nthat creates an array from an input string by splitting on a given delimiter. \nThis is similar to functions in perl, php, and undoubtedly other languages. It \nshould work nicely in conjunction with array_values().\n\nSorry for the long mail and thanks for any feedback!\n\nJoe\n\n", "msg_date": "Sun, 08 Dec 2002 08:49:00 -0800", "msg_from": "Joe Conway <mail@joeconway.com>", "msg_from_op": true, "msg_subject": "proposal: array utility functions phase 1" }, { "msg_contents": "Joe Conway <mail@joeconway.com> writes:\n> [ much snipped ]\n> The first function borrows from an idea Nigel Andrews had -- i.e. expand an \n> array into rows (and possibly columns). It currently works like this:\n\n> -- 1D array\n> test=# select * from array_values('{101,102,103,104}'::int[]) as (a int, b1 int);\n\n> Now on to the TODO item. Given the array_values() function, here's what I was\n> thinking of to implement listing members of a group:\n\n> CREATE OR REPLACE FUNCTION pg_get_grolist(text) RETURNS INT[] AS 'SELECT \n> grolist FROM pg_group WHERE groname = $1' LANGUAGE 'sql';\n\n> CREATE TYPE pg_grolist_rec AS (array_index int, member_name text);\n\n> CREATE OR REPLACE FUNCTION group_list(text) RETURNS SETOF pg_grolist_rec AS \n> 'SELECT g.id, pg_get_userbyid(g.usesysid)::text AS member_name FROM \n> array_values(pg_get_grolist($1)) AS g(id int, usesysid int)' LANGUAGE 'sql';\n\n\nThis crystallizes something that has been bothering me for awhile: the\ntable function syntax is severely hobbled (not to say crippled :-() by\nthe fact that the function arguments have to be constants. You really\ndon't want to have to invent intermediate functions every time you want\na slightly different query --- yet this technique seems to require *two*\nbespoke functions for every query, one on each end of the array_values()\nfunction.\n\nThe original Berkeley syntax, messy as it was, at least avoided this\nproblem. For example, I believe this same problem could be solved\n(approximately) with\n\n\tselect array_values(grolist) from pg_group where groname = 'g2'\n\n--- getting the users shown as names instead of numbers would take an\nextra level of SELECT, but I leave the details to the reader.\n\nI think we ought to try to find a way that table functions could be used\nwith inputs that are taken from tables. In a narrow sense you can do\nthis already, with a sub-SELECT:\n\n\tselect * from my_table_func((select x from ...));\n\nbut (a) the sub-select can only return a single value, and (b) you can't\nget at any of the other columns in the row the sub-select is selecting.\nFor instance it won't help me much to do\n\n\tselect * from\n\tarray_values((select grolist from pg_group where groname = 'g2'))\n\nif I want to show the group's grosysid as well.\n\nI know I'm not explaining this very well (I'm only firing on one\ncylinder today :-(), but basically I think we need to step back and take\nanother look at the mechanism before we start inventing tons of helper\nfunctions to make up for the lack of adequate mechanism.\n\n\nAs for array_values() itself, it seems fairly inelegant to rely on the\nuser to get the input and output types to match up. Now that we have\nan \"anyarray\" pseudotype, I think it wouldn't be unreasonable to hack up\nsome kluge in the parser to allow reference to the element type of such\nan argument --- that is, you'd say something like\n\n\tcreate function array_values(anyarray) returns setof anyarray_element\n\nand the parser would automatically understand what return type to assign\nto any particular use of array_values. (Since type resolution is done\nbottom-up, I see no logical difficulty here, though the implementation\nmight be a bit of a wart...)\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 08 Dec 2002 17:06:40 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: proposal: array utility functions phase 1 " }, { "msg_contents": "Tom Lane wrote:\n> This crystallizes something that has been bothering me for awhile: the\n> table function syntax is severely hobbled (not to say crippled :-() by\n> the fact that the function arguments have to be constants. You really\n> don't want to have to invent intermediate functions every time you want\n> a slightly different query --- yet this technique seems to require *two*\n> bespoke functions for every query, one on each end of the array_values()\n> function.\n\nIt did for me too. I was thinking along these lines while working on the \nconnectby function, but this work really makes it clear.\n\n> The original Berkeley syntax, messy as it was, at least avoided this\n> problem. For example, I believe this same problem could be solved\n> (approximately) with\n> \n> \tselect array_values(grolist) from pg_group where groname = 'g2'\n\nYes, this is exactly what I was yearning to do. Was there a spec or technical \nreason (or both) for not allowing the following?\n\n select * from array_values(g.grolist), pg_group g where g.groname = 'g2';\n\nIt seems like you could treat it like a one-to-many join between pg_group and \nthe function. I'm sure this is a bad idea and breaks down for more complex \nexamples, but I often have found myself wishing I could do exactly that.\n\n\n> I think we ought to try to find a way that table functions could be used\n> with inputs that are taken from tables. In a narrow sense you can do\n> this already, with a sub-SELECT:\n> \n> \tselect * from my_table_func((select x from ...));\n> \n> but (a) the sub-select can only return a single value, and (b) you can't\n> get at any of the other columns in the row the sub-select is selecting.\n> For instance it won't help me much to do\n> \n> \tselect * from\n> \tarray_values((select grolist from pg_group where groname = 'g2'))\n> \n> if I want to show the group's grosysid as well.\n\nYou could do something like:\n select * from array_values('pg_group','grolist') ...\nand repeat the rest of pg_group's columns for each row produced from grolist \nin the output (this is closer to what Nigel did, IIRC). This even works in the \ncurrent table function implementation. It does not get around the issue of \nspecifying querytime column refs though.\n\n> I know I'm not explaining this very well (I'm only firing on one\n> cylinder today :-(), but basically I think we need to step back and take\n> another look at the mechanism before we start inventing tons of helper\n> functions to make up for the lack of adequate mechanism.\n\nNope, you're explaining it just fine -- it's what I've been thinking for a \nwhile, but couldn't articulate myself.\n\n\n> As for array_values() itself, it seems fairly inelegant to rely on the\n> user to get the input and output types to match up. Now that we have\n> an \"anyarray\" pseudotype, I think it wouldn't be unreasonable to hack up\n> some kluge in the parser to allow reference to the element type of such\n> an argument --- that is, you'd say something like\n> \n> \tcreate function array_values(anyarray) returns setof anyarray_element\n> \n> and the parser would automatically understand what return type to assign\n> to any particular use of array_values. (Since type resolution is done\n> bottom-up, I see no logical difficulty here, though the implementation\n> might be a bit of a wart...)\n\nThat doesn't quite work as written (you'd have to account for the array index \ncolumn or lose it -- which loses any ability to get position in the array), \nand has even more problems with the array_values('pg_group','grolist') approach.\n\nHow ugly/difficult would it be to allow the planner to interrogate the \nfunction and let the function report back a tupledesc based on the actual \nruntime input parameters? Kind of a special mode of function call that the \nfunction could detect and respond to differently than during execution (to \navoid excessive runtime an/or side effects -- just form a tupledesc and return \nit). Then the planner could move forward without requiring a specific declared \nreturn composite type or a return type of record with a runtime query column \ndefinition.\n\nJoe\n\n", "msg_date": "Sun, 08 Dec 2002 17:56:37 -0800", "msg_from": "Joe Conway <mail@joeconway.com>", "msg_from_op": true, "msg_subject": "Re: proposal: array utility functions phase 1" }, { "msg_contents": "Joe Conway <mail@joeconway.com> writes:\n> Yes, this is exactly what I was yearning to do. Was there a spec or technical\n> reason (or both) for not allowing the following?\n\n> select * from array_values(g.grolist), pg_group g where g.groname = 'g2';\n\nThis seems fairly unworkable to me as-is. By definition, WHERE selects\nfrom a cross-product of the FROM tables; to make the above do what you\nwant, you'd have to break that fundamental semantics. The semantics of\nexplicit JOIN cases would be broken too.\n\nWhat we need is some kind of explicit multi-level SELECT operation.\nPerhaps it would help to think about the analogy of aggregates of\naggregate functions, which are impossible to express properly in a\nsingle SELECT but work nicely given subselect-in-FROM.\nSubselect-in-FROM doesn't seem to get this job done though.\n\nRight offhand I don't see any reasonable syntax other than\nfunction-in-the-SELECT-list, which shoots us right back into the\nmessinesses of the Berkeley implementation. However, we do now have the\nprecedent of the table-function AS clause. Does it help any to do\nsomething like\n\n\tSELECT grosysid, array_values(grolist) AS (array_index,member_id)\n\tFROM pg_group where groname = 'g2';\n\n(Again you could wrap this in an outer SELECT to transform the\nmember_ids to member_names.)\n\nThe real problem with the Berkeley approach shows up when you consider\nwhat happens with multiple table functions called in a single SELECT.\nThe code we currently have produces the cross-product of the implied\nrows (or at least it tries to, I seem to recall that it doesn't\nnecessarily get it right). That's pretty unpleasant, and though you can\nfilter the rows in an outer SELECT, there's no way to optimize the\nimplementation into a smarter-than-nested-loop join.\n\nIt seems like somehow we need a level of FROM/WHERE producing some base\nrows, and then a set of table function calls to apply to each of the\nbase rows, and then another level of WHERE to filter the results of the\nfunction calls (in particular to provide join conditions to identify\nwhich rows to match up in the function outputs). I don't see any way to\ndo this without inventing new SELECT clauses out of whole cloth\n... unless SQL99's WITH clause helps, but I don't think it does ...\n\n> How ugly/difficult would it be to allow the planner to interrogate the \n> function and let the function report back a tupledesc based on the actual \n> runtime input parameters?\n\nParse-time, not run-time. It could be done --- IIRC, the auxiliary\n\"function info\" call we introduced in the V1 fmgr protocol was\ndeliberately designed to allow expansion in this sort of direction.\nBut it would have to take a tupledesc (or some similar static\ndescription) and return another one.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 13:13:49 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: proposal: array utility functions phase 1 " }, { "msg_contents": "Tom Lane wrote:\n> This seems fairly unworkable to me as-is. By definition, WHERE selects\n> from a cross-product of the FROM tables; to make the above do what you\n> want, you'd have to break that fundamental semantics. The semantics of\n> explicit JOIN cases would be broken too.\n> \n> What we need is some kind of explicit multi-level SELECT operation.\n> Perhaps it would help to think about the analogy of aggregates of\n> aggregate functions, which are impossible to express properly in a\n> single SELECT but work nicely given subselect-in-FROM.\n> Subselect-in-FROM doesn't seem to get this job done though.\n> \n> Right offhand I don't see any reasonable syntax other than\n> function-in-the-SELECT-list, which shoots us right back into the\n> messinesses of the Berkeley implementation. However, we do now have the\n> precedent of the table-function AS clause. Does it help any to do\n> something like\n> \n> \tSELECT grosysid, array_values(grolist) AS (array_index,member_id)\n> \tFROM pg_group where groname = 'g2';\n\nAfter further thought, and ignoring the difficulty of implementation, what \nseems ideal is to be able to specify 'setof <datatype>' or 'setof \n<composite-type>' as an input to the function, and fire the function once for \neach row of the input. Basically, allow anything that now qualifies as a FROM \nitem -- a table reference, a subselect with AS clause, another table function, \nor maybe even a join clause. Some (totally contrived) examples of how it would \nlook:\n\ncreate table foo1(f1 int, f2 text);\ninsert into foo1 values(1,'a');\ninsert into foo1 values(2,'b');\ninsert into foo1 values(3,'c');\n\ncreate table foo2(f1 int, f2 text);\ninsert into foo2 values(1,'w');\ninsert into foo2 values(1,'x');\ninsert into foo2 values(2,'y');\ninsert into foo2 values(2,'z');\n\ncreate function funcfoo1(setof foo1) returns setof foo2 as 'select * from foo2 \nwhere foo2.f1 = $1.f1' language 'sql';\nselect * from funcfoo1(foo1);\n f1 f2\n----+-----\n 1 | w\n 1 | x\n 2 | y\n 2 | z\n\nselect * from funcfoo1((select * from foo1 where f1=1) as t);\n f1 f2\n----+-----\n 1 | w\n 1 | x\n\n\nWhat do you think?\n\n\n> (Again you could wrap this in an outer SELECT to transform the\n> member_ids to member_names.)\n> \n> The real problem with the Berkeley approach shows up when you consider\n> what happens with multiple table functions called in a single SELECT.\n> The code we currently have produces the cross-product of the implied\n> rows (or at least it tries to, I seem to recall that it doesn't\n> necessarily get it right). That's pretty unpleasant, and though you can\n> filter the rows in an outer SELECT, there's no way to optimize the\n> implementation into a smarter-than-nested-loop join.\n\nWhat if there was a way to declare that a table function returns sorted \nresults, and on which column(s)?\n\n\n> It seems like somehow we need a level of FROM/WHERE producing some base\n> rows, and then a set of table function calls to apply to each of the\n> base rows, and then another level of WHERE to filter the results of the\n> function calls (in particular to provide join conditions to identify\n> which rows to match up in the function outputs). I don't see any way to\n> do this without inventing new SELECT clauses out of whole cloth\n> ... unless SQL99's WITH clause helps, but I don't think it does ...\n\nIs this still needed given my approach above?\n\n\n>>How ugly/difficult would it be to allow the planner to interrogate the \n>>function and let the function report back a tupledesc based on the actual \n>>runtime input parameters?\n> \n> \n> Parse-time, not run-time. It could be done --- IIRC, the auxiliary\n> \"function info\" call we introduced in the V1 fmgr protocol was\n> deliberately designed to allow expansion in this sort of direction.\n> But it would have to take a tupledesc (or some similar static\n> description) and return another one.\n\nNice! I'll dig in to that a bit.\n\nThanks,\n\nJoe\n\n\n\n", "msg_date": "Mon, 09 Dec 2002 19:48:13 -0800", "msg_from": "Joe Conway <mail@joeconway.com>", "msg_from_op": true, "msg_subject": "Re: proposal: array utility functions phase 1" }, { "msg_contents": "Tom Lane wrote:\n> It seems like somehow we need a level of FROM/WHERE producing some base\n> rows, and then a set of table function calls to apply to each of the\n> base rows, and then another level of WHERE to filter the results of the\n> function calls (in particular to provide join conditions to identify\n> which rows to match up in the function outputs). I don't see any way to\n> do this without inventing new SELECT clauses out of whole cloth\n> ... unless SQL99's WITH clause helps, but I don't think it does ...\n\nWell, maybe this is a start. It allows a table function's input parameter to \nbe declared with setof. The changes involved primarily:\n\n1) a big loop in ExecMakeTableFunctionResult so that functions with set \nreturning arguments get called for each row of the argument,\n and\n2) aways initializing the tuplestore in ExecMakeTableFunctionResult and \npassing that to the function, even when SFRM_Materialize mode is used.\n\nThe result looks like:\n\ncreate table foot(f1 text, f2 text);\ninsert into foot values('a','b');\ninsert into foot values('c','d');\ninsert into foot values('e','f');\n\ncreate or replace function test2() returns setof foot as 'select * from foot \norder by 1 asc' language 'sql';\ncreate or replace function test(setof foot) returns foot as 'select $1.f1, \n$1.f2' language 'sql';\n\nregression=# select * from test(test2());\n f1 | f2\n----+----\n a | b\n c | d\n e | f\n(3 rows)\n\nI know it doesn't solve all the issues discussed, but is it a step forward? \nSuggestions?\n\nJoe", "msg_date": "Wed, 11 Dec 2002 14:27:55 -0800", "msg_from": "Joe Conway <mail@joeconway.com>", "msg_from_op": true, "msg_subject": "Re: proposal: array utility functions phase 1" }, { "msg_contents": "Joe Conway wrote:\n> Tom Lane wrote:\n> \n>> It seems like somehow we need a level of FROM/WHERE producing some base\n>> rows, and then a set of table function calls to apply to each of the\n>> base rows, and then another level of WHERE to filter the results of the\n>> function calls (in particular to provide join conditions to identify\n>> which rows to match up in the function outputs). I don't see any way to\n>> do this without inventing new SELECT clauses out of whole cloth\n>> ... unless SQL99's WITH clause helps, but I don't think it does ...\n> \n> \n> Well, maybe this is a start. It allows a table function's input \n> parameter to be declared with setof. The changes involved primarily:\n> \n> 1) a big loop in ExecMakeTableFunctionResult so that functions with set \n> returning arguments get called for each row of the argument,\n> and\n> 2) aways initializing the tuplestore in ExecMakeTableFunctionResult and \n> passing that to the function, even when SFRM_Materialize mode is used.\n> \n> The result looks like:\n> \n> create table foot(f1 text, f2 text);\n> insert into foot values('a','b');\n> insert into foot values('c','d');\n> insert into foot values('e','f');\n> \n> create or replace function test2() returns setof foot as 'select * from \n> foot order by 1 asc' language 'sql';\n> create or replace function test(setof foot) returns foot as 'select \n> $1.f1, $1.f2' language 'sql';\n> \n> regression=# select * from test(test2());\n> f1 | f2\n> ----+----\n> a | b\n> c | d\n> e | f\n> (3 rows)\n> \n> I know it doesn't solve all the issues discussed, but is it a step \n> forward? Suggestions?\n> \n\nThe patch was no longer applying cleanly to cvs, so I updated it. Also moved \nfrom HACKERS to PATCHES. Passes all regression tests (well -- geometry still \nfails, but that is unrelated).\n\nJoe", "msg_date": "Fri, 13 Dec 2002 09:01:49 -0800", "msg_from": "Joe Conway <mail@joeconway.com>", "msg_from_op": true, "msg_subject": "SETOF input parameters (was Re: [HACKERS] proposal: array utility\n\tfunctions phase 1)" } ]
[ { "msg_contents": "Looking at my problem with changed cursor behaviour in 7.3 again, I\nnoticed something interesting: a cursor in 7.3 apparently does not let \nyou scroll back to its first row at all! Neither a \"move backward all\"\nor a \"move -n\" where n is equal to or greater than the cursor's current \nposition, will let you fetch any more rows from the cursor. Scrolling \nback to the second row does work though.\n\nI can work around this in libpqxx, without too much trouble, but I'm \nnot sure I should have to.\n\n\nJeroen\n\n", "msg_date": "Sun, 8 Dec 2002 19:28:03 +0100", "msg_from": "\"Jeroen T. Vermeulen\" <jtv@xs4all.nl>", "msg_from_op": true, "msg_subject": "More on cursors in 7.3" }, { "msg_contents": "\"Jeroen T. Vermeulen\" <jtv@xs4all.nl> writes:\n> Looking at my problem with changed cursor behaviour in 7.3 again, I\n> noticed something interesting: a cursor in 7.3 apparently does not let \n> you scroll back to its first row at all!\n\nOh?\n\nregression=# begin;\nBEGIN\nregression=# declare c cursor for select * from int8_tbl;\nDECLARE CURSOR\nregression=# fetch all from c;\n q1 | q2\n------------------+-------------------\n 123 | 456\n 123 | 4567890123456789\n 4567890123456789 | 123\n 4567890123456789 | 4567890123456789\n 4567890123456789 | -4567890123456789\n(5 rows)\n\nregression=# move backward all in c;\nMOVE 5\nregression=# fetch all from c;\n q1 | q2\n------------------+-------------------\n 123 | 456\n 123 | 4567890123456789\n 4567890123456789 | 123\n 4567890123456789 | 4567890123456789\n 4567890123456789 | -4567890123456789\n(5 rows)\n\nregression=#\n\n\nI believe it is true though that backing up a cursor only works for\ncertain plan types (seqscan, indexscan, sort, maybe a couple others).\nThat has always been true --- 7.3 is no better nor worse than prior\nreleases.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 08 Dec 2002 16:28:38 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: More on cursors in 7.3 " }, { "msg_contents": "On Sun, Dec 08, 2002 at 04:28:38PM -0500, Tom Lane wrote:\n> \n> I believe it is true though that backing up a cursor only works for\n> certain plan types (seqscan, indexscan, sort, maybe a couple others).\n> That has always been true --- 7.3 is no better nor worse than prior\n> releases.\n\nAh, I didn't know that. I guess the plan for \"select * from pg_tables\"\nmust have changed in 7.3 then.\n\nIs any of this described in the docs somewhere?\n\n\nJeroen\n\n", "msg_date": "Sun, 8 Dec 2002 22:33:16 +0100", "msg_from": "\"Jeroen T. Vermeulen\" <jtv@xs4all.nl>", "msg_from_op": true, "msg_subject": "Re: More on cursors in 7.3" }, { "msg_contents": "\"Jeroen T. Vermeulen\" <jtv@xs4all.nl> writes:\n> Ah, I didn't know that. I guess the plan for \"select * from pg_tables\"\n> must have changed in 7.3 then.\n\n[looks...] Yeah, there's a join to pg_namespace in there now.\n\n> Is any of this described in the docs somewhere?\n\nFraid not.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 08 Dec 2002 17:09:09 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: More on cursors in 7.3 " }, { "msg_contents": "On Sun, Dec 08, 2002 at 05:09:09PM -0500, Tom Lane wrote:\n> \n> > Is any of this described in the docs somewhere?\n> \n> Fraid not.\n\nDamn & blast. I was rather counting on cursors that could back up for\nmy nifty CachedResult class (which acts more or less like a normal result \nset but transparently fetches rows on demand).\n\nNow if I understood a bit more of what's going on here, at least I could\ndocument it...\n\n\nJeroen\n\n", "msg_date": "Sun, 8 Dec 2002 23:13:49 +0100", "msg_from": "\"Jeroen T. Vermeulen\" <jtv@xs4all.nl>", "msg_from_op": true, "msg_subject": "Re: More on cursors in 7.3" }, { "msg_contents": "\"Jeroen T. Vermeulen\" <jtv@xs4all.nl> writes:\n> Now if I understood a bit more of what's going on here, at least I could\n> document it...\n\nWell, you could dig through backend/executor/node*.c and see which of\nthe node types pay attention to es_direction. To a first approximation\nit looks like these do:\n\nFunctionscan\nAppend\nIndexscan\nMergejoin\nLimit\nMaterial\nSubqueryscan\nSeqscan\nSort\nTidscan\n\nalthough I have not thought about which other upper plan nodes might be\nokay (ie, they're safe if their input nodes are). Also, a Material or\nSort node will hide any direction-unsafety in its input.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 08 Dec 2002 17:28:22 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: More on cursors in 7.3 " }, { "msg_contents": "On Sun, Dec 08, 2002 at 05:28:22PM -0500, Tom Lane wrote:\n> \n> Well, you could dig through backend/executor/node*.c and see which of\n> the node types pay attention to es_direction. To a first approximation\n> it looks like these do:\n \nI'll be honest with you: I don't know much about the internals and this\nis pure Greek to me... And I never was much good at Greek in school.\n\n\n> although I have not thought about which other upper plan nodes might be\n> okay (ie, they're safe if their input nodes are). Also, a Material or\n> Sort node will hide any direction-unsafety in its input.\n\nMore Greek, I'm afraid. :-(\n\n\nJeroen\n\n", "msg_date": "Sun, 8 Dec 2002 23:58:14 +0100", "msg_from": "\"Jeroen T. Vermeulen\" <jtv@xs4all.nl>", "msg_from_op": true, "msg_subject": "Re: More on cursors in 7.3" } ]
[ { "msg_contents": "(Doh! Forgot to attach first time.)\n\nThis partially nails this TODO item:\n\n* Add schema, cast, and conversion backslash commands to psql\n\nI had to create a new publically available function,\npg_conversion_is_visible, as it seemed to be missing from the catalogs.\nThis required me to do no small amount of hacking around in namespace.c\n\nI have updated the \\? help and sgml docs.\n\n\\dc - list conversions [PATTERN]\n\\dC - list casts\n\nI didn't support patterns with casts as there's nothing obvious to match\nagainst.\n\nI've tested it a fair bit and I can't see any problems. Feel free to do a\nonce over on my backend changes tho.\n\nChris", "msg_date": "Sun, 8 Dec 2002 20:39:11 -0800", "msg_from": "\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au>", "msg_from_op": true, "msg_subject": "show casts and conversions in psql (2nd try)" }, { "msg_contents": "Christopher Kings-Lynne writes:\n\n> \\dc - list conversions [PATTERN]\n> \\dC - list casts\n\nWhat are we going to use for collations?\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Mon, 9 Dec 2002 20:48:55 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: show casts and conversions in psql (2nd try)" }, { "msg_contents": "\n\n> Christopher Kings-Lynne writes:\n> \n> > \\dc - list conversions [PATTERN]\n> > \\dC - list casts\n> \n> What are we going to use for collations?\n\n\\dn Is the only letter left in collations that hasn't been used!\n\nChris\n\n", "msg_date": "Mon, 9 Dec 2002 12:10:24 -0800", "msg_from": "\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au>", "msg_from_op": true, "msg_subject": "Re: show casts and conversions in psql (2nd try)" }, { "msg_contents": "[ moved to hackers from pgsql-patches ]\n\n\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au> writes:\n> Peter wrote:\n>> Christopher Kings-Lynne writes:\n>>> \\dc - list conversions [PATTERN]\n>>> \\dC - list casts\n>> \n>> What are we going to use for collations?\n\n> \\dn Is the only letter left in collations that hasn't been used!\n\n... and that was already proposed for \"show schemas\" (namespaces).\n\nI'm inclined to think it's time to bite the bullet and go over to\nwords rather than single characters to identify the \\d target\n(viz, \"\\dschema\", \"\\dcast\", etc, presumably with unique abbreviations\nbeing allowed, as well as special cases for the historical single\ncharacters).\n\nThe issue here is what do we do with the existing \"\\d[istvS]\" behavior\n(for instance, \"\\dsit\" means \"list sequences, indexes, and tables\").\nIs that useful enough to try to preserve, or do we just bit-bucket it?\nIf we do try to preserve it, how should it work?\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 16:05:43 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "psql's \\d commands --- end of the line for 1-character identifiers?" }, { "msg_contents": "> ... and that was already proposed for \"show schemas\" (namespaces).\n>\n> I'm inclined to think it's time to bite the bullet and go over to\n> words rather than single characters to identify the \\d target\n> (viz, \"\\dschema\", \"\\dcast\", etc, presumably with unique abbreviations\n> being allowed, as well as special cases for the historical single\n> characters).\n\nHmmm...I'm not certain that the \\d commands really NEED to have a logical\nlink to the actual thing you're listing. If you just made \\dh for schemas,\npeople would look it up and then remember it from then on. It's probably\nnot a huge deal.\n\nWe could do DESCRIBE commands as well. Also, what happened to the\nINFORMATION_SCHEMA proposal? Wasn't Peter E doing something with that?\nWhat happened to it?\n\n> The issue here is what do we do with the existing \"\\d[istvS]\" behavior\n> (for instance, \"\\dsit\" means \"list sequences, indexes, and tables\").\n> Is that useful enough to try to preserve, or do we just bit-bucket it?\n> If we do try to preserve it, how should it work?\n\nI'd much rather it were preserved, and I'm sure most people would as well.\n\nChris\n\n", "msg_date": "Mon, 9 Dec 2002 13:22:20 -0800", "msg_from": "\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au>", "msg_from_op": true, "msg_subject": "Re: psql's \\d commands --- end of the line for 1-character\n\tidentifiers?" }, { "msg_contents": "Tom Lane kirjutas T, 10.12.2002 kell 02:05:\n> [ moved to hackers from pgsql-patches ]\n> \n> \"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au> writes:\n> > Peter wrote:\n> >> Christopher Kings-Lynne writes:\n> >>> \\dc - list conversions [PATTERN]\n> >>> \\dC - list casts\n> >> \n> >> What are we going to use for collations?\n> \n> > \\dn Is the only letter left in collations that hasn't been used!\n> \n> ... and that was already proposed for \"show schemas\" (namespaces).\n> \n> I'm inclined to think it's time to bite the bullet and go over to\n> words rather than single characters to identify the \\d target\n> (viz, \"\\dschema\", \"\\dcast\", etc, presumably with unique abbreviations\n> being allowed, as well as special cases for the historical single\n> characters).\n> \n> The issue here is what do we do with the existing \"\\d[istvS]\" behavior\n> (for instance, \"\\dsit\" means \"list sequences, indexes, and tables\").\n> Is that useful enough to try to preserve, or do we just bit-bucket it?\n> If we do try to preserve it, how should it work?\n\nWhy not use \\D for \"long\" ids ?\n\nSomewhat similar to -? and --help for command line.\n\n-- \nHannu Krosing <hannu@tm.ee>\n", "msg_date": "10 Dec 2002 02:25:12 +0500", "msg_from": "Hannu Krosing <hannu@tm.ee>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for" }, { "msg_contents": "Hannu Krosing <hannu@tm.ee> writes:\n> Why not use \\D for \"long\" ids ?\n\nSeems like a fine idea to me. (I had actually started to think of\n\"\\s<something>\" for \"show\", but was just observing that that would\ncreate conflicts against existing commands, when your message arrived.\n\"\\D<something>\" works though.)\n\nAny objections out there?\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 17:13:02 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for " }, { "msg_contents": "At 05:13 PM 9/12/2002 -0500, Tom Lane wrote:\n>Seems like a fine idea to me.\n\nDitto.\n\n>\"\\D<something>\" works though.)\n>\n>Any objections out there?\n\nMy only complaint here is being forced to use the 'shift' key on commands \nthat will be common. I would prefer any other lower case char: \\b, \\j, \\k , \n\\m, \\n, \\u, \\v, and \\y are available. I'd vote for \\v (view), or \\k \n(command). The go with:\n\n \\v schema\nor\n \\k show schema\n\n(I'd vote for \\v).\n\n\n\n\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Tue, 10 Dec 2002 10:12:31 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for " }, { "msg_contents": "Christopher Kings-Lynne wrote:\n> We could do DESCRIBE commands as well. Also, what happened to the\n> INFORMATION_SCHEMA proposal? Wasn't Peter E doing something with that?\n> What happened to it?\n> \n> > The issue here is what do we do with the existing \"\\d[istvS]\" behavior\n> > (for instance, \"\\dsit\" means \"list sequences, indexes, and tables\").\n> > Is that useful enough to try to preserve, or do we just bit-bucket it?\n> > If we do try to preserve it, how should it work?\n> \n> I'd much rather it were preserved, and I'm sure most people would as well.\n\nI was going to say the opposite, that it isn't needed. Oh well.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Mon, 9 Dec 2002 20:36:49 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for 1-character" }, { "msg_contents": "At 01:22 PM 9/12/2002 -0800, Christopher Kings-Lynne wrote:\n>Hmmm...I'm not certain that the \\d commands really NEED to have a logical\n>link to the actual thing you're listing.\n\nThis is the perspective a person with good memory, unlike me. In find it \nuseful to be able to derive commands from common-sense rules, even if it \nmeans a little more typing.\n\n\n>We could do DESCRIBE commands as well. Also, what happened to the\n\nThis would be fine, so long as the standard does not get in the way of \ndisplaying postgres-specific information (eg. function attrs).\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Tue, 10 Dec 2002 12:55:51 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for" }, { "msg_contents": "On Tue, Dec 10, 2002 at 12:55:51PM +1100, Philip Warner wrote:\n> At 01:22 PM 9/12/2002 -0800, Christopher Kings-Lynne wrote:\n> >Hmmm...I'm not certain that the \\d commands really NEED to have a logical\n> >link to the actual thing you're listing.\n> \n> This is the perspective a person with good memory, unlike me. In find it \n> useful to be able to derive commands from common-sense rules, even if it \n> means a little more typing.\n\nWould it work to make \\d tab-completable in a way that showed both the\ncommands that are available and the objects they describe? e.g.\n\n\\d<tab> would show something like\n\\dt [tables]\t\\ds [sequences]\t\\dv [views] ...\n\n(the way it's shown now shows what completions are available, but not\nwhat they mean. Also, both \\d and \\D should be shown in any case)\n\n-- \nAlvaro Herrera (<alvherre[a]dcc.uchile.cl>)\n\"Los romanticos son seres que mueren de deseos de vida\"\n", "msg_date": "Tue, 10 Dec 2002 01:55:38 -0300", "msg_from": "Alvaro Herrera <alvherre@dcc.uchile.cl>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for" }, { "msg_contents": "At 01:55 AM 10/12/2002 -0300, Alvaro Herrera wrote:\n>\\d<tab> would show something like\n>\\dt [tables] \\ds [sequences] \\dv [views] ...\n>\n>(the way it's shown now shows what completions are available, but not\n>what they mean. Also, both \\d and \\D should be shown in any case)\n\nThis would be OK, but I'd be very happy with DESCRIBE, especially if \ntab-completion meant I could type 'DESC<tab>TAB<tab><name>' instead of \n'DESCRIBE TABLE <name>'.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Tue, 10 Dec 2002 17:16:26 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for" }, { "msg_contents": "> >\"\\D<something>\" works though.)\n> >\n> >Any objections out there?\n>\n> My only complaint here is being forced to use the 'shift' key on commands\n> that will be common.\n\n\\dd perhaps?\n\n", "msg_date": "Tue, 10 Dec 2002 00:18:12 -0600", "msg_from": "\"Matthew T. O'Connor\" <matthew@zeut.net>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for " }, { "msg_contents": "Philip Warner <pjw@rhyme.com.au> writes:\n> This would be OK, but I'd be very happy with DESCRIBE, especially if \n> tab-completion meant I could type 'DESC<tab>TAB<tab><name>' instead of \n> 'DESCRIBE TABLE <name>'.\n\nThat's quicker than <backslash><shift><D><unshift><t><space><name> ?\n\nI don't want to sound like I've got some kind of religious objection\nto DESCRIBE, but it was just a couple of hours ago that someone was\nobjecting to \"\\D\" because it'd require touching the shift key. Let's\nget a bit realistic on the ease-of-typing arguments here.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Dec 2002 01:25:53 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for " }, { "msg_contents": "On Mon, 2002-12-09 at 23:12, Philip Warner wrote:\n> At 05:13 PM 9/12/2002 -0500, Tom Lane wrote:\n> >Seems like a fine idea to me.\n> \n> Ditto.\n> \n> >\"\\D<something>\" works though.)\n> >\n> >Any objections out there?\n> \n> My only complaint here is being forced to use the 'shift' key on commands \n> that will be common.\n\nOn most european keyboards you alreday have to use \"AltGr\" to get to \\\nso using an extra shift is not too bad ;)\n\n\n-- \nHannu Krosing <hannu@tm.ee>\n", "msg_date": "10 Dec 2002 17:05:42 +0000", "msg_from": "Hannu Krosing <hannu@tm.ee>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for" }, { "msg_contents": "\n\n> At 01:25 AM 10/12/2002 -0500, Tom Lane wrote:\n> >Let's\n> >get a bit realistic on the ease-of-typing arguments here.\n>\n> It's a fair cop, but don't forget the memory argument as well - I did say\nI\n> was happy with \\d<tab> providing prompts, and DESCRIBE is a little more\n> portable & memorable than \\d[heiroglyphic].\n\nI think the problem with DESCRIBE is that it's supposed to just return a\nrecordset. I don't see it showing fk's, indexes, rules, etc. as well...\n\nChris\n\n", "msg_date": "Tue, 10 Dec 2002 11:47:04 -0800", "msg_from": "\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au>", "msg_from_op": true, "msg_subject": "Re: psql's \\d commands --- end of the line for " }, { "msg_contents": "Peter Eisentraut <peter_e@gmx.net> writes:\n> Christopher Kings-Lynne writes:\n>> We could do DESCRIBE commands as well. Also, what happened to the\n>> INFORMATION_SCHEMA proposal? Wasn't Peter E doing something with that?\n>> What happened to it?\n\n> Ooops. Yeah, let's get this in. Where should I put it?\n\nHow do you mean \"where\"? The spec says it's gotta be called\ninformation_schema, no? What's left to decide?\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Dec 2002 15:03:05 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for 1-character\n\tidentifiers?" }, { "msg_contents": "Alvaro Herrera writes:\n\n> Would it work to make \\d tab-completable in a way that showed both the\n> commands that are available and the objects they describe? e.g.\n>\n> \\d<tab> would show something like\n> \\dt [tables]\t\\ds [sequences]\t\\dv [views] ...\n\nThat won't work. The actual completion and the view of the alternatives\nif the completion is ambiguous is driven by the same data.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Tue, 10 Dec 2002 21:04:07 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for" }, { "msg_contents": "Christopher Kings-Lynne writes:\n\n> We could do DESCRIBE commands as well. Also, what happened to the\n> INFORMATION_SCHEMA proposal? Wasn't Peter E doing something with that?\n> What happened to it?\n\nOoops. Yeah, let's get this in. Where should I put it?\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Tue, 10 Dec 2002 21:04:59 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for 1-character\n\tidentifiers?" }, { "msg_contents": "> > We could do DESCRIBE commands as well. Also, what happened to the\n> > INFORMATION_SCHEMA proposal? Wasn't Peter E doing something with that?\n> > What happened to it?\n>\n> Ooops. Yeah, let's get this in. Where should I put it?\n\nI wouldn't mind having a look at the patch. Where do you implement this\nkind of thing? Where in the code do you create system views and schemas?\nJust add to the initdb script or something?\n\nAdding this should allow us to move around 20 items from the sql99\nunsupported list to the supported, which would be sweet.\n\nChris\n\n", "msg_date": "Tue, 10 Dec 2002 13:51:10 -0800", "msg_from": "\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au>", "msg_from_op": true, "msg_subject": "INFORMATION_SCHEMA" }, { "msg_contents": "> > We could do DESCRIBE commands as well. Also, what happened to the\n> > INFORMATION_SCHEMA proposal? Wasn't Peter E doing something with that?\n> > What happened to it?\n>\n> Ooops. Yeah, let's get this in. Where should I put it?\n\nI wouldn't mind having a look at the patch. Where do you implement this\nkind of thing? Where in the code do you create system views and schemas?\nJust add to the initdb script or something?\n\nAdding this should allow us to move around 20 items from the sql99\nunsupported list to the supported, which would be sweet.\n\nChris\n\n", "msg_date": "Tue, 10 Dec 2002 13:51:45 -0800", "msg_from": "\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au>", "msg_from_op": true, "msg_subject": "INFORMATION_SCHEMA" }, { "msg_contents": "\nYour patch has been added to the PostgreSQL unapplied patches list at:\n\n\thttp://momjian.postgresql.org/cgi-bin/pgpatches\n\nI will try to apply it within the next 48 hours.\n\n---------------------------------------------------------------------------\n\n\nChristopher Kings-Lynne wrote:\n> (Doh! Forgot to attach first time.)\n> \n> This partially nails this TODO item:\n> \n> * Add schema, cast, and conversion backslash commands to psql\n> \n> I had to create a new publically available function,\n> pg_conversion_is_visible, as it seemed to be missing from the catalogs.\n> This required me to do no small amount of hacking around in namespace.c\n> \n> I have updated the \\? help and sgml docs.\n> \n> \\dc - list conversions [PATTERN]\n> \\dC - list casts\n> \n> I didn't support patterns with casts as there's nothing obvious to match\n> against.\n> \n> I've tested it a fair bit and I can't see any problems. Feel free to do a\n> once over on my backend changes tho.\n> \n> Chris\n\n[ Attachment, skipping... ]\n\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 5: Have you checked our extensive FAQ?\n> \n> http://www.postgresql.org/users-lounge/docs/faq.html\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 10 Dec 2002 22:47:04 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: show casts and conversions in psql (2nd try)" }, { "msg_contents": "Tom Lane writes:\n\n> How do you mean \"where\"? The spec says it's gotta be called\n> information_schema, no? What's left to decide?\n\nI was more wondering where in the source tree to put and how to install\nit. I would suggest the following: I have a file information_schema.sql\nwhich contains the commands to create the information schema in a\nuser-space manner. I'd put it into backend/catalog, from there it gets\ninstalled in the same place the .bki files go, and initdb installs it into\ntemplate1 near the end.\n\nMaybe someone wants to make a case for putting the information schema\ndefinitions into a bunch of separate .sql files, one for each item defined\nin the standard?\n\nWe also have to think whether we want to magically represent the\ninformation_schema in the search path as the spec prescribes. I would\nsuggest a \"no\" for that. The names of the objects in the information\nschema are way to obvious to put in the default namespace (pun intended\nfor amusement).\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Wed, 11 Dec 2002 20:57:12 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for 1-character\n\tidentifiers?" }, { "msg_contents": "Peter Eisentraut <peter_e@gmx.net> writes:\n> Tom Lane writes:\n>> How do you mean \"where\"?\n\n> I was more wondering where in the source tree to put and how to install\n> it. I would suggest the following: I have a file information_schema.sql\n> which contains the commands to create the information schema in a\n> user-space manner. I'd put it into backend/catalog, from there it gets\n> installed in the same place the .bki files go, and initdb installs it into\n> template1 near the end.\n\nAh. That works for me.\n\n> Maybe someone wants to make a case for putting the information schema\n> definitions into a bunch of separate .sql files, one for each item defined\n> in the standard?\n\nOne file per item is way overkill ... especially when there's no obvious\nreason that people would want some but not all of these items.\n\n> We also have to think whether we want to magically represent the\n> information_schema in the search path as the spec prescribes.\n\nWe already have the necessary mechanism to let this be set on a\nper-installation (or even per-database or per-user) basis. I agree with\nyour thought that we need not make it the factory default.\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 11 Dec 2002 23:14:43 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: psql's \\d commands --- end of the line for 1-character\n\tidentifiers?" }, { "msg_contents": "\nThis is the patch description.\n\n---------------------------------------------------------------------------\n\nChristopher Kings-Lynne wrote:\n> (Doh! Forgot to attach first time.)\n> \n> This partially nails this TODO item:\n> \n> * Add schema, cast, and conversion backslash commands to psql\n> \n> I had to create a new publically available function,\n> pg_conversion_is_visible, as it seemed to be missing from the catalogs.\n> This required me to do no small amount of hacking around in namespace.c\n> \n> I have updated the \\? help and sgml docs.\n> \n> \\dc - list conversions [PATTERN]\n> \\dC - list casts\n> \n> I didn't support patterns with casts as there's nothing obvious to match\n> against.\n> \n> I've tested it a fair bit and I can't see any problems. Feel free to do a\n> once over on my backend changes tho.\n> \n> Chris\n\n[ Attachment, skipping... ]\n\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 5: Have you checked our extensive FAQ?\n> \n> http://www.postgresql.org/users-lounge/docs/faq.html\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Thu, 12 Dec 2002 16:03:09 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: show casts and conversions in psql (2nd try)" } ]
[ { "msg_contents": "Vince, Peter:\n\nI can definitely understand someone not wanting to *participate* in\nmarketing/advocacy of PostgreSQL. However, your being opposed to\npromoting PostgreSQL as an organized activity *at all* baffles me. How\ncan you be against promoting PostgreSQL? Don't you want poeple to use\nyour code?\n\nFor me, it's not just a matter of preference, but of necessity; if\nPostgres becomes obscure, I stop being able to participate in the\nproject. While there are people on this list who are fortunate enough\nto be able to code whatever they want and still get paid, for a lot of\npeople, our participation hinges on the cycle:\n\nPostgres Users --> Postgres Contracts --> Postgres Jobs --> Postgres\nContributors --> Improvement *and Promotion* of Postgres --> Postgres\nUsers ...\n\nThe Promotion part of that step is *not* dispensable; all of the best\nfeatures in the world are not going to expand the Postgres commmunity\nif people haven't heard of it, can't find it, and know a lot more about\nMySQL anyway. While this may not be true for everybody, some of us\nhave clients or bosses who do read trade periodicals and demand that we\nfollow their technology reccomendations. I already have one client\nusing MySQL because of MySQL's \"much more professional\" web site and\n\"better support\" and \"better performance\".\n\nFrankly, if we blow off marketing PostgreSQL as \"irrelevant\", we\n*deserve* to get steamrollered by MySQL. \n\nI think it's terrific that Postgres is a real, programmer-centric,\ndemocratic Open Source project. I believe that programmers and\ncontributors should lead the project, and decide features and schedules\nbased on technical and not marketing reasons. Nobody on the Advocacy\nteam is trying to take control of the project and turn it into a\ndot-com.\n\nBut once Postgres has been packaged, we need to have a group making a\nloud enough noise to get the world to pay attention. I'm not asking\neveryone on this list to participate, but I am asking everyone on this\nlist to recognize the utility of the effort.\n\n-Josh Berkus\n\n\n\n\n\n\n", "msg_date": "Sun, 08 Dec 2002 22:54:47 -0800", "msg_from": "\"Josh Berkus\" <josh@agliodbs.com>", "msg_from_op": true, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" }, { "msg_contents": "On Sun, 8 Dec 2002, Josh Berkus wrote:\n\n> But once Postgres has been packaged, we need to have a group making a\n> loud enough noise to get the world to pay attention. I'm not asking\n> everyone on this list to participate, but I am asking everyone on this\n> list to recognize the utility of the effort.\n\nHere are my main problems with it.\n\n1) They're marketing to those that are already sold on it.\n2) They are, or at least were, insisting that I join their list to\n stay informed on what they're doing.\n3) They need to learn HOW to market from someone who knows (not me)\n how or they'll never be taken seriously.\n\nThat's all I'm going to say on this subject.\n\nVince.\n-- \n Fast, inexpensive internet service 56k and beyond! http://www.pop4.net/\n http://www.meanstreamradio.com http://www.unknown-artists.com\n Internet radio: It's not file sharing, it's just radio.\n\n", "msg_date": "Mon, 9 Dec 2002 07:29:41 -0500 (EST)", "msg_from": "Vince Vielhaber <vev@michvhf.com>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" }, { "msg_contents": "On Mon, 09 Dec 2002 07:29:55 -0500, Vince Vielhaber wrote:\n> On Sun, 8 Dec 2002, Josh Berkus wrote:\n> \n>> But once Postgres has been packaged, we need to have a group making a\n>> loud enough noise to get the world to pay attention. I'm not asking\n>> everyone on this list to participate, but I am asking everyone on this\n>> list to recognize the utility of the effort.\n> \n> Here are my main problems with it.\n> \n> 1) They're marketing to those that are already sold on it. \n\nI think we've already shown why it doesn't hurt to market to the\nconverted. I'll add that if you compare the 7.2 press release with the\n7.3 press release, you'll see none of the technical content was removed. \n\n> 2) They are,\n> or at least were, insisting that I join their list to\n> stay informed on what they're doing.\n\nI think it was only suggested that you join since you obviously have a\nlot of feedback you'd like to give to the group. Since a lot of people on\n-hackers don't want to be involved in the process, it seemed a bad idea\nto post all of the detail work to this list.\n\n> 3) They need to learn HOW to market from someone who knows (not me)\n> how or they'll never be taken seriously.\n> \n\nI've seen more posts saying that \"until you get a decent website your\nnot going to be taken seriously\" than anything else, by far. While I'm\nhoping that's not entirely true, I do agree that until we get a coordinated and\nopen web development process the advocacy group is going to have a much\nharder go of it.\n\nRobert Treat\n", "msg_date": "Mon, 09 Dec 2002 08:41:42 -0500", "msg_from": "Robert Treat <xzilla@users.sourceforge.net>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" }, { "msg_contents": "Vince Vielhaber <vev@michvhf.com> writes:\n> 1) They're marketing to those that are already sold on it.\n\nI think the upshot of the prior discussion was that the outside press\nrelease shouldn't have been used as the release announcement for the\nexisting mailing lists. Fine, they made a one-time mistake.\n\n> 2) They are, or at least were, insisting that I join their list to\n> stay informed on what they're doing.\n\nIt seems to me that people have made it perfectly clear that they don't\nwant to hear about marketing on the -hackers or -general lists. Taking\nit to a marketing-specific list seems like exactly the right response.\nWhere do you think it should be discussed?\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 10:31:41 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group " }, { "msg_contents": "Vince,\n\n> Here are my main problems with it.\n> \n> 1) They're marketing to those that are already sold on it.\n\nFirst off ... not \"they\", \"you\". I'm a member of Advocacy; so are\nRobert, Justin, Neil, Marc, Bruce and several other members of this\nlist. The advocacy group is not some privately sponsored bunch of\nmarketeers; *we* are your fellow contributors.\n\nYes, we should have released a different version of the announcement to\nthe internal lists. I believe that I have already explained how that\nhappened.\n\n> 2) They are, or at least were, insisting that I join their list to\n> stay informed on what they're doing.\n\nUnless you don't want to stay informed. In which case, you're welcome\nnot to, and one or more Advocacy people will join wwwdevel to keep\nlinks synchronized. Nobody's going to make you do anything. This is\nOpen Source.\n\n> 3) They need to learn HOW to market from someone who knows (not me)\n> how or they'll never be taken seriously.\n\nOne of our volunteers is a professional PR person. Two are periodical\nwriters. I started (with 2 partners) the OpenOffice.org Marketing\nProject, which was cited by one columnist (Amy Wohl) as a better\nvolunteer marketing team than Sun could put together for a\nmillion-dollar budget (paraphrased). 3 of us are small business\nowners. I think we have as much or more combined experience as the\nmarketing department of any start-up, without the baggage.\n\nAlso, half a marketing effort is better than none. At the very least,\nwe need to keep Postgres in the press, else we are likely to see\nPostgreSQL fade into permanent obscurity. The technology world is full\nof technically good but poorly marketed products -- FoxPro anyone?\nParadox? Beta video? Amiga?\n\nLast week I got a 5-page long database developer survey from EvansData.\n It mentioned 10 other database platforms -- including Ingres! -- but\nnot PostgreSQL. I personally don't want to see that again.\n\nSure, we got off to a rocky start. However, I will point out that our\nfirst release happened to fall on a major American holiday; this made\nit extra hard to organize the effort, and things didn't work out well.\n But the answer to that is not to abandon the effort, but to plan and\nprepare better in the future.\n\nI would also be grateful if us folks on the Advocacy team could look to\nHackers to make sure that we *aren't* going off on a tangent, or\npushing Postgres in a way that's inconsistent with the development\ngoals for the database. We *want* Advocacy to be an integral part of\nthe Postgres community, serving the general goal of making Postgres the\nbest possible ORDBMS in existence. \n\n-Josh Berkus\n\n", "msg_date": "Mon, 09 Dec 2002 09:15:46 -0800", "msg_from": "\"Josh Berkus\" <josh@agliodbs.com>", "msg_from_op": true, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" }, { "msg_contents": "Robert Treat writes:\n\n> I think we've already shown why it doesn't hurt to market to the\n> converted. I'll add that if you compare the 7.2 press release with the\n> 7.3 press release, you'll see none of the technical content was removed.\n\nCompare the 7.3 release notes, written for the most part by Bruce\nMomjian and revised by a couple of other developers, to the \"press\nrelease\", written by people who were obviously ill-informed.\n\nRelease notes:\n\n Schemas\n Schemas allow users to create objects in their own namespace so\n two people or applications can have tables with the same name.\n There is also a public schema for shared tables. Table/index\n creation can be restricted by removing permissions on the\n public schema.\n\nPress release:\n\n Schemas\n PostgreSQL now joins the handful of ORDBMS's to support\n the SQL 92 Schema specification, improving both enterprise\n database management and security through the use of namespaces.\n\nThis not only removes all information about the actual use of schemas,\nit contains completely bogus information, because SQL 92 is obsolete,\nthere is no \"SQL Schema specification\", and none of this has to do\nwith being an ORDBMS. And besides, whose hands were used to do the\ncounting?\n\n\nRelease notes:\n\n Drop Column\n PostgreSQL now supports the ALTER TABLE ... DROP COLUMN\n functionality.\n\nPress release:\n\n <void>\n\n\nRelease notes:\n\n Table Functions\n Functions returning multiple rows and/or multiple columns are\n now much easier to use than before. You can call such a \"table\n function\" in the SELECT FROM clause, treating its output like a\n table. Also, PL/pgSQL functions can now return sets.\n\nPress release:\n\n Table Functions\n PostgreSQL version 7.3 has greatly simplified returning result sets\n of rows and columns in database functions. This significantly\n enhances the useability of stored procedures in PostgreSQL, and will\n make it even easier to port Oracle applications to PostgreSQL.\n\nAgain, this removes all details about how the feature can be used, and\nagain it inserts completely bogus information. There are no \"sets of\ncolumns\", and PostgreSQL does not have stored procedures. Also, it\nmakes it look as though PostgreSQL exists merely to reimplement\nOracle.\n\n\nRelease notes:\n\n Prepared Queries\n PostgreSQL now supports prepared queries, for improved\n performance.\n\nPress release:\n\n - Prepared queries for maximized performance on common requests.\n\nI'm curious to know how the marketing department determined that this\nis, in fact, the maximal performance.\n\n\nRelease notes:\n\n Dependency Tracking\n PostgreSQL now records object dependencies, which allows\n improvements in many areas. \"DROP\" statements now take either\n CASCADE or RESTRICT to control whether dependent objects are\n also dropped.\n\nPress release:\n\n - Enhanced dependency tracking for complex databases.\n\nAgain, all relevant information dropped, replaced by marketing fluff.\n\n\nRelease notes:\n\n Privileges\n Functions and procedural languages now have privileges, and\n functions can be defined to run with the privileges of their\n creator.\n\nPress release:\n\n Security Advances\n In response to community demands, PostgreSQL has added schema,\n function, and other permissions and settings to increase the database\n administrator's granular control over security.\n\nInformation dropped, replaced by broad and repetitive verbiage. But\nat least they didn't write, \"in response to market pressures\".\n\n\nAnd my personal favorite is this:\n\nRelease notes:\n\n Internationalization\n Both multibyte and locale support are now always enabled.\n\nPress release:\n\n - Supports data in many international characters sets (UNICODE, EUC_JP,\n EUC_CN, EUC_KR, JOHAB, EUC_TW, ISO 8859-1 ECMA-94, KOI8, WIN1256, etc...)\n\nThat is just plain wrong. Support for various character sets is years\nold.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Mon, 9 Dec 2002 18:50:24 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" }, { "msg_contents": "On Monday 09 December 2002 12:50, Peter Eisentraut wrote:\n> Compare the 7.3 release notes, written for the most part by Bruce\n> Momjian and revised by a couple of other developers, to the \"press\n> release\", written by people who were obviously ill-informed.\n\nIf people want to see the details, let them read the release-notes themselves, \nand let it be the detail document. A press release of the detail that the \nrelease notes have will not get any 'press' -- and I say that wearing my \nradio broadcaster hat, where I have personally approved or disapproved 'press \nreleases' in news stories in the past. Getting 'press' is what a 'press \nrelease' is all about.\n\nSo, IMHO, the pgsql-announce mailing list should get the press release along \nwith the other 'outside' press outlets -- and the developers' lists (since \nhackers is far from the only one) should, IMHO again, get a copy of the \nrelease notes.\n\n> And my personal favorite is this:\n\n> Release notes:\n\n> Internationalization\n> Both multibyte and locale support are now always enabled.\n>\n> Press release:\n>\n> - Supports data in many international characters sets (UNICODE,\n> EUC_JP, EUC_CN, EUC_KR, JOHAB, EUC_TW, ISO 8859-1 ECMA-94, KOI8, WIN1256,\n> etc...)\n\n> That is just plain wrong. Support for various character sets is years\n> old.\n\nIt IS true that the current release supports all of these. The blanket \n'Supports' statement above quoted was not true in the blanket case until the \n'support' became default, since there were cases that this would not be true. \nSupport != 'if you pass the right parameters to configure this will work', at \nleast not at the press release level.\n-- \nLamar Owen\nWGCR Internet Radio\n1 Peter 4:11\n", "msg_date": "Mon, 9 Dec 2002 13:34:44 -0500", "msg_from": "Lamar Owen <lamar.owen@wgcr.org>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" }, { "msg_contents": "Peter Eisentraut <peter_e@gmx.net> writes:\n\n> Robert Treat writes:\n> \n> > I think we've already shown why it doesn't hurt to market to the\n> > converted. I'll add that if you compare the 7.2 press release with the\n> > 7.3 press release, you'll see none of the technical content was removed.\n> \n> Compare the 7.3 release notes, written for the most part by Bruce\n> Momjian and revised by a couple of other developers, to the \"press\n> release\", written by people who were obviously ill-informed.\n\n<snip for brevity>\n\nSo does this mean that you are volunteering to proofread the next\nmarketing announcement? I would wager that only a PostgreSQL\ndeveloper (such as yourself) could have picked out the inconsistencies\nthat you were able to find. The press release might have seemed\n\"obviously ill-informed\" to you, but it seemed just fine to me, and I\ncan guarantee you that I am at least an order of magnitude more\ninformed about PostgreSQL than the average manager.\n\nThe difference between the press release and the Release Notes is the\nintended audience. The folks that the press release is aimed at\nprobably don't have any idea that SQL 92 is obsolete, or that\ninternationalization has been supported for years. Chances are good\nthat they will skim over the new features entirely.\n\nWhat *is* important to these people, however, are the customer\ntestimonials at the beginning of the press release and the list of\nhappy customers at the end.\n\nOnce management has read the press release they can ask their\ndevelopers to read the Release Notes. Press releases don't supercede\nRelease Notes, they complement them. The difference between the 7.3\nRelease Notes and the press release is that I could give the press\nrelease to my boss.\n\nPostgreSQL desperately needs marketing help. In fact, at this point I\nwould say that PostgreSQL needs more marketing help than it needs\ndevelopment work. Technically PostgreSQL is clearly a winner, but\ndespite its myriad features and impressive performance PostgreSQL is\nstill not being deployed nearly as much as it *should* be. The team\nthat has been assembled to market PostgreSQL has some fairly\nimpressive credentials. They are certainly *much* better than what\nyou would expect considering how much they are getting paid :).\n\nIn short, if you want to help the folks writing the press releases,\nthen that's fine and dandy. But if all you want to do is throw rocks\nat the people doing the marketing, then that's another story\naltogether.\n\nJason Earl\n", "msg_date": "09 Dec 2002 11:36:25 -0700", "msg_from": "Jason Earl <jason.earl@simplot.com>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" }, { "msg_contents": "Jason Earl <jason.earl@simplot.com> writes:\n> Peter Eisentraut <peter_e@gmx.net> writes:\n>> Compare the 7.3 release notes, written for the most part by Bruce\n>> Momjian and revised by a couple of other developers, to the \"press\n>> release\", written by people who were obviously ill-informed.\n\n> So does this mean that you are volunteering to proofread the next\n> marketing announcement? I would wager that only a PostgreSQL\n> developer (such as yourself) could have picked out the inconsistencies\n> that you were able to find.\n\nFWIW, the press release looked fine to me too (and yes, I saw it in\nadvance).\n\n> The difference between the press release and the Release Notes is the\n> intended audience.\n\nExactly. The level of detail in the release notes is aimed at hackers\n(and usually gets criticized as \"insufficient\" by them ;-)), but a press\nrelease has entirely different purposes.\n\n> In short, if you want to help the folks writing the press releases,\n> then that's fine and dandy.\n\nOne error that I think the advocacy team made is that they didn't invite\nreview of the press release from a wider part of the community.\nAlthough I generally agree with the viewpoint that marketing issues\nshould be on a separate list and not on -hackers or -general, I think it\nwouldn't be out of place to send one message to those lists saying \"a\ndraft of the press release for <event FOO> is up at <this URL>, please\nsend comments to <advocacy mail list>.\" That seems like a reasonable\ncompromise between filling the lists with unwanted material and having\npeople feel that they were excluded from the process.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 14:14:52 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group " }, { "msg_contents": "Josh Berkus writes:\n\n> I can definitely understand someone not wanting to *participate* in\n> marketing/advocacy of PostgreSQL. However, your being opposed to\n> promoting PostgreSQL as an organized activity *at all* baffles me. How\n> can you be against promoting PostgreSQL?\n\nI'm not against promoting PostgreSQL. I'm against promoting PostgreSQL in\nways that embarrass me.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Mon, 9 Dec 2002 20:49:18 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" }, { "msg_contents": "\nPeter,\n\n> > I can definitely understand someone not wanting to *participate* in\n> > marketing/advocacy of PostgreSQL. However, your being opposed to\n> > promoting PostgreSQL as an organized activity *at all* baffles me. How\n> > can you be against promoting PostgreSQL?\n> \n> I'm not against promoting PostgreSQL. I'm against promoting PostgreSQL in\n> ways that embarrass me.\n\nWhat, specifically, were you embarassed by? \n\n-- \n-Josh Berkus\n Aglio Database Solutions\n San Francisco\n\n", "msg_date": "Mon, 9 Dec 2002 12:07:40 -0800", "msg_from": "Josh Berkus <josh@agliodbs.com>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" }, { "msg_contents": "Peter Eisentraut wrote:\n<snip>\n> Press release:\n> \n> - Supports data in many international characters sets (UNICODE, EUC_JP,\n> EUC_CN, EUC_KR, JOHAB, EUC_TW, ISO 8859-1 ECMA-94, KOI8, WIN1256, etc...)\n> \n> That is just plain wrong. Support for various character sets is years\n> old.\n\nSure is. Notice it didn't say \"just added\" or \"added with this release\"?\n\nIt just says \"supports\". It's to highlight the fact that it can be used \nfor non-English character sets. Sure, a whole bunch of people know \nthis, but the main target of the press release is people new to \nPostgreSQL that don't.\n\n:-)\n\nRegards and best wishes,\n\nJustin Clift\n\n-- \n\"My grandfather once told me that there are two kinds of people: those\nwho work and those who take the credit. He told me to try to be in the\nfirst group; there was less competition there.\"\n- Indira Gandhi\n\n", "msg_date": "Tue, 10 Dec 2002 11:02:00 +1100", "msg_from": "Justin Clift <justin@postgresql.org>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" } ]
[ { "msg_contents": "hi,\n When I check out the pgsql from cvs and I complile it, an error occured .\n\ndir: pgsql/src/interfaces/ecpg/preproc\nbison -y -d preproc.y\nerro information:\n preproc.y:5559: fatal error: maximum table size (32767) exceeded.\n\nHowever, I used the source from the ftp, find preproc.c in there. gmake will skip the\nstep(bison -y -d preproc.y) and succeeded.\n\nwho can tell me why?\n　　　　　　　\n　　　　　　　　　　2002-12-09\n\n\n\n", "msg_date": "Mon, 9 Dec 2002 14:58:19 +0800", "msg_from": "\"bigapple\" <bigapple__2002@163.com>", "msg_from_op": true, "msg_subject": "bison error" }, { "msg_contents": "On Mon, 2002-12-09 at 01:58, bigapple wrote:\n> When I check out the pgsql from cvs and I complile it, an error occured .\n> \n> dir: pgsql/src/interfaces/ecpg/preproc\n> bison -y -d preproc.y\n> erro information:\n> preproc.y:5559: fatal error: maximum table size (32767) exceeded.\n\nYou need to use Bison 1.50 or greater, due to a limitation in previous\nversions of Bison.\n\nCheers,\n\nNeil\n-- \nNeil Conway <neilc@samurai.com> || PGP Key ID: DB3C29FC\n\n\n\n", "msg_date": "09 Dec 2002 02:07:01 -0500", "msg_from": "Neil Conway <neilc@samurai.com>", "msg_from_op": false, "msg_subject": "Re: bison error" }, { "msg_contents": "bigapple wrote:\n> hi,\n> When I check out the pgsql from cvs and I complile it, an error occured .\n> \n> dir: pgsql/src/interfaces/ecpg/preproc\n> bison -y -d preproc.y\n> erro information:\n> preproc.y:5559: fatal error: maximum table size (32767) exceeded.\n> \n\nYou need at least version 1.5 of bison. Last time I checked, the latest out\nwas 1.75\n\n> However, I used the source from the ftp, find preproc.c in there. gmake will skip the\n> step(bison -y -d preproc.y) and succeeded.\n> \n> who can tell me why?\n\nSource distributions contain preprocessed files so that you can compile\nPostgreSQL without needing bison installed. I think you only need bison if you\nget source code from CVS (or want to hack the grammar in the source distribution).\n\nHTH,\n\nJoe\n\n", "msg_date": "Sun, 08 Dec 2002 23:08:18 -0800", "msg_from": "Joe Conway <mail@joeconway.com>", "msg_from_op": false, "msg_subject": "Re: bison error" } ]
[ { "msg_contents": "Does anybody know where I can find a newer DBD::Pg module for Windows\nNT? The only pre-compiled one I can find is 0.98, which is based on a\nPostgreSQL 7.0 library set.\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "09 Dec 2002 08:51:36 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": true, "msg_subject": "DBD::Pg module on Windows" } ]
[ { "msg_contents": "Attached is a patch against DBD::Pg 1.20 which\nfixes the driver-specific function 'table_attributes'\nfor use with PostgreSQL 7.3 while maintaining\nbackwards compatibility with older PostgreSQL\nversions.\n\nTo avoid voodoo with PostgreSQL version numbers\na check is made whether pg_relcheck exists and\nthe appropriate query (either 7.3 or pre 7.3)\nexecuted. Tested with 7.3 and 7.1.3.\n\nThis is hopefully a one off problem requiring this\nkind of check. I haven't had a chance to look at\nevery function in Pg.pm, but it seems this is the\nonly one \"broken\" by 7.3.\n\nComments, corrections etc. welcome!\n\nIan Barwick\nbarwick@gmx.net", "msg_date": "Mon, 9 Dec 2002 16:31:00 +0100", "msg_from": "Ian Barwick <barwick@gmx.net>", "msg_from_op": true, "msg_subject": "Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "Ian Barwick <barwick@gmx.net> writes:\n> To avoid voodoo with PostgreSQL version numbers\n> a check is made whether pg_relcheck exists and\n> the appropriate query (either 7.3 or pre 7.3)\n> executed.\n\nI would think that looking at version number (select version())\nwould be a much cleaner approach. Or do you think that direct\nexamination of pg_class is a version-independent operation?\n\nThis inquiry into pg_relcheck's existence is already arguably wrong\nin 7.3 (since it's not taking account of which schema pg_relcheck\nmight be found in) and it can only go downhill in future versions.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 11:03:52 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem " }, { "msg_contents": "On Monday 09 December 2002 17:03, Tom Lane wrote:\n> Ian Barwick <barwick@gmx.net> writes:\n> > To avoid voodoo with PostgreSQL version numbers\n> > a check is made whether pg_relcheck exists and\n> > the appropriate query (either 7.3 or pre 7.3)\n> > executed.\n>\n> I would think that looking at version number (select version())\n> would be a much cleaner approach. Or do you think that direct\n> examination of pg_class is a version-independent operation?\n\nNo, but I was hoping it will remain stable for long enough\nfor what is basically a temporary work around until a revised version of \nDBD::Pg can be produced. It doesn't make any more assumptions \nabout pg_class than are made elsewhere in the current Pg.pm.\n\n> This inquiry into pg_relcheck's existence is already arguably wrong\n> in 7.3 (since it's not taking account of which schema pg_relcheck\n> might be found in) and it can only go downhill in future versions.\n\nDoh. Knew I had to be missing something obvious. (Of course,\nanyone using current DBD::Pg with 7.3 as is will have to take\nextra care with system tables and schema namespaces anyway.)\n\nSo out with the candle wax and pins ;-). Am I right\nin thinking that the string returned by SELECT version()\nstarts with the word \"PostgreSQL\" followed by:\n a space; \n a single digit indicating the major version number;\n a full stop / decimal point;\n a single digit indicating the minor version number;\nand either \"interim release\" number (e.g. \".1\" in the case of 7.3.1), or\n\"devel\", \"rc1\" etc. ?\nAnd that this has been true since 6.x and will continue for the forseeable \nfuture (i.e. far far longer than the intended lifespan of attached patch)?\n\n\nIan Barwick\nbarwick@gmx.net\n\nAttached: revised patch", "msg_date": "Tue, 10 Dec 2002 00:34:31 +0100", "msg_from": "Ian Barwick <barwick@gmx.net>", "msg_from_op": true, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "Ian Barwick <barwick@gmx.net> writes:\n> So out with the candle wax and pins ;-). Am I right\n> in thinking that the string returned by SELECT version()\n> starts with the word \"PostgreSQL\" followed by:\n> a space;\n> a single digit indicating the major version number;\n> a full stop / decimal point;\n> a single digit indicating the minor version number;\n> and either \"interim release\" number (e.g. \".1\" in the case of 7.3.1), or\n> \"devel\", \"rc1\" etc. ?\n\nI would recommend parsing it the same way pg_dump does, see\n_check_database_version(). Looks like it's basically skip-11-chars-\nand-sscanf. Note the lack of assumptions about numbers of digits.\n\nIn the next protocol version update (hopefully 7.4) I would like to see\nthe basic version string (eg, \"7.3.1\" or \"7.4devel\") delivered to the\nclient automatically during connection startup and then available from a\nlibpq inquiry function. This would eliminate the need to call version()\nexplicitly and to know that you must skip \"PostgreSQL \" in its output.\nHowever, it will only help for clients/libraries that are willing to\ndeal exclusively with 7.4-or-newer backends, so it will take a few\nreleases to become really useful.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 18:47:30 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem " }, { "msg_contents": "\n(crossposting to hackers)\n\nOn Tuesday 10 December 2002 00:47, Tom Lane wrote:\n> In the next protocol version update (hopefully 7.4) I would like to see\n> the basic version string (eg, \"7.3.1\" or \"7.4devel\") delivered to the\n> client automatically during connection startup and then available from a\n> libpq inquiry function. This would eliminate the need to call version()\n> explicitly and to know that you must skip \"PostgreSQL \" in its output.\n\nSomething along the lines of \n char *PQversion(const PGconn *conn) ?\n\n> However, it will only help for clients/libraries that are willing to\n> deal exclusively with 7.4-or-newer backends, so it will take a few\n> releases to become really useful.\n\nSounds good to me. Is it on the todo-list? (Couldn't see it there).\n\nIan Barwick\nbarwick@gmx.net\n\n", "msg_date": "Tue, 10 Dec 2002 16:38:03 +0100", "msg_from": "Ian Barwick <barwick@gmx.net>", "msg_from_op": true, "msg_subject": "Re: [INTERFACES] Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "Ian Barwick <barwick@gmx.net> writes:\n> Sounds good to me. Is it on the todo-list? (Couldn't see it there).\n\nProbably not; Bruce for some reason has resisted listing protocol change\ndesires as an identifiable TODO category. There are a couple of threads\nin the pghackers archives over the last year or so that discuss the\ndifferent things we want to do, though. (Improving the error-reporting\nframework and fixing the COPY protocol are a couple of biggies I can\nrecall offhand.)\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Dec 2002 10:48:04 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: [INTERFACES] Patch for DBD::Pg pg_relcheck problem " }, { "msg_contents": "Ian Barwick writes:\n > On Tuesday 10 December 2002 00:47, Tom Lane wrote:\n > > In the next protocol version update (hopefully 7.4) I would like to see\n > > the basic version string (eg, \"7.3.1\" or \"7.4devel\") delivered to the\n > > client automatically during connection startup and then available from a\n > > libpq inquiry function. This would eliminate the need to call version()\n > > explicitly and to know that you must skip \"PostgreSQL \" in its output.\n > Something along the lines of \n > char *PQversion(const PGconn *conn) ?\n\nProbably:\n\n int PQversion(const PGconn *conn)\n\nwould be better, and easier to parse? For example the value returned\nfor 7.3.1 would be 7003001; for 7.4 7004000; for 101.10.2\n101010002. This allows simple numerical tests...\n\nLee.\n", "msg_date": "Tue, 10 Dec 2002 15:54:07 +0000", "msg_from": "Lee Kindness <lkindness@csl.co.uk>", "msg_from_op": false, "msg_subject": "Re: [INTERFACES] Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "(no followup to dbi-dev@perl.org, getting a little OT there)\n\nOn Tuesday 10 December 2002 16:54, Lee Kindness wrote:\n> Ian Barwick writes:\n> > Something along the lines of\n> > char *PQversion(const PGconn *conn) ?\n>\n> Probably:\n>\n> int PQversion(const PGconn *conn)\n>\n> would be better, and easier to parse? For example the value returned\n> for 7.3.1 would be 7003001; for 7.4 7004000; for 101.10.2\n> 101010002. This allows simple numerical tests...\n\nSounds logical - I was evidently thinking in Perl ;-). \n\nFor reference pg_dump currently parses the SELECT version() string\ninto an integer thus:\n\n7.2 70200\n7.2.1 70201\n7.3devel 70300\n7.3rc1 70300\n7.3.1 70301\n7.3.99 70399\n7.399.399 110299\n101.10.2 1011002\n\n(and just for fun:\n\"11i Enterprise Edition with Bells and Whistles \"\nreturns -1 ;-)\n\nwhich works with minor release numbers of 99\nand below.\n\nIan Barwick\nbarwick@gmx.net\n\n", "msg_date": "Tue, 10 Dec 2002 17:35:47 +0100", "msg_from": "Ian Barwick <barwick@gmx.net>", "msg_from_op": true, "msg_subject": "Re: [INTERFACES] Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "\nIan, can I please have a context diff, diff -c?\n\n---------------------------------------------------------------------------\n\nIan Barwick wrote:\n> On Monday 09 December 2002 17:03, Tom Lane wrote:\n> > Ian Barwick <barwick@gmx.net> writes:\n> > > To avoid voodoo with PostgreSQL version numbers\n> > > a check is made whether pg_relcheck exists and\n> > > the appropriate query (either 7.3 or pre 7.3)\n> > > executed.\n> >\n> > I would think that looking at version number (select version())\n> > would be a much cleaner approach. Or do you think that direct\n> > examination of pg_class is a version-independent operation?\n> \n> No, but I was hoping it will remain stable for long enough\n> for what is basically a temporary work around until a revised version of \n> DBD::Pg can be produced. It doesn't make any more assumptions \n> about pg_class than are made elsewhere in the current Pg.pm.\n> \n> > This inquiry into pg_relcheck's existence is already arguably wrong\n> > in 7.3 (since it's not taking account of which schema pg_relcheck\n> > might be found in) and it can only go downhill in future versions.\n> \n> Doh. Knew I had to be missing something obvious. (Of course,\n> anyone using current DBD::Pg with 7.3 as is will have to take\n> extra care with system tables and schema namespaces anyway.)\n> \n> So out with the candle wax and pins ;-). Am I right\n> in thinking that the string returned by SELECT version()\n> starts with the word \"PostgreSQL\" followed by:\n> a space; \n> a single digit indicating the major version number;\n> a full stop / decimal point;\n> a single digit indicating the minor version number;\n> and either \"interim release\" number (e.g. \".1\" in the case of 7.3.1), or\n> \"devel\", \"rc1\" etc. ?\n> And that this has been true since 6.x and will continue for the forseeable \n> future (i.e. far far longer than the intended lifespan of attached patch)?\n> \n> \n> Ian Barwick\n> barwick@gmx.net\n> \n> Attached: revised patch\n> \n> \n> \n> \n> \n> \n\n[ Attachment, skipping... ]\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 10 Dec 2002 22:49:12 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "Tom Lane wrote:\n> Ian Barwick <barwick@gmx.net> writes:\n> > Sounds good to me. Is it on the todo-list? (Couldn't see it there).\n> \n> Probably not; Bruce for some reason has resisted listing protocol change\n> desires as an identifiable TODO category. There are a couple of threads\n> in the pghackers archives over the last year or so that discuss the\n> different things we want to do, though. (Improving the error-reporting\n> framework and fixing the COPY protocol are a couple of biggies I can\n> recall offhand.)\n\nListing protocol changes seemed too low-level for the TODO list, but I\nhave kept the email messages. Today I updated the TODO list and added a\nsection for them.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 10 Dec 2002 22:50:37 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: [INTERFACES] Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "Perhaps compression should be added to the list of protocol changes. \nThis way, we can allow for per packet evaluation for compression.\n\n\n-- \nGreg Copeland <greg@copelandconsulting.net>\nCopeland Computer Consulting\n\n\nOn Tue, 2002-12-10 at 21:50, Bruce Momjian wrote:\n> Tom Lane wrote:\n> > Ian Barwick <barwick@gmx.net> writes:\n> > > Sounds good to me. Is it on the todo-list? (Couldn't see it there).\n> > \n> > Probably not; Bruce for some reason has resisted listing protocol change\n> > desires as an identifiable TODO category. There are a couple of threads\n> > in the pghackers archives over the last year or so that discuss the\n> > different things we want to do, though. (Improving the error-reporting\n> > framework and fixing the COPY protocol are a couple of biggies I can\n> > recall offhand.)\n> \n> Listing protocol changes seemed too low-level for the TODO list, but I\n> have kept the email messages. Today I updated the TODO list and added a\n> section for them.\n\n\n", "msg_date": "11 Dec 2002 13:44:42 -0600", "msg_from": "Greg Copeland <greg@CopelandConsulting.Net>", "msg_from_op": false, "msg_subject": "Re: [INTERFACES] Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "\nAdded to TODO.\n\n---------------------------------------------------------------------------\n\nGreg Copeland wrote:\n> Perhaps compression should be added to the list of protocol changes. \n> This way, we can allow for per packet evaluation for compression.\n> \n> \n> -- \n> Greg Copeland <greg@copelandconsulting.net>\n> Copeland Computer Consulting\n> \n> \n> On Tue, 2002-12-10 at 21:50, Bruce Momjian wrote:\n> > Tom Lane wrote:\n> > > Ian Barwick <barwick@gmx.net> writes:\n> > > > Sounds good to me. Is it on the todo-list? (Couldn't see it there).\n> > > \n> > > Probably not; Bruce for some reason has resisted listing protocol change\n> > > desires as an identifiable TODO category. There are a couple of threads\n> > > in the pghackers archives over the last year or so that discuss the\n> > > different things we want to do, though. (Improving the error-reporting\n> > > framework and fixing the COPY protocol are a couple of biggies I can\n> > > recall offhand.)\n> > \n> > Listing protocol changes seemed too low-level for the TODO list, but I\n> > have kept the email messages. Today I updated the TODO list and added a\n> > section for them.\n> \n> \n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 11 Dec 2002 19:19:27 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: [INTERFACES] Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "\nThanks. Patch applied. David, time to package up a new version of DBD:Pg?\n\n---------------------------------------------------------------------------\n\nIan Barwick wrote:\n> On Monday 09 December 2002 17:03, Tom Lane wrote:\n> > Ian Barwick <barwick@gmx.net> writes:\n> > > To avoid voodoo with PostgreSQL version numbers\n> > > a check is made whether pg_relcheck exists and\n> > > the appropriate query (either 7.3 or pre 7.3)\n> > > executed.\n> >\n> > I would think that looking at version number (select version())\n> > would be a much cleaner approach. Or do you think that direct\n> > examination of pg_class is a version-independent operation?\n> \n> No, but I was hoping it will remain stable for long enough\n> for what is basically a temporary work around until a revised version of \n> DBD::Pg can be produced. It doesn't make any more assumptions \n> about pg_class than are made elsewhere in the current Pg.pm.\n> \n> > This inquiry into pg_relcheck's existence is already arguably wrong\n> > in 7.3 (since it's not taking account of which schema pg_relcheck\n> > might be found in) and it can only go downhill in future versions.\n> \n> Doh. Knew I had to be missing something obvious. (Of course,\n> anyone using current DBD::Pg with 7.3 as is will have to take\n> extra care with system tables and schema namespaces anyway.)\n> \n> So out with the candle wax and pins ;-). Am I right\n> in thinking that the string returned by SELECT version()\n> starts with the word \"PostgreSQL\" followed by:\n> a space; \n> a single digit indicating the major version number;\n> a full stop / decimal point;\n> a single digit indicating the minor version number;\n> and either \"interim release\" number (e.g. \".1\" in the case of 7.3.1), or\n> \"devel\", \"rc1\" etc. ?\n> And that this has been true since 6.x and will continue for the forseeable \n> future (i.e. far far longer than the intended lifespan of attached patch)?\n> \n> \n> Ian Barwick\n> barwick@gmx.net\n> \n> Attached: revised patch\n> \n> \n> \n> \n> \n> \n\n[ Attachment, skipping... ]\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Sat, 14 Dec 2002 18:37:01 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "On Sunday 15 December 2002 00:37, Bruce Momjian wrote:\n> Thanks. Patch applied. David, time to package up a new version of DBD:Pg?\n\nSorry Bruce, I've missed a couple of days through illness and\nam just catching up. I think I was gong to submit a slightly different\nversion of the patch plus a couple of notes for the README file.\nCan you hold off 24 hours or so?\n\nIan Barwick\nbarwick@gmx.net\n\n", "msg_date": "Sun, 15 Dec 2002 00:53:57 +0100", "msg_from": "Ian Barwick <barwick@gmx.net>", "msg_from_op": true, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "Ian Barwick wrote:\n> On Sunday 15 December 2002 00:37, Bruce Momjian wrote:\n> > Thanks. Patch applied. David, time to package up a new version of DBD:Pg?\n> \n> Sorry Bruce, I've missed a couple of days through illness and\n> am just catching up. I think I was gong to submit a slightly different\n> version of the patch plus a couple of notes for the README file.\n> Can you hold off 24 hours or so?\n\nUntil we release it, the patch can just sit in CVS. Just send over a\nnew version and I will back out the old one and put in the new one. I\nwill not do any release until you are ready.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Sat, 14 Dec 2002 20:09:19 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "On Saturday, December 14, 2002, at 03:37 PM, Bruce Momjian wrote:\n\n> Thanks. Patch applied. David, time to package up a new version of \n> DBD:Pg?\n\nAfter we figure out what to do with NULLs, I think.\n\nBest,\n\nDavid\n\n-- \nDavid Wheeler AIM: dwTheory\ndavid@wheeler.net ICQ: 15726394\nhttp://david.wheeler.net/ Yahoo!: dew7e\n Jabber: Theory@jabber.org\n\n", "msg_date": "Sat, 14 Dec 2002 17:36:08 -0800", "msg_from": "David Wheeler <david@wheeler.net>", "msg_from_op": false, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "David Wheeler wrote:\n> On Saturday, December 14, 2002, at 03:37 PM, Bruce Momjian wrote:\n> \n> > Thanks. Patch applied. David, time to package up a new version of \n> > DBD:Pg?\n> \n> After we figure out what to do with NULLs, I think.\n\nI thought you had that figured out, by flagging the param as bytea\nsomehow? Are you thinking of something automatic?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Sat, 14 Dec 2002 20:46:53 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "On Saturday, December 14, 2002, at 05:46 PM, Bruce Momjian wrote:\n\n> I thought you had that figured out, by flagging the param as bytea\n> somehow? Are you thinking of something automatic?\n\nNo. Baldur Kristinsson points out that some people may pass a string \nwith NUL characters to DBD::Pg to be inserted into something other than \na bytea column. We have to figure out what to do when they do that. \nBaldur Kristinsson thinks we should just quietly strip them out (and \nthrow a warning). Others think that it should throw an exception and \ncroak. Tim said:\n\n The driver should always try to be as transparent as possible.\n The general principle is \"don't mes with the users data unless\n they've specifically asked you to (cf. ChopBlanks)\".\n\nSince, however, nulls aren't really allowed in any PostgreSQL data type \n(except bytea, and then only if specifically bound as such to a \nprepared statement), I'm not sure what to do about this. We can't leave \nthe data alone unless we just want PostgreSQL to throw an error (likely \nto be a mysterious error, as the user won't know why her data is \ngetting truncated).\n\nI think...throw an exception, since PostgreSQL can't handle the null \nbyte. Then it will be up to the user to clean up her data, and we won't \nhave to touch it.\n\nThoughts?\n\nDavid\n\n-- \nDavid Wheeler AIM: dwTheory\ndavid@wheeler.net ICQ: 15726394\nhttp://david.wheeler.net/ Yahoo!: dew7e\n Jabber: Theory@jabber.org\n\n", "msg_date": "Sat, 14 Dec 2002 17:58:17 -0800", "msg_from": "David Wheeler <david@wheeler.net>", "msg_from_op": false, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "David Wheeler wrote:\n> Since, however, nulls aren't really allowed in any PostgreSQL data type \n> (except bytea, and then only if specifically bound as such to a \n> prepared statement), I'm not sure what to do about this. We can't leave \n> the data alone unless we just want PostgreSQL to throw an error (likely \n> to be a mysterious error, as the user won't know why her data is \n> getting truncated).\n> \n> I think...throw an exception, since PostgreSQL can't handle the null \n> byte. Then it will be up to the user to clean up her data, and we won't \n> have to touch it.\n\nYep, throw an error, and maybe point to bytea as the solution, until we\nhave a better one. ;-)\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Sat, 14 Dec 2002 21:35:07 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "On Sunday 15 December 2002 02:09, Bruce Momjian wrote:\n>\n> Until we release it, the patch can just sit in CVS. Just send over a\n> new version and I will back out the old one and put in the new one. I\n> will not do any release until you are ready.\n\nHaving looked it again I think it can stay as it is.\n(It is not very elegant but will fix things until something\nbetter comes along).\n\nAttached is a supplemental patch with an addition for \nthe POD documentation in Pg.pm regarding the status\nof schema support in DBD::Pg. \n\n\nIan Barwick\nbarwick@gmx.net", "msg_date": "Mon, 16 Dec 2002 00:36:26 +0100", "msg_from": "Ian Barwick <barwick@gmx.net>", "msg_from_op": true, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "\nWhere are we on the release of a new DBDpg version? As I remember the\nonly open item is handling binary values, but at this point, maybe we\nshould just push out a release and fix it later.\n\n---------------------------------------------------------------------------\n\nBruce Momjian wrote:\n> \n> Thanks. Patch applied. David, time to package up a new version of DBD:Pg?\n> \n> ---------------------------------------------------------------------------\n> \n> Ian Barwick wrote:\n> > On Monday 09 December 2002 17:03, Tom Lane wrote:\n> > > Ian Barwick <barwick@gmx.net> writes:\n> > > > To avoid voodoo with PostgreSQL version numbers\n> > > > a check is made whether pg_relcheck exists and\n> > > > the appropriate query (either 7.3 or pre 7.3)\n> > > > executed.\n> > >\n> > > I would think that looking at version number (select version())\n> > > would be a much cleaner approach. Or do you think that direct\n> > > examination of pg_class is a version-independent operation?\n> > \n> > No, but I was hoping it will remain stable for long enough\n> > for what is basically a temporary work around until a revised version of \n> > DBD::Pg can be produced. It doesn't make any more assumptions \n> > about pg_class than are made elsewhere in the current Pg.pm.\n> > \n> > > This inquiry into pg_relcheck's existence is already arguably wrong\n> > > in 7.3 (since it's not taking account of which schema pg_relcheck\n> > > might be found in) and it can only go downhill in future versions.\n> > \n> > Doh. Knew I had to be missing something obvious. (Of course,\n> > anyone using current DBD::Pg with 7.3 as is will have to take\n> > extra care with system tables and schema namespaces anyway.)\n> > \n> > So out with the candle wax and pins ;-). Am I right\n> > in thinking that the string returned by SELECT version()\n> > starts with the word \"PostgreSQL\" followed by:\n> > a space; \n> > a single digit indicating the major version number;\n> > a full stop / decimal point;\n> > a single digit indicating the minor version number;\n> > and either \"interim release\" number (e.g. \".1\" in the case of 7.3.1), or\n> > \"devel\", \"rc1\" etc. ?\n> > And that this has been true since 6.x and will continue for the forseeable \n> > future (i.e. far far longer than the intended lifespan of attached patch)?\n> > \n> > \n> > Ian Barwick\n> > barwick@gmx.net\n> > \n> > Attached: revised patch\n> > \n> > \n> > \n> > \n> > \n> > \n> \n> [ Attachment, skipping... ]\n> \n> -- \n> Bruce Momjian | http://candle.pha.pa.us\n> pgman@candle.pha.pa.us | (610) 359-1001\n> + If your life is a hard drive, | 13 Roberts Road\n> + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Sun, 29 Dec 2002 23:57:32 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem" }, { "msg_contents": "\nPatch applied. Thanks.\n\n---------------------------------------------------------------------------\n\nIan Barwick wrote:\n> On Sunday 15 December 2002 02:09, Bruce Momjian wrote:\n> >\n> > Until we release it, the patch can just sit in CVS. Just send over a\n> > new version and I will back out the old one and put in the new one. I\n> > will not do any release until you are ready.\n> \n> Having looked it again I think it can stay as it is.\n> (It is not very elegant but will fix things until something\n> better comes along).\n> \n> Attached is a supplemental patch with an addition for \n> the POD documentation in Pg.pm regarding the status\n> of schema support in DBD::Pg. \n> \n> \n> Ian Barwick\n> barwick@gmx.net\n\n[ Attachment, skipping... ]\n\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 2: you can get off all lists at once with the unregister command\n> (send \"unregister YourEmailAddressHere\" to majordomo@postgresql.org)\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Sun, 29 Dec 2002 23:59:21 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Patch for DBD::Pg pg_relcheck problem" } ]
[ { "msg_contents": "\nI didn't read my email this weekend, so I am sorry to be late getting\nback to you on this.\n\nFirst, let me say I am excited about this patch. Several people have\nasked for IPv6 support, but you are the first person to actually submit\na patch for it.\n\nI want to comment on the patch a bit because IPv6 is usually a difficult\nfeature to add, considering the many platforms we support.\n\nFirst, I am running BSD/OS, so I have the bind 8.2.3 Inet routines in my\nlibc. However, I do not have IPv6 enabled in my kernel. Your patch\nwill have to deal with such systems, _and_ platforms that don't support\nIPv6 at all.\n\nIt appears the patch does a few things:\n\n\tchanges comparison of port->raddr.sa.sa_family == AF_INET to\n\t\tcall to a new function isAF_INETx\n\tchanges inet_aton() call to a new function SockAddr_pton()\n\textensively modifies StreamServerPort\n\tadds modifies StreamServerPortSubAFUNIX1/2\n\tmodifies connectDBStart\n\t\n\nTo better integrate your patch, I have:\n\n\treformatted to fit our coding style\n\trenamed file v6utils.c to ipv6.c\n\tremoved 'configure' link code to use Makefile link\n\t\tlike md5.c\n\tadded code to create IPv4 port if IPv6 port creation fails\n\nI need to continue reviewing the code and add a configure test to\ndetermine if IPv6 can compile on the machine. The fallback to IPv4\nshould be fine on machines that don't have IPv6 in their kernels.\n\nThe current patch can be downloaded from:\n\n\tftp://candle.pha.pa.us/pub/postgresql/mypatches/ipv6\n\n---------------------------------------------------------------------------\n\nNigel Kukard wrote:\n> Hrmmm,\n> \n> Is the compiling or running of postgresql a problem?\n> \n> if its the compiling we could add #ifdef's maybe to check\n> if we have INET6 support... or even add it to the configure.in,\n> if its running of postgresql which is the problem i have no clue\n> how to get around that.\n> \n> \n> thoughts?\n> \n> \n> On Fri, 6 Dec 2002, Bruce Momjian wrote:\n> \n> > \n> > I now have the INETv6 patch working using IPv4 on my machine, and I\n> > don't have IPv6 enabled in my kernel. Tomorrow, I will review the\n> > entire patch for portability issues, then post it so others can test it.\n> > I am pretty sure it is going to fail if your machine isn't INET6 aware,\n> > which may be many.\n> > \n> > ---------------------------------------------------------------------------\n> > \n> > > > The INETv6 patch was rejected because of this report, and an error on\n> > > > postmaster startup from BSD/OS:\n> > > > \n> > > > \tLOG: FATAL: StreamServerPort: getaddrinfo2() failed: hostname nor\n> > > > \tservname provided, or not known\n> > > > \n> > > > Please submit a new patch that addresses these issues. I can work with\n> > > > you to do testing.\n> > > > \n> > > > ---------------------------------------------------------------------------\n> > > > \n> > > > Joe Conway wrote:\n> > > > > Bruce Momjian wrote:\n> > > > > > Fixing now. This just isn't my night --- another patch with a missing\n> > > > > > file.\n> > > > > > \n> > > > > \n> > > > > OK - I can run configure and make now, but I'm getting these warnings:\n> > > > > \n> > > > > In file included from ../../../../src/include/libpq/libpq.h:22,\n> > > > > from printtup.c:20:\n> > > > > ../../../../src/include/libpq/v6util.h:3: warning: `struct addrinfo' declared \n> > > > > inside parameter list\n> > > > > ../../../../src/include/libpq/v6util.h:3: warning: its scope is only this \n> > > > > definition or declaration, which is probably not what you want.\n> > > > > ../../../../src/include/libpq/v6util.h:5: warning: `struct addrinfo' declared \n> > > > > inside parameter list\n> > > > > \n> > > > > lots of similar warnings to the above -- and:\n> > > > > \n> > > > > auth.c: In function `ClientAuthentication':\n> > > > > auth.c:414: warning: passing arg 1 of `isAF_INETx' from incompatible pointer type\n> > > > > gcc -O2 -g -Wall -Wmissing-prototypes -Wmissing-declarations \n> > > > > -I../../../src/include -c -o crypt.o crypt.c -MMD\n> > > > > In file included from ../../../src/include/libpq/libpq.h:22,\n> > > > > from crypt.c:24:\n> > > > > \n> > > > > Joe\n> > > > > \n> > > > > \n> > > > > ---------------------------(end of broadcast)---------------------------\n> > > > > TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org\n> > > > > \n> > > > \n> > > > \n> > > \n> > > -- \n> > > \n> > > \n> > > Nigel Kukard (Chief Executive Officer)\n> > > Lando Technologies Africa (Pty) Ltd\n> > > nigel@lando.co.za www.lando.co.za\n> > > Tel: 083 399 5822 Fax: 086 1100036\n> > > Hoheisen Park Bellville, Cape Town\n> > > National Internet Service Provider\n> > > \n> > > \n> > > The best language to use is the language that was designed for\n> > > what you want to use it for - 1997\n> > > \n> > > \n> > > =====================================================================\n> > > \n> > > Disclaimer\n> > > ----------\n> > > The contents of this message and any attachments are intended \n> > > solely for the addressee's use and may be legally privileged and/or \n> > > confidential information. This message may not be retained, \n> > > distributed, copied or used if you are not he addressee of this \n> > > message. If this message was sent to you in error, please notify \n> > > the sender immediately by reply e-mail and then destroy the message \n> > > and any copies thereof.\n> > > \n> > > Opinions, conclusions and other information in this message may be \n> > > personal to the sender and is not that of Lando Technologies Africa \n> > > or any of it's subsideries, associated companies or principals and \n> > > is therefore not endorsed by any of the Lando groups of companies. \n> > > Due to e-maill communication being insecure, Lando groups of \n> > > companies do not guarantee confidentiality, security, accuracy or \n> > > performance of the e-mail. Any liability for viruses is excluded \n> > > to the fullest extent.\n> > > \n> > > \n> > \n> > \n> \n> -- \n> \n> \n> Nigel Kukard (Chief Executive Officer)\n> Lando Technologies Africa (Pty) Ltd\n> nigel@lando.co.za www.lando.co.za\n> Tel: 083 399 5822 Fax: 086 1100036\n> Hoheisen Park Bellville, Cape Town\n> National Internet Service Provider\n> \n> \n> The best language to use is the language that was designed for\n> what you want to use it for - 1997\n> \n> \n> =====================================================================\n> \n> Disclaimer\n> ----------\n> The contents of this message and any attachments are intended \n> solely for the addressee's use and may be legally privileged and/or \n> confidential information. This message may not be retained, \n> distributed, copied or used if you are not he addressee of this \n> message. If this message was sent to you in error, please notify \n> the sender immediately by reply e-mail and then destroy the message \n> and any copies thereof.\n> \n> Opinions, conclusions and other information in this message may be \n> personal to the sender and is not that of Lando Technologies Africa \n> or any of it's subsideries, associated companies or principals and \n> is therefore not endorsed by any of the Lando groups of companies. \n> Due to e-maill communication being insecure, Lando groups of \n> companies do not guarantee confidentiality, security, accuracy or \n> performance of the e-mail. Any liability for viruses is excluded \n> to the fullest extent.\n> \n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Mon, 9 Dec 2002 13:08:29 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: IPv6 patch rejected" } ]
[ { "msg_contents": "Below is a short list of TODOs on sequences I wish to tackle over the\nnext week.\n\n\nCREATE SEQUENCE:\n- Addition of NO MAXVALUE and NO MINVALUE options, which use the system\nimplementation settings -- for SQL2002 compliance, and makes ALTER\nSEQUENCE slightly easier.\n\nALTER SEQUENCE:\n- Supports RESTART WITH, + options from Create Statement (including NO\nMAXVALUE, NO MINVALUE).\n\n\nModify init_params to deal with seq->options only. This allows\nAlterSequence to use it as well.\n\n\nOk, this is where it gets confusing. Right now setval() is implemented\nin such a manner that it cannot be rolled back (see SETVAL NOTE below),\nbut I'd like ALTER SEQUENCE to be transaction safe. Can I assume that\na standard simple_heap_update() is valid against the sequence, so long\nas I set xmin = FrozenTransactionId and create 2 XLog records similarly\nto DefineSequence?\n\nNow, do I need to do anything to clear the cache of other backends, or\nsimply let them play themselves out. I'm leaning towards the latter, as\nnextval() appears to read in the min / max value from the sequence\nbuffer.\n\n\nA transaction safe alter sequence, implemented in the standard method,\nwill result in two tuples. Doing this many times could make sequences\nquite slow. It looks like read_info() depends on a single value value\nin the sequence table. Do I need to do something more complex like a\nrelfileswap, generating a fresh tuple for it -- all parts of\nDefineSequence() except the DefineRelation() step? \n\n\nAnyway, I'll fiddle with the above two approaches, but would appreciate\ninput where appropriate.\n\n\n-- SETVAL NOTE --\na=# select nextval('test');\n nextval \n---------\n 2\n(1 row)\n\na=# begin;\nBEGIN\na=# select setval('test', 50);\n setval \n--------\n 50\n(1 row)\n\na=# select nextval('test');\n nextval \n---------\n 51\n(1 row)\n\na=# rollback;\nROLLBACK\na=# select nextval('test');\n nextval \n---------\n 52\n(1 row)\n\n\n\n\n\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "09 Dec 2002 15:49:50 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": true, "msg_subject": "Sequence Cleanup" }, { "msg_contents": "Rod Taylor kirjutas T, 10.12.2002 kell 01:49:\n> Below is a short list of TODOs on sequences I wish to tackle over the\n> next week.\n...\n> Ok, this is where it gets confusing. Right now setval() is implemented\n> in such a manner that it cannot be rolled back (see SETVAL NOTE below),\n> but I'd like ALTER SEQUENCE to be transaction safe.\n\nAll *val('seqname') functions are transaction-unsafe, i.e. live outside\ntransactions. \n\nWhy would you want alter transaction to be transaction safe ? \n\n\n-- \nHannu Krosing <hannu@tm.ee>\n", "msg_date": "10 Dec 2002 02:13:11 +0500", "msg_from": "Hannu Krosing <hannu@tm.ee>", "msg_from_op": false, "msg_subject": "Re: Sequence Cleanup" }, { "msg_contents": "Rod Taylor <rbt@rbt.ca> writes:\n> I'd like ALTER SEQUENCE to be transaction safe.\n\nI think that's inherently impossible without breaking the existing\nbehavior of setval/nextval, which is something we will not accept.\nALTER SEQUENCE would be better thought of as a form of setval with\neven more parameters, but not fundamentally different semantics.\n\n> Can I assume that\n> a standard simple_heap_update() is valid against the sequence,\n\nYou can be certain that it is not. There can be only one tuple in a\nsequence table.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 16:19:54 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Sequence Cleanup " } ]
[ { "msg_contents": "http://webhosting.yahoo.com/ps/wh/prod/p3.html\n\n\"MySQL® is considered the most popular open source database in the world. Fast and powerful, it is perfect for high-traffic, heavy-load sites.\"\n\n", "msg_date": "Mon, 9 Dec 2002 16:36:28 -0500", "msg_from": "\"Ned Lilly\" <ned@nedscape.com>", "msg_from_op": true, "msg_subject": "Yahoo hosting service using MySQL" } ]
[ { "msg_contents": "Using cvs source of Dec 4 15:13:\n\ntest=# \\d amount\n Table \"public.amount\"\n Column | Type | Modifiers \n--------+---------+--------------------------------------------------------\n id | integer | not null default nextval('public.amount_id_seq'::text)\n value | integer | \n\ntest=# select \"value\" from amount;\n value \n-------\n(0 rows)\n\ntest=# select value from amount;\nserver closed the connection unexpectedly\n This probably means the server terminated abnormally\n before or while processing the request.\nThe connection to the server was lost. Attempting reset: Failed.\n!#\n\n\nNot very graceful... (and this is different to the geqo prob I saw with my\nlibpq++ program, cured with geqo=false (Thanks Tom!))\n\nProgram received signal SIGSEGV, Segmentation fault.\ntransformTargetEntry (pstate=0x82e235c, node=0x82e21fc, expr=0x0, colname=0x0, \n resjunk=0 '\\000') at parse_target.c:61\n61 if (IsA(expr, RangeVar))\n(gdb) bt\n#0 transformTargetEntry (pstate=0x82e235c, node=0x82e21fc, expr=0x0, \n colname=0x0, resjunk=0 '\\000') at parse_target.c:61\n#1 0x80c2af0 in transformTargetList (pstate=0x82e235c, targetlist=0x82e2234)\n at parse_target.c:192\n#2 0x80a9c00 in transformSelectStmt (pstate=0x82e235c, stmt=0x82e22b4)\n at analyze.c:1654\n#3 0x80a7feb in transformStmt (pstate=0x82e235c, parseTree=0x82e22b4, \n extras_before=0xbfbfa9b0, extras_after=0xbfbfa9b4) at analyze.c:308\n#4 0x80a7bd0 in parse_analyze (parseTree=0x82e22b4, parentParseState=0x0)\n at analyze.c:147\n#5 0x813e1c0 in pg_analyze_and_rewrite (parsetree=0x82e22b4) at postgres.c:408\n#6 0x813e4a3 in pg_exec_query_string (query_string=0x82e201c, dest=Remote, \n parse_context=0x828c8ac) at postgres.c:696\n#7 0x813f935 in PostgresMain (argc=5, argv=0xbfbfaca8, \n username=0x826a925 \"prlw1\") at postgres.c:2016\n#8 0x8124332 in DoBackend (port=0x826a800) at postmaster.c:2293\n#9 0x8123c25 in BackendStartup (port=0x826a800) at postmaster.c:1915\n#10 0x8122e0b in ServerLoop () at postmaster.c:1002\n#11 0x8122972 in PostmasterMain (argc=3, argv=0x8266030) at postmaster.c:781\n#12 0x80fb135 in main (argc=3, argv=0xbfbfb4b4) at main.c:209\n#13 0x8069e84 in ___start ()\n\nCheers,\n\nPatrick\n", "msg_date": "Mon, 9 Dec 2002 23:39:17 +0000", "msg_from": "Patrick Welche <prlw1@newn.cam.ac.uk>", "msg_from_op": true, "msg_subject": "SIGSEGV" }, { "msg_contents": "Patrick Welche <prlw1@newn.cam.ac.uk> writes:\n> test=# select value from amount;\n> server closed the connection unexpectedly\n\nThis is a known bug also (in the domain-constraint patch, which has\nturned VALUE into a reserved word, a rather unpleasant price to pay\nfor the feature IMHO). Rod claimed his latest patch fixes it, but\nAFAIK that hasn't been applied yet.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 19:04:34 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: SIGSEGV " }, { "msg_contents": "On Mon, 2002-12-09 at 19:04, Tom Lane wrote:\n> Patrick Welche <prlw1@newn.cam.ac.uk> writes:\n> > test=# select value from amount;\n> > server closed the connection unexpectedly\n> \n> This is a known bug also (in the domain-constraint patch, which has\n> turned VALUE into a reserved word, a rather unpleasant price to pay\n> for the feature IMHO). Rod claimed his latest patch fixes it, but\n> AFAIK that hasn't been applied yet.\n\nIt was applied a couple of days ago by Bruce.\n\na=# select value;\nERROR: VALUE is not allowed in expression for node 732\n\nError message isn't great, but it doesn't crash anymore.\n\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "09 Dec 2002 19:48:30 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": false, "msg_subject": "Re: SIGSEGV" } ]
[ { "msg_contents": "Peter, Robert, Jason, Vince, Justin, et al.:\n\nFirst off, I'd like to ask everyone to CUT IT OUT WITH THE $^*&^@** FLAMING, \nALREADY! People are *attacking* each other instead of disagreeing. \nSeveral posters seem to be taking to opportunity to say everything in the \nmost insulting way possible, even when the actual source of disagreement is \nsmall. Perhaps we should declare a moritorium on this topic for 48 hours to \nlet everyone calm down?\n\nIn case we don't, my response:\n\nPETER, it's obvious that the press release team would have benefitted from \nyour copy-editing of the press release. You have several good points about \nplaces where we did not do the best possible job in the difficult task of \ntranslating technical notes into a form the general press would understand. \nI wrote a lot of the paragraphs you take issue with, and I don't deny that \nthey could stand improvement.\n\nWould you be willing to act as a reviewer on future press releases? That way, \nwe can get the benefit of your insight in a manner that will benefit the \npress release process.\n\n-- \n-Josh Berkus\n Aglio Database Solutions\n San Francisco\n\n", "msg_date": "Mon, 9 Dec 2002 16:16:53 -0800", "msg_from": "Josh Berkus <josh@agliodbs.com>", "msg_from_op": true, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" }, { "msg_contents": "Josh Berkus <josh@agliodbs.com> writes:\n> First off, I'd like to ask everyone to CUT IT OUT WITH THE $^*&^@** FLAMING, \n> ALREADY! People are *attacking* each other instead of disagreeing.\n\nAmen. This was first time 'round for the advocacy group, and it's not\nsurprising that there are some things they did wrong, or at least could\nhave done better. Can't we discuss the matter like a group of\nreasonably friendly people? I think we all have the same end goals\nin mind, so I don't see the need for unpleasantness.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 20:04:03 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group " }, { "msg_contents": "Tom Lane wrote:\n> Josh Berkus <josh@agliodbs.com> writes:\n> > First off, I'd like to ask everyone to CUT IT OUT WITH THE $^*&^@** FLAMING, \n> > ALREADY! People are *attacking* each other instead of disagreeing.\n> \n> Amen. This was first time 'round for the advocacy group, and it's not\n> surprising that there are some things they did wrong, or at least could\n> have done better. Can't we discuss the matter like a group of\n> reasonably friendly people? I think we all have the same end goals\n> in mind, so I don't see the need for unpleasantness.\n\nAgreed. Here's a story:\n\nMyself and a few people wanted live animals for a manger scene on our\nchurch lawn for Christmas. Many thought it was a bad idea, but we went\nahead anyway. It was a huge success, but then people complained we\ndidn't have enough people on the lawn to greet the hundreds of visitors.\n\nMoral of the story: if you take risks, expect folks to complain. And,\neven if you succeed, others will complain you didn't anticipate the\nsuccess.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Mon, 9 Dec 2002 20:47:06 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] PostgreSQL Global Development Group" } ]
[ { "msg_contents": "\nIt should have worked, but edit Makefile.shlib and remove that offending\nexport from the link line. That may fix it.\n\n---------------------------------------------------------------------------\n\nShibashish wrote:\n> Dear Sir,\n> \n> I use SCO Open Server 5.0.5 on an intel box. Although I have installed and\n> used PostgreSQL on Linux, setting it on SCO has not been successful :$\n> I have downloaded the latest version ie \"Postgresql-7.3\" I have also tried\n> installing postgresql-7.1 and postgresql-7.2.3, but never succeeded.\n> \n> I also installed \"ant\" package for using java. I have \"tcl8.0,\" \"tk8.0,\"\n> \"itclsh3.0\" and \"itkwish3.0\" installed in my system.\n> \"gcc version 2.7.2.1\"\n> \"Java 2 SDK, Standard Edition, v. 1.2.1\"\n> \"GNU Make 3.80\"\n> \n> My \"configure\" command was as following ...\n> ./configure --prefix=/data/pgsql --with-perl --with-tcl\n> --with-tclconfig=/data/tcl/lib/ --with-tkconfig=/data/tk/lib/\n> --with-java --without-readline --without-zlib\n> \n> The output has been attached as file \"configoutput.txt\"\n> \n> Then i give the \"make\" command. The compiling stops on an error and\n> exits after some time. The output has been attached as file\n> \"makeoutput.txt\"\n> \n> I'd be thankful to you if you can help me out sort the problem. I got your\n> mail-ids from the net and came to know that you are working on the similar\n> lines.\n> \n> Also, in the file doc/FAQ_SCO, does the section \"Compiling PostgreSQL 7.1\n> with the UDK\" apply to release 7.3 of Postgresql ? I am not using UDK.\n> \n> Waiting for a quick response from your end. kindly inform me if you have\n> already solved the problem, or whether any patch is available. Any\n> documentation or url will be highly helpful.\n> \n> Thanking You in anticipation.\n> \n> with regards\n> Shibashish\n> \n> shiba@it.iitb.ac.in\n> On Yahoo Messenger : shib_leo\n> \n> Software Engineer\n> IIT Bombay, India.\n\nContent-Description: \n\n[ Attachment, skipping... ]\n\nContent-Description: \n\n[ Attachment, skipping... ]\n\nContent-Description: \n\n[ Attachment, skipping... ]\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Mon, 9 Dec 2002 20:08:49 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: PostgreSQL 7.3 Installation on SCO" } ]
[ { "msg_contents": "Can anyone tell me how to get a tuple's TupleDesc if all that is known is the tid? Or is there an easy way to step through a tuple, retrieving the data and data type from each field?\n\nThanks,\nNate Sommer\n\n\n\n\n\n\n\nCan anyone tell me how to get a tuple's TupleDesc \nif all that is known is the tid?  Or is there an easy way to step through a \ntuple, retrieving the data and data type from each field?\n \nThanks,\nNate Sommer", "msg_date": "Mon, 9 Dec 2002 23:07:52 -0500", "msg_from": "\"Nate Sommer\" <sommena@earlham.edu>", "msg_from_op": true, "msg_subject": "tuple descriptors?" }, { "msg_contents": "\"Nate Sommer\" <sommena@earlham.edu> writes:\n> Can anyone tell me how to get a tuple's TupleDesc if all that is known is t=\n> he tid? Or is there an easy way to step through a tuple, retrieving the da=\n> ta and data type from each field?\n\nTupledescs are generally associated with tables (relations) more easily\nthan with specific tuples. What exactly is your context here?\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Dec 2002 23:20:07 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: tuple descriptors? " }, { "msg_contents": "> > Can anyone tell me how to get a tuple's TupleDesc if all that is known\nis t=\n> > he tid? Or is there an easy way to step through a tuple, retrieving the\nda=\n> > ta and data type from each field?\n>\n> Tupledescs are generally associated with tables (relations) more easily\n> than with specific tuples. What exactly is your context here?\n>\n> regards, tom lane\n\nI'm a student taking a database systems course, and as a project option I\nchose to work on one of PostgreSQL's todo list items, namely \"auto-delete\nlarge objects when referencing row is deleted.\" The main point of the\nproject is to become more comfortable tackling large amounts of code.\nWorking through the PostgreSQL code has been very interesting, and thus far\nI've worked independently, but now I'm getting a little overwhelmed. What\nI'd like to do is add some code to the heap_delete function that checks the\ntuple being deleted for oids, compares those oids to the loids in the\npg_largeobject relation, and deletes rows accordingly. I thought using\nTupleDescs would be helpful, but I guess my understanding of them is off.\n\nNate Sommer\n\n", "msg_date": "Mon, 9 Dec 2002 23:54:15 -0500", "msg_from": "\"Nate Sommer\" <sommena@earlham.edu>", "msg_from_op": true, "msg_subject": "Re: tuple descriptors? " }, { "msg_contents": "\"Nate Sommer\" <sommena@earlham.edu> writes:\n>> Tupledescs are generally associated with tables (relations) more easily\n>> than with specific tuples. What exactly is your context here?\n\n> What I'd like to do is add some code to the heap_delete function that\n> checks the tuple being deleted for oids, compares those oids to the\n> loids in the pg_largeobject relation, and deletes rows accordingly.\n\nAh. Well, heap_delete has trivial access to the appropriate tupledesc:\nrelation->rd_att (or more cleanly RelationGetDescr(relation)) gives it\nto you.\n\nNot sure how large a can of worms you wanted to open here, but some\ncreepy-crawlies I can finger offhand include:\n\n* don't forget heap_update's obsoleted tuple (but only when the\n replacement tuple contains a different LO oid).\n* [ extra credit ] don't forget heap_truncate. (If you can figure out\n how to do this bit without sacrificing the fundamental performance\n advantage of heap_truncate, then you're wasting your time dealing with\n us mere mortals...)\n* scanning pg_largeobject anytime someone wants to delete a tuple that\n includes an OID will be a serious performance hit, especially for\n updates on system catalogs --- it could even open the potential for\n deadlocks. Not to mention the obvious infinite-recursion problem:\n pg_largeobject itself has an OID column. Possibly you could finesse\n most of these issues by only doing the special processing for \"lo\"\n columns not \"oid\" columns, but that seems like a cheat. Is there a\n better way?\n* OIDs are not guaranteed unique across different system catalogs.\n Maybe there isn't a better way --- certainly deleting LO 42 because\n someone deleted pg_proc 42 wouldn't be happy-making. Within the\n catalogs we take care to know from context which catalog an OID must\n refer to, but a trigger that works on \"any OID column\" is at risk.\n\nYou've done pretty well already to identify heap_delete as a plausible\nplace to hack this, though. Soldier on ...\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Dec 2002 00:37:40 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: tuple descriptors? " }, { "msg_contents": "\"Nate Sommer\" <sommena@earlham.edu> writes:\n> I'm a student taking a database systems course, and as a project option I\n> chose to work on one of PostgreSQL's todo list items, namely \"auto-delete\n> large objects when referencing row is deleted.\" The main point of the\n> project is to become more comfortable tackling large amounts of code.\n\nBTW, I went to a pretty outstanding tutorial on just that topic this\npast summer at OSCON: Glenn Vanderburg on \"Using the Source: Software\nArchaeology for Users of Open Source Software\"\n(http://conferences.oreillynet.com/cs/os2002/view/e_sess/2869)\nwhich could be described in two seconds as \"don't be afraid to get\nyour hands dirty\", and in full length as how to be effective while\ndoing so.\n\nThe pikers at O'Reilly seem not to have made Glenn's slides available\non-line (isn't this a direct violation of their speakers' agreement?\nMine are there...) but I'm sure Glenn would be pleased to send 'em to\nyou on request.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Dec 2002 01:56:53 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: tuple descriptors? " }, { "msg_contents": "At 01:56 AM 10/12/2002 -0500, Tom Lane wrote:\n> but I'm sure Glenn would be pleased to send 'em to\n>you on request.\n\nDo you have an email address - the O'Reilly site also seems not to have one...\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Tue, 10 Dec 2002 18:05:45 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": false, "msg_subject": "Re: tuple descriptors? " }, { "msg_contents": "Philip Warner <pjw@rhyme.com.au> writes:\n> At 01:56 AM 10/12/2002 -0500, Tom Lane wrote:\n>> but I'm sure Glenn would be pleased to send 'em to\n>> you on request.\n\n> Do you have an email address - the O'Reilly site also seems not to have one...\n\nHrm, you're right. I think there was one in the hardcopy slides, but\nthat's not beside me at the moment. Before we deluge Glenn with\nrequests, let me ask him for a URL on behalf of the group ...\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Dec 2002 02:11:11 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: tuple descriptors? " } ]
[ { "msg_contents": "I have had similar troubles, related to oid overflow. I had to modify pg_dump \nto properly cast queries that contain oids. This is against 7.1.3 source. The \npatch was hacked quickly, in order to get a corrupted database reloaded, and \nthis while I was traveling in another country... so it is far from perfect but \nsaved my database(s). It also fixes other oid-related troubles of pg_dump.\n\nSee attached file.\n\nDaniel\n\n>>>Brian Fujito said:\n > Thanks for your input--\n > \n > I've tried both ways:\n > \n > createlang/droplang from the command line as user postgres\n > \n > and:\n > \n > CREATE FUNCTION plpgsql_call_handler () RETURNS OPAQUE AS\n > '/usr/lib/pgsql/plpgsql.so' LANGUAGE 'C';\n > \n > CREATE TRUSTED PROCEDURAL LANGUAGE 'plpgsql'\n > HANDLER plpgsql_call_handler\n > LANCOMPILER 'PL/pgSQL';\n > \n > I'm using pg_dump (not pg_dumpall) on the specific database on which\n > I created the language.\n > \n > I realize 7.0.3 is ancient (same with 7.1)... I just don't have the\n > time to deal with an upgrade quite yet. Soon enough :) In the mean\n > time, a stop-gap solution would definitely be appreciated.\n > \n > Thank you,\n > Brian\n > \n > \n > On Mon, 2002-12-09 at 14:28, Tom Lane wrote:\n > > Brian Fujito <brian@lightsource.com> writes:\n > > > I'm running postgres-7.0.3 on RH7.1 using RedHat RPMs.\n > > > I recently added plpgsql as a language to one of my databases, \n > > > and now when I try to do a pg_dump, I get:\n > > \n > > > \"dumpProcLangs(): handler procedure for language plpgsql not found\"\n > > \n > > > If I drop the language, pg_dump works fine, but if I add it back (and\n > > > even if I restart postgres), I get the same error.\n > > \n > > What exactly are you doing to drop and re-add the language? I should\n > > think CREATE LANGUAGE would fail if the handler proc isn't there.\n > > \n > > (Also, are you doing pg_dump or pg_dumpall? If the latter, maybe the\n > > failure is occurring in a different database than the one you are\n > > changing.)\n > > \n > > But having said that, 7.0.3 is ancient history ... you really are\n > > overdue for an upgrade. With my Red Hat fedora on, I'd say the same\n > > about your choice of OS version too.\n > > \n > > \t\t\tregards, tom lane\n > \n > \n > \n > ---------------------------(end of broadcast)---------------------------\n > TIP 3: if posting/reading through Usenet, please send an appropriate\n > subscribe-nomail command to majordomo@postgresql.org so that your\n > message can get through to the mailing list cleanly", "msg_date": "Tue, 10 Dec 2002 09:06:17 +0200", "msg_from": "Daniel Kalchev <daniel@digsys.bg>", "msg_from_op": true, "msg_subject": "Re: \"dumpProcLangs(): handler procedure for language " } ]
[ { "msg_contents": "At 01:56 AM 10/12/2002 -0500, Tom Lane wrote:\n> but I'm sure Glenn would be pleased to send 'em to\n>you on request.\n\nI've found a link:\n\n http://www.delphis.com/java/java.html\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Tue, 10 Dec 2002 18:08:25 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: tuple descriptors? " } ]
[ { "msg_contents": "Sorry for this being so long but I want to describe this as thoroughly as possible.\n\nI am having two problems with foreign keys. One is a performance problem and the other is a deadlock problem but they are related to the same root cause. I am running PostgreSQL 7.3 (the released version).\n\nI have isolated it down to a simple test:\n\nGiven the following database:\n\ncreate table names (\n id integer not null,\n name text not null,\n primary key (id)\n);\n\ncreate table orders (\n id integer not null,\n nameid integer,\n description text,\n primary key (id)\n);\nalter table orders add constraint oname foreign key(nameid) references names(id);\n\ninsert into names values (1,'bob');\ninsert into names values (2,'fred');\ninsert into names values (3,'sam');\n\ninsert into orders values (1,1,'bob1');\ninsert into orders values (2,1,'bob2');\ninsert into orders values (3,1,'bob3');\ninsert into orders values (4,2,'fred1');\ninsert into orders values (5,3,'sam1');\n\nTo reproduce the bug, start psql on the database in two different shells.\n\nIn shell A:\n\nbegin;\nupdate orders set description='bob1-1' where id=1;\n\n\nIn shell B:\n\nbegin;\nupdate orders set description='bob2-1' where id=2;\n\n\nThe update in shell B will blocuntilll you do a \"commit;\" or \"rollback;\" in shell A. This blocking should not occur.\n\nThe problem is that the update in shell A causes a\n\nSELECT 1 FROM ONLY \"public\".\"names\" x WHERE \"id\" = $1 FOR UPDATE OF x\n\nstatement to be generated internally as part of the foreign key checking. For shell A this works fine but when shell B executes this line it blocks until the transaction in shell A does a commit or rollback.\n\nThe purpose of this SELECT seems to be two-fold:\n 1. To make sure that row in the target table exists.\n2. To make sure that the row does not get deleted or that column in that row does not get changed until the commit happens because other transactions cannot see the changes until the commit happens.\n\nAs a test I went into \"backend/utils/adt/ri_triggers.c\" and removed the \"FOR UPDATE OF X\" from the foreign key checks and the concurrency issues disappeared. This still make check 1 happen but removed the safety net of check 2.\n\nThe \"FOR UPDATE OF X\" seems to grab a lock that cannot be shared so the second foreign key select must wait until the first one releases. Is there a weaker lock that can applied to the foreign key check in ri_triggers.c? Is a new type of lock \"FOR FKEY OF X\" required?\n\nThis really drags down our system when we get alot of traffic. It also also causes deadlocks.\n\nDEADLOCK\n--------\n\nThe example is a very simple case but in my application where I have more tables and multiple foreign keys I run into deadlocks.\n\nIn the simplest case I have multiple \"information\" tables that are the targets of foreign keys. I have 2 \"data\" tables that have foreign keys into the information tables. If I am inserting/updating rows in tables \"data1\" and \"data2\".\n\nHere is an example I made up to (hopefully) make this clear:\n\ncreate table names (\n id integer not null,\n name text not null,\n primary key (id)\n);\n\ncreate table billaddr (\n id integer not null,\n address text not null,\n primary key (id)\n);\n\ncreate table shipaddr (\n id integer not null,\n address text not null,\n primary key (id)\n);\n\ncreate table phone_orders (\n id integer not null,\n nameid integer,\n billid integer,\n shipid integer,\n description text,\n primary key (id)\n);\nalter table phone_orders add constraint poname foreign key(nameid) references names(id);\nalter table phone_orders add constraint pobaddr foreign key(billid) references billaddr(id);\nalter table phone_orders add constraint posaddr foreign key(shipid) references shipaddr(id);\n\ncreate table web_orders (\n id integer not null,\n nameid integer,\n billid integer,\n shipid integer,\n description text,\n primary key (id)\n);\nalter table web_orders add constraint woname foreign key(nameid) references names(id);\nalter table web_orders add constraint wobaddr foreign key(billid) references billaddr(id);\nalter table web_orders add constraint wosaddr foreign key(shipid) references shipaddr(id);\n\ninsert into names values (1,'bob');\ninsert into names values (2,'fred');\ninsert into names values (3,'sam');\n\ninsert into billaddr values (1,'123 main st');\ninsert into billaddr values (2,'456 minor ave');\n\ninsert into shipaddr values (1,'789 major ct');\ninsert into shipaddr values (2,'912 losers lane');\n\ninsert into phone_orders values (1,1,1,1,'phone order 1');\ninsert into phone_orders values (2,2,2,2,'phone order 2');\n\ninsert into web_orders values (1,1,1,1,'web order 1');\ninsert into web_orders values (2,2,2,2,'web order 2');\n\n\nOnce again start psql on the database in two different shells.\n\nIn shell A:\n\nbegin;\nupdate phone_orders set description='phone order 1-1' where id=1;\n\n\nIn shell B:\n\nbegin;\nupdate web_orders set description='web order 1-1' where id=1;\n\nIf the PostgreSQL server acquires the foreign key locks in a different order on the web-orders and phone_orders tables then you will get a deadlock. When I ran this exact case I did not see this but I have seen it when running my application. I reorganized by schema to try to get the locks to be acquired in the same order all the time but I could not get it to work. Even if I could get this to work it would not solve the performance issue.\n\n\nI saw a discussion on this from March:\n\nhttp://archives.postgresql.org/pgsql-hackers/2002-03/msg01156.php\n\nThese does not seem to be a resolution to this yet.\n\nAre there any plans to fix this soon? This is a serious problem for us.\n\nThanks\n\nBrian Walker\n\n\n\n\n\n\n\n", "msg_date": "Tue, 10 Dec 2002 10:47:16 -0600", "msg_from": "\"Brian Walker\" <brianw@mcsdallas.com>", "msg_from_op": true, "msg_subject": "Problem with foreign keys (performance and deadlocks)" }, { "msg_contents": "On Tue, 10 Dec 2002, Brian Walker wrote:\n\n> As a test I went into \"backend/utils/adt/ri_triggers.c\" and removed\n> the \"FOR UPDATE OF X\" from the foreign key checks and the concurrency\n> issues disappeared. This still make check 1 happen but removed the\n> safety net of check 2.\n\nRight, at which point you lose the guarantees of the constraint.\n\n> The \"FOR UPDATE OF X\" seems to grab a lock that cannot be shared so\n> the second foreign key select must wait until the first one releases.\n> Is there a weaker lock that can applied to the foreign key check in\n> ri_triggers.c? Is a new type of lock \"FOR FKEY OF X\" required?\n\nNot in the same way and sort of (it's more involved than a new kind of\nlock).\n\n> I saw a discussion on this from March:\n>\n> http://archives.postgresql.org/pgsql-hackers/2002-03/msg01156.php\n>\n> These does not seem to be a resolution to this yet.\n>\n> Are there any plans to fix this soon? This is a serious problem for us.\n\nThere are plans to fix it. Right now, soon is relative to how much time I\ncan spend on it. I think a version of my very early testing patch ended\nup on -general (with one bug that was mentioned afterwards that has an\neasy fix) that I was asking for information on. As I say in that message,\nit's not a prime time patch since at the last it lets through bad data\n(less so than removing the for update from the original patch) and it\nreally needs better management of its local data and it's really ugly. I\nhaven't had any time recently, but I'm hoping to get a new testing patch\nout by early January.\n\n", "msg_date": "Tue, 10 Dec 2002 10:22:36 -0800 (PST)", "msg_from": "Stephan Szabo <sszabo@megazone23.bigpanda.com>", "msg_from_op": false, "msg_subject": "Re: Problem with foreign keys (performance and deadlocks)" }, { "msg_contents": "\nI posted this message to psql-bugs in December but I have not heard if this has been fixed or not. This is holding up our development and testing of our product using Linux and PostgreSQL.\n\nThanks\n\n\n\n********** ORIGINAL MESSAGE ***************\n\nSorry for this being so long but I want to describe this as thoroughly as possible.\n\nI am having two problems with foreign keys. One is a performance problem and the other is a deadlock problem but they are related to the same root cause. I am running PostgreSQL 7.3 (the released version).\n\nI have isolated it down to a simple test:\n\nGiven the following database:\n\ncreate table names (\n id integer not null,\n name text not null,\n primary key (id)\n);\n\ncreate table orders (\n id integer not null,\n nameid integer,\n description text,\n primary key (id)\n);\nalter table orders add constraint oname foreign key(nameid) references names(id);\n\ninsert into names values (1,'bob');\ninsert into names values (2,'fred');\ninsert into names values (3,'sam');\n\ninsert into orders values (1,1,'bob1');\ninsert into orders values (2,1,'bob2');\ninsert into orders values (3,1,'bob3');\ninsert into orders values (4,2,'fred1');\ninsert into orders values (5,3,'sam1');\n\nTo reproduce the bug, start psql on the database in two different shells.\n\nIn shell A:\n\nbegin;\nupdate orders set description='bob1-1' where id=1;\n\n\nIn shell B:\n\nbegin;\nupdate orders set description='bob2-1' where id=2;\n\n\nThe update in shell B will blocuntilll you do a \"commit;\" or \"rollback;\" in shell A. This blocking should not occur.\n\nThe problem is that the update in shell A causes a\n\nSELECT 1 FROM ONLY \"public\".\"names\" x WHERE \"id\" = $1 FOR UPDATE OF x\n\nstatement to be generated internally as part of the foreign key checking. For shell A this works fine but when shell B executes this line it blocks until the transaction in shell A does a commit or rollback.\n\nThe purpose of this SELECT seems to be two-fold:\n 1. To make sure that row in the target table exists.\n2. To make sure that the row does not get deleted or that column in that row does not get changed until the commit happens because other transactions cannot see the changes until the commit happens.\n\nAs a test I went into \"backend/utils/adt/ri_triggers.c\" and removed the \"FOR UPDATE OF X\" from the foreign key checks and the concurrency issues disappeared. This still make check 1 happen but removed the safety net of check 2.\n\nThe \"FOR UPDATE OF X\" seems to grab a lock that cannot be shared so the second foreign key select must wait until the first one releases. Is there a weaker lock that can applied to the foreign key check in ri_triggers.c? Is a new type of lock \"FOR FKEY OF X\" required?\n\nThis really drags down our system when we get alot of traffic. It also also causes deadlocks.\n\nDEADLOCK\n--------\n\nThe example is a very simple case but in my application where I have more tables and multiple foreign keys I run into deadlocks.\n\nIn the simplest case I have multiple \"information\" tables that are the targets of foreign keys. I have 2 \"data\" tables that have foreign keys into the information tables. If I am inserting/updating rows in tables \"data1\" and \"data2\".\n\nHere is an example I made up to (hopefully) make this clear:\n\ncreate table names (\n id integer not null,\n name text not null,\n primary key (id)\n);\n\ncreate table billaddr (\n id integer not null,\n address text not null,\n primary key (id)\n);\n\ncreate table shipaddr (\n id integer not null,\n address text not null,\n primary key (id)\n);\n\ncreate table phone_orders (\n id integer not null,\n nameid integer,\n billid integer,\n shipid integer,\n description text,\n primary key (id)\n);\nalter table phone_orders add constraint poname foreign key(nameid) references names(id);\nalter table phone_orders add constraint pobaddr foreign key(billid) references billaddr(id);\nalter table phone_orders add constraint posaddr foreign key(shipid) references shipaddr(id);\n\ncreate table web_orders (\n id integer not null,\n nameid integer,\n billid integer,\n shipid integer,\n description text,\n primary key (id)\n);\nalter table web_orders add constraint woname foreign key(nameid) references names(id);\nalter table web_orders add constraint wobaddr foreign key(billid) references billaddr(id);\nalter table web_orders add constraint wosaddr foreign key(shipid) references shipaddr(id);\n\ninsert into names values (1,'bob');\ninsert into names values (2,'fred');\ninsert into names values (3,'sam');\n\ninsert into billaddr values (1,'123 main st');\ninsert into billaddr values (2,'456 minor ave');\n\ninsert into shipaddr values (1,'789 major ct');\ninsert into shipaddr values (2,'912 losers lane');\n\ninsert into phone_orders values (1,1,1,1,'phone order 1');\ninsert into phone_orders values (2,2,2,2,'phone order 2');\n\ninsert into web_orders values (1,1,1,1,'web order 1');\ninsert into web_orders values (2,2,2,2,'web order 2');\n\n\nOnce again start psql on the database in two different shells.\n\nIn shell A:\n\nbegin;\nupdate phone_orders set description='phone order 1-1' where id=1;\n\n\nIn shell B:\n\nbegin;\nupdate web_orders set description='web order 1-1' where id=1;\n\nIf the PostgreSQL server acquires the foreign key locks in a different order on the web-orders and phone_orders tables then you will get a deadlock. When I ran this exact case I did not see this but I have seen it when running my application. I reorganized by schema to try to get the locks to be acquired in the same order all the time but I could not get it to work. Even if I could get this to work it would not solve the performance issue.\n\n\nI saw a discussion on this from March:\n\nhttp://archives.postgresql.org/pgsql-hackers/2002-03/msg01156.php\n\nThese does not seem to be a resolution to this yet.\n\nAre there any plans to fix this soon? This is a serious problem for us.\n\nThanks\n\nBrian Walker\n\n\n\n", "msg_date": "Tue, 04 Mar 2003 13:21:15 -0600", "msg_from": "\"Brian Walker\" <brianw@mcsdallas.com>", "msg_from_op": true, "msg_subject": "Problem with foreign keys (performance and deadlocks)" } ]
[ { "msg_contents": "I've been looking at the recently-committed ALTER DOMAIN patch, and I\nthink it's got some serious if not fatal problems. Specifically, the\napproach to adding/dropping constraints associated with domains doesn't\nwork.\n\n1. Insufficient locking, guise 1: there's no protection against someone\nelse dropping a column or whole table between the time you find a\npg_attribute entry in get_rels_with_domain and the time you actually\nprocess it in AlterDomainNotNull or AlterDomainAddConstraint. The code\nappears to think that holding RowExclusiveLock on pg_attribute will\nprotect it somehow, but that doesn't (and shouldn't) do any such thing.\nThis will result in at best an elog and at worst coredump when you try\nto scan the no-longer-present table or column.\n\n2. Insufficient locking, guise 2: there's no protection against someone\nelse adding a column or table while you're processing an ALTER DOMAIN,\neither. This means that constraint checks will be missed. Example:\n\n<< backend 1 >>\n\nregression=# create domain mydom int4;\nCREATE DOMAIN\nregression=# begin;\nBEGIN\nregression=# alter domain mydom set not null;\nALTER DOMAIN\n\n<< don't commit yet; in backend 2 do >>\n\nregression=# create table foo (f1 mydom);\nCREATE TABLE\nregression=# insert into foo values(null);\nINSERT 149688 1\n\n<< now in backend 1: >>\n\nregression=# commit;\nCOMMIT\n\n<< now in backend 2: >>\n\nregression=# insert into foo values(null);\nERROR: Domain mydom does not allow NULL values\nregression=# select * from foo;\n f1\n----\n\n(1 row)\n\nNot a very watertight domain constraint, is it? The begin/commit is not\nnecessary to cause a failure, it just makes it easy to make the window\nfor failure wide enough to hit in a manually entered example.\n\n3. Too much locking, guise 1: the ALTER DOMAIN command will acquire\nexclusive lock on every table that it scans, and will hold all these\nlocks until it commits. This can easily result in deadlocks --- against\nother ALTER DOMAIN commands, or just against any random other\ntransaction that is unlucky enough to try to write any two tables\ntouched by the ALTER DOMAIN. AFAICS you don't need an exclusive lock,\nyou just want to prevent updates of the table until the domain changes\nare committed, so ShareLock would be sufficient; that would reduce but\nnot eliminate the risk of deadlock.\n\n4. Too much locking, guise 2: the ExclusiveLock acquired on pg_class by\nget_rels_with_domain has no useful effect, since it's released again\nat the end of the scan; it does manage to shut down most sorts of schema\nchanges while get_rels_with_domain runs, however. This is bad enough,\nbut:\n\n5. Performance sucks. In the regression database on my machine, \"alter\ndomain mydom set not null\" takes over six seconds --- that's for a\nfreshly created domain that's not used *anywhere*. This can be blamed\nentirely on the inefficient implementation of get_rels_with_domain.\nIn a database with more tables performance would get much worse; it's\nbasically O(N^2). And it's holding ExclusiveLock on pg_class the whole\ntime :-(. (A reasonably efficient way to make the same search would be\nto use pg_depend to look for columns that depend on the domain type ---\nthis might find a few indirect dependencies, but it would certainly be\nlots faster than repeated seqscans over pg_attribute.)\n\n6. Permission bogosity: as per discussion yesterday, ownership of a\nschema does not grant ownership rights on contained objects.\n\n7. No mechanism for causing constraint changes to actually propagate\nafter they are made. This is more a fault of the design of the domain\nconstraint patch than it is of the alter patch, but nonetheless alter is\nwhat exposes it. The problem is particularly acute because you chose to\ninsert a domain's constraint expressions into coercion operations at\nexpression parsing time, which is far too early. A stored rule that has\na coerce-to-domain operation in it will have a frozen idea of what\nconstraints it should be enforcing. Probably the expression tree should\njust have a \"CoerceToDomain foo\" node in it, and at executor startup\nthis node would have to look to the pg_type entry for foo to see exactly\nwhat it should be enforcing at the moment.\n\n\nSome of these are fixable, but I don't actually see any fix for point 2\nshort of creating some entirely new locking convention. Currently, only\nrelations can be locked, but you'd really need an enforceable lock on\nthe type itself to make a watertight solution, I think. Since we've\nnever had any sort of supported ALTER TYPE operation before, the issue\nhasn't come up before ...\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Dec 2002 12:39:52 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": true, "msg_subject": "Problems with ALTER DOMAIN patch" }, { "msg_contents": "On Tue, 2002-12-10 at 12:39, Tom Lane wrote:\n> I've been looking at the recently-committed ALTER DOMAIN patch, and I\n> think it's got some serious if not fatal problems. Specifically, the\n> approach to adding/dropping constraints associated with domains doesn't\n> work.\n> \n> 1. Insufficient locking, guise 1: there's no protection against someone\n> else dropping a column or whole table between the time you find a\n\nOk.. I obviously have to spend some time to figure out how locking works\nand exactly what it affects.\n\nI had incorrectly assumed that since dropping a column required removal\nof the pg_attribute entry, that holding a RowExclusive on it would\nprevent that.\n\n> 2. Insufficient locking, guise 2: there's no protection against someone\n> else adding a column or table while you're processing an ALTER DOMAIN,\n> either. This means that constraint checks will be missed. Example:\n\nLocking the entry in pg_type doesn't prevent that? Afterall, something\ndoes a test to see if the type exists prior to allowing the client to\nadd it.\n\n> 3. Too much locking, guise 1: the ALTER DOMAIN command will acquire\n> exclusive lock on every table that it scans, and will hold all these\n> locks until it commits. This can easily result in deadlocks --- against\n> other ALTER DOMAIN commands, or just against any random other\n> transaction that is unlucky enough to try to write any two tables\n> touched by the ALTER DOMAIN. AFAICS you don't need an exclusive lock,\n> you just want to prevent updates of the table until the domain changes\n> are committed, so ShareLock would be sufficient; that would reduce but\n> not eliminate the risk of deadlock.\n\nI noticed a completed TODO item that allows multiple locks to be\nobtained simultaneously, and had intended on using that for this -- but\nwas having a hard time tracking down an example.\n\n> 4. Too much locking, guise 2: the ExclusiveLock acquired on pg_class by\n> get_rels_with_domain has no useful effect, since it's released again\n> at the end of the scan; it does manage to shut down most sorts of schema\n> changes while get_rels_with_domain runs, however. This is bad enough,\n> but:\n\nYeah... Trying to transfer the lock to the attributes -- which as you've\nshown doesn't do what I thought.\n\n> 5. Performance sucks. In the regression database on my machine, \"alter\n> domain mydom set not null\" takes over six seconds --- that's for a\n> freshly created domain that's not used *anywhere*. This can be blamed\n> entirely on the inefficient implementation of get_rels_with_domain.\n\nYes, I need to (and intend to) redo this with dependencies, but hadn't\nfigured out how. I'm surprised it took 6 seconds though. I hardly\nnotice any delay on a database with ~30 tables in it.\n\n> 6. Permission bogosity: as per discussion yesterday, ownership of a\n> schema does not grant ownership rights on contained objects.\n\nPatch submitted yesterday to correct this.\n\n> 7. No mechanism for causing constraint changes to actually propagate\n> after they are made. This is more a fault of the design of the domain\n> constraint patch than it is of the alter patch, but nonetheless alter is\n> what exposes it. The problem is particularly acute because you chose to\n> insert a domain's constraint expressions into coercion operations at\n> expression parsing time, which is far too early. A stored rule that has\n> a coerce-to-domain operation in it will have a frozen idea of what\n> constraints it should be enforcing. Probably the expression tree should\n> just have a \"CoerceToDomain foo\" node in it, and at executor startup\n> this node would have to look to the pg_type entry for foo to see exactly\n> what it should be enforcing at the moment.\n\nThanks for the explanations. I'll see if I can 1) fix my poor knowledge\nof locking, 2) Add to my notes that I need to test stuff with Rules from\nnow on, and 3) correct the above items.\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "10 Dec 2002 17:23:49 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": false, "msg_subject": "Re: Problems with ALTER DOMAIN patch" }, { "msg_contents": "Rod Taylor <rbt@rbt.ca> writes:\n>> 2. Insufficient locking, guise 2: there's no protection against someone\n>> else adding a column or table while you're processing an ALTER DOMAIN,\n>> either. This means that constraint checks will be missed. Example:\n\n> Locking the entry in pg_type doesn't prevent that?\n\nIf there were such a thing as \"locking the entry in pg_type\", it might\nprevent that, but (a) there isn't, and (b) your code wouldn't invoke it \nif there were. Reading a row should surely not be tantamount to\ninvoking an exclusive lock on it.\n\nIn any case, other backends might have the pg_type entry in their\nsyscaches, in which case their references to the type would be quite\nfree of any actual read of the pg_type row that might fall foul of\nyour hypothetical lock.\n\nTo make this work in a reliable way, there needs to be some concept\nof acquiring a lock on the type as an entity, in the same way that\nLockRelation acquires a lock on a relation as an entity --- which has\nonly the loosest possible connection to the notion of a lock on the\nrelation's pg_class row. We have no such locks on types at present,\nbut I think it may be time to invent 'em.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Dec 2002 22:56:56 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": true, "msg_subject": "Re: Problems with ALTER DOMAIN patch " }, { "msg_contents": "On Tue, 2002-12-10 at 22:56, Tom Lane wrote:\n> Rod Taylor <rbt@rbt.ca> writes:\n> >> 2. Insufficient locking, guise 2: there's no protection against someone\n> >> else adding a column or table while you're processing an ALTER DOMAIN,\n> >> either. This means that constraint checks will be missed. Example:\n> \n> > Locking the entry in pg_type doesn't prevent that?\n> \n> If there were such a thing as \"locking the entry in pg_type\", it might\n> prevent that, but (a) there isn't, and (b) your code wouldn't invoke it \n> if there were. Reading a row should surely not be tantamount to\n> invoking an exclusive lock on it.\n\nHrm... Yes.. I came to that conclusion while walking home. My concepts\nof locking, and what actually happens in PostgreSQL are two completely\ndifferent things.\n\n> In any case, other backends might have the pg_type entry in their\n> syscaches, in which case their references to the type would be quite\n> free of any actual read of the pg_type row that might fall foul of\n> your hypothetical lock.\n\nSo... Basically I'm cooked.\n\n> relation's pg_class row. We have no such locks on types at present,\n> but I think it may be time to invent 'em.\n\nI'd be happy to use them once created.\n\nThanks again for the help.\n\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "10 Dec 2002 23:09:26 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": false, "msg_subject": "Re: Problems with ALTER DOMAIN patch" }, { "msg_contents": "Rod Taylor wrote:\n> > relation's pg_class row. We have no such locks on types at present,\n> > but I think it may be time to invent 'em.\n> \n> I'd be happy to use them once created.\n> \n> Thanks again for the help.\n\nWhere does that leave the patch _until_ they are created?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 10 Dec 2002 23:15:36 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Problems with ALTER DOMAIN patch" }, { "msg_contents": "Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> Where does that leave the patch _until_ they are created?\n\nI'd say \"it's under death sentence unless fixed before 7.4 release\".\nI don't want to back it out in toto right now, because that will\ninterfere with other edits I'm in process of making (and also Rod\nincluded some necessary fixes to the domain-constraint patch in the\nalter-domain patch; which wasn't too clean of him but it's done).\n\nFor now, please put \"fix or disable ALTER DOMAIN\" on the must-do-\nbefore-7.4 part of TODO.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Dec 2002 23:54:33 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": true, "msg_subject": "Re: Problems with ALTER DOMAIN patch " }, { "msg_contents": "Rod Taylor <rbt@rbt.ca> writes:\n> On Tue, 2002-12-10 at 22:56, Tom Lane wrote:\n>> relation's pg_class row. We have no such locks on types at present,\n>> but I think it may be time to invent 'em.\n\n> I'd be happy to use them once created.\n\nI think you misunderstood me ;=) ... that was a none-too-subtle\nsuggestion that *you* should go invent 'em, seeing as how you're the\none pushing the feature that makes 'em necessary.\n\nThe lock manager itself deals with lock tags that could be almost\nanything. We currently only use lock tags that represent relations or\nspecific pages in relations, but I see no reason that there couldn't\nalso be lock tags representing types --- or other basic catalog entries.\n(I am trying hard to repress the thought that we may already need\nlocking on other classes of entities as well.) What we need now is a\nlittle thought about exactly how to represent these different lock tags\n(should be easy), and about what semantics to assign to different lock\nmodes applied to pg_type entities (perhaps not so easy).\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 11 Dec 2002 00:05:55 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": true, "msg_subject": "Re: Problems with ALTER DOMAIN patch " }, { "msg_contents": "On Wed, 2002-12-11 at 00:05, Tom Lane wrote:\n> Rod Taylor <rbt@rbt.ca> writes:\n> > On Tue, 2002-12-10 at 22:56, Tom Lane wrote:\n> >> relation's pg_class row. We have no such locks on types at present,\n> >> but I think it may be time to invent 'em.\n> \n> > I'd be happy to use them once created.\n> \n> I think you misunderstood me ;=) ... that was a none-too-subtle\n\nNope... I understood. But it could take quite a while.\n\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "11 Dec 2002 09:48:40 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": false, "msg_subject": "Re: Problems with ALTER DOMAIN patch" }, { "msg_contents": "Rod Taylor wrote:\n-- Start of PGP signed section.\n> On Wed, 2002-12-11 at 00:05, Tom Lane wrote:\n> > Rod Taylor <rbt@rbt.ca> writes:\n> > > On Tue, 2002-12-10 at 22:56, Tom Lane wrote:\n> > >> relation's pg_class row. We have no such locks on types at present,\n> > >> but I think it may be time to invent 'em.\n> > \n> > > I'd be happy to use them once created.\n> > \n> > I think you misunderstood me ;=) ... that was a none-too-subtle\n> \n> Nope... I understood. But it could take quite a while.\n\nI have an idea. Rather than doing some complex locking for types, why\ndon't we just restrict ALTER DOMAIN to cases where we are the only one\nattached to the database, as seen in dropdb(). Seems like an easy way\nto get the feature in without adding a new locking method. Also, it\nwould allow the regression test to work too because no one is attached\nto 'regression' at the time of the test.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 11 Dec 2002 17:57:21 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Problems with ALTER DOMAIN patch" }, { "msg_contents": "Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> I have an idea. Rather than doing some complex locking for types, why\n> don't we just restrict ALTER DOMAIN to cases where we are the only one\n> attached to the database, as seen in dropdb().\n\nYech!\n\n> would allow the regression test to work too because no one is attached\n> to 'regression' at the time of the test.\n\nNo, we'd just get random reports of unreproducible failures.\n\nFWIW, I frequently attach to the regression database with a manual psql\nwhile the regression tests are running. It's a useful thing to do ---\nI've more than once found bugs that way.\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 11 Dec 2002 18:35:54 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": true, "msg_subject": "Re: Problems with ALTER DOMAIN patch " }, { "msg_contents": "\nIt is an idea if no better one can be found, unless we don't want ALTER\nDOMAIN at all, which doesn't seem good.\n\n---------------------------------------------------------------------------\n\nTom Lane wrote:\n> Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> > I have an idea. Rather than doing some complex locking for types, why\n> > don't we just restrict ALTER DOMAIN to cases where we are the only one\n> > attached to the database, as seen in dropdb().\n> \n> Yech!\n> \n> > would allow the regression test to work too because no one is attached\n> > to 'regression' at the time of the test.\n> \n> No, we'd just get random reports of unreproducible failures.\n> \n> FWIW, I frequently attach to the regression database with a manual psql\n> while the regression tests are running. It's a useful thing to do ---\n> I've more than once found bugs that way.\n> \n> \t\t\tregards, tom lane\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 11 Dec 2002 19:00:08 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Problems with ALTER DOMAIN patch" }, { "msg_contents": "On Wed, 2002-12-11 at 19:00, Bruce Momjian wrote:\n> It is an idea if no better one can be found, unless we don't want ALTER\n> DOMAIN at all, which doesn't seem good.\n\nI'll make a proposal for 'Object' locks as suggested, and we'll see\nwhere we go from there.\n\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "12 Dec 2002 09:45:20 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": false, "msg_subject": "Re: Problems with ALTER DOMAIN patch" } ]
[ { "msg_contents": "\nOK, I wonder if adding -ldl will help. You need to link to the library\ncontaining the ldopen function.\n\n---------------------------------------------------------------------------\n\nShibashish wrote:\n> Thanks for the help. I edited the src/makefiles/Makefile.sco and removed\n> the export. But the compile still hangs up after the following errors.\n> I tried the following combinations too ....\n> export_dynamic = -Wl,-Bexport\n> export_dynamic = -Wl\n> #export_dynamic = -Wl,-Bexport {stops at the following output}\n> \n> I will send the full output if u need.\n> \n> -------------------------------------\n> make[4]: Leaving directory\n> `/data/postgres/postgresql-7.3/src/backend/utils/mb'\n> /usr/ccs/bin/ld -r -o SUBSYS.o fmgrtab.o adt/SUBSYS.o cache/SUBSYS.o\n> error/SUBSYS.o fmgr/SUBSYS.o hash/SUBSYS.o init/SUBSYS.o misc/SUBSYS.o\n> mmgr/SUBSYS.o sort/SUBSYS.o time/SUBSYS.o mb/SUBSYS.o\n> make[3]: Leaving directory\n> `/data/postgres/postgresql-7.3/src/backend/utils'\n> gcc -O2 -Wall -Wmissing-prototypes -Wmissing-declarations -L../../src/port\n> access/SUBSYS.o bootstrap/SUBSYS.o catalog/SUBSYS.o parser/SUBSYS.o\n> commands/SUBSYS.o executor/SUBSYS.o lib/SUBSYS.o libpq/SUBSYS.o\n> main/SUBSYS.o nodes/SUBSYS.o optimizer/SUBSYS.o port/SUBSYS.o\n> postmaster/SUBSYS.o regex/SUBSYS.o rewrite/SUBSYS.o storage/SUBSYS.o\n> tcop/SUBSYS.o utils/SUBSYS.o -lPW -lgen -lld -lsocket -lnsl -lm -lpgport\n> -o postgres\n> undefined first referenced\n> symbol in file\n> _dlopen utils/SUBSYS.o\n> _dlerror utils/SUBSYS.o\n> _dlsym utils/SUBSYS.o\n> _dlclose utils/SUBSYS.o\n> i386ld fatal: Symbol referencing errors. No output written to postgres\n> make[2]: *** [postgres] Error 1\n> make[2]: Leaving directory `/data/postgres/postgresql-7.3/src/backend'\n> make[1]: *** [all] Error 2\n> make[1]: Leaving directory `/data/postgres/postgresql-7.3/src'\n> make: *** [all] Error 2\n> -------------------------------------\n> \n> from Shibashish\n> \n> \n> On Mon, 9 Dec 2002, Bruce Momjian wrote:\n> \n> >\n> > It should have worked, but edit Makefile.shlib and remove that offending\n> > export from the link line. That may fix it.\n> >\n> > ---------------------------------------------------------------------------\n> \n> \n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 10 Dec 2002 14:25:03 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: PostgreSQL 7.3 Installation on SCO" } ]
[ { "msg_contents": "With this pg_hba.conf (line numbers from vi, of course):\n\n 48 # TYPE DATABASE USER IP-ADDRESS IP-MASK METHOD 49 \n 50 local all all ident sameuser\n 51 host all 127.0.0.1 127.0.0.1 ident s ameuser\n 52 \n\nwe naturally get a parse error because of the missing user column entry\nin line 51. But in the log we see:\n\nDec 10 19:27:42 linda postgres[10944]: [8] LOG: parse_hba: invalid\nsyntax in pg_hba.conf file at line 95, token \"ident\"\n\nIn a more complicated file, a bogus line number is going to make\ndebugging very tricky. I tried following this in gdb, but haven't\nmanaged to track it through the fork of the new backend.\n\n-- \nOliver Elphick Oliver.Elphick@lfix.co.uk\nIsle of Wight, UK http://www.lfix.co.uk/oliver\nGPG: 1024D/3E1D0C1C: CA12 09E0 E8D5 8870 5839 932A 614D 4C34 3E1D 0C1C\n ========================================\n \"I beseech you therefore, brethren, by the mercies of \n God, that ye present your bodies a living sacrifice, \n holy, acceptable unto God, which is your reasonable \n service.\" Romans 12:1 \n\n", "msg_date": "10 Dec 2002 19:35:26 +0000", "msg_from": "Oliver Elphick <olly@lfix.co.uk>", "msg_from_op": true, "msg_subject": "pg_hba.conf parse error gives wrong line number" }, { "msg_contents": "\nI see the problem with the line number here. I will work on a fix now. \nThanks.\n\n\n---------------------------------------------------------------------------\n\nOliver Elphick wrote:\n> With this pg_hba.conf (line numbers from vi, of course):\n> \n> 48 # TYPE DATABASE USER IP-ADDRESS IP-MASK METHOD 49 \n> 50 local all all ident sameuser\n> 51 host all 127.0.0.1 127.0.0.1 ident s ameuser\n> 52 \n> \n> we naturally get a parse error because of the missing user column entry\n> in line 51. But in the log we see:\n> \n> Dec 10 19:27:42 linda postgres[10944]: [8] LOG: parse_hba: invalid\n> syntax in pg_hba.conf file at line 95, token \"ident\"\n> \n> In a more complicated file, a bogus line number is going to make\n> debugging very tricky. I tried following this in gdb, but haven't\n> managed to track it through the fork of the new backend.\n> \n> -- \n> Oliver Elphick Oliver.Elphick@lfix.co.uk\n> Isle of Wight, UK http://www.lfix.co.uk/oliver\n> GPG: 1024D/3E1D0C1C: CA12 09E0 E8D5 8870 5839 932A 614D 4C34 3E1D 0C1C\n> ========================================\n> \"I beseech you therefore, brethren, by the mercies of \n> God, that ye present your bodies a living sacrifice, \n> holy, acceptable unto God, which is your reasonable \n> service.\" Romans 12:1 \n> \n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 4: Don't 'kill -9' the postmaster\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 10 Dec 2002 15:00:52 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: pg_hba.conf parse error gives wrong line number" }, { "msg_contents": "OK, the following patch fixes the bug. The code wasn't handling\ncomments properly in dealing the the line count. I will backpatch this\ninto 7.3.\n\n---------------------------------------------------------------------------\n\nOliver Elphick wrote:\n> With this pg_hba.conf (line numbers from vi, of course):\n> \n> 48 # TYPE DATABASE USER IP-ADDRESS IP-MASK METHOD 49 \n> 50 local all all ident sameuser\n> 51 host all 127.0.0.1 127.0.0.1 ident s ameuser\n> 52 \n> \n> we naturally get a parse error because of the missing user column entry\n> in line 51. But in the log we see:\n> \n> Dec 10 19:27:42 linda postgres[10944]: [8] LOG: parse_hba: invalid\n> syntax in pg_hba.conf file at line 95, token \"ident\"\n> \n> In a more complicated file, a bogus line number is going to make\n> debugging very tricky. I tried following this in gdb, but haven't\n> managed to track it through the fork of the new backend.\n> \n> -- \n> Oliver Elphick Oliver.Elphick@lfix.co.uk\n> Isle of Wight, UK http://www.lfix.co.uk/oliver\n> GPG: 1024D/3E1D0C1C: CA12 09E0 E8D5 8870 5839 932A 614D 4C34 3E1D 0C1C\n> ========================================\n> \"I beseech you therefore, brethren, by the mercies of \n> God, that ye present your bodies a living sacrifice, \n> holy, acceptable unto God, which is your reasonable \n> service.\" Romans 12:1 \n> \n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 4: Don't 'kill -9' the postmaster\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n\nIndex: src/backend/libpq/hba.c\n===================================================================\nRCS file: /cvsroot/pgsql-server/src/backend/libpq/hba.c,v\nretrieving revision 1.90\ndiff -c -c -r1.90 hba.c\n*** src/backend/libpq/hba.c\t6 Dec 2002 04:37:02 -0000\t1.90\n--- src/backend/libpq/hba.c\t11 Dec 2002 22:13:10 -0000\n***************\n*** 93,98 ****\n--- 93,99 ----\n next_token(FILE *fp, char *buf, const int bufsz)\n {\n \tint\t\t\tc;\n+ \tchar\t *start_buf = buf;\n \tchar\t *end_buf = buf + (bufsz - 1);\n \tbool\t\tin_quote = false;\n \tbool\t\twas_quote = false;\n***************\n*** 115,121 ****\n \t\t\t{\n \t\t\t\twhile ((c = getc(fp)) != EOF && c != '\\n')\n \t\t\t\t\t;\n! \t\t\t\tcontinue;\n \t\t\t}\n \n \t\t\tif (buf >= end_buf)\n--- 116,125 ----\n \t\t\t{\n \t\t\t\twhile ((c = getc(fp)) != EOF && c != '\\n')\n \t\t\t\t\t;\n! \t\t\t\t/* If only comment, consume EOL too; return EOL */\n! \t\t\t\tif (c != EOF && buf == start_buf)\n! \t\t\t\t\tc = getc(fp);\n! \t\t\t\tbreak;\n \t\t\t}\n \n \t\t\tif (buf >= end_buf)", "msg_date": "Wed, 11 Dec 2002 17:16:05 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: pg_hba.conf parse error gives wrong line number" } ]
[ { "msg_contents": "Hi,\n I am doing some experiments on dead tuples, I am looking of reusing the\ndead tuples apace in a particular page during the \"Update\".This patch \nis meant for the tables\nwhich are heavily updated to avoid vacuum very frequently.By using it \nwill arrest the size of\ntable for heavily updated table. The algorithm works like this:\n 1) During the update it check for the dead tuples in the current \npage(page that contain\n the tuple that need to be updated). If it finds any dead tuples it uses \nthe dead tuple space\n by ovewriting on dead tuple. The checking of dead tuple is very similer \nto the task that of\nlazy vaccum.\n 2) If it cannot find any dead tuple it proceed as usual by inserting \nat the end of table .\n\n Performance Effect:\n 1) The CPU processing will be slighly more for the update, but io \nprocessing is\nexactly same\n 2) The size of table grows slower under heavy update , so vacuum is \nnot required very frequently.\n The total processing for update is more or less same even after \ndoing large number of updates without vacuum.\n\nDoes it breaks anythings by overwriting the dead tuples ?.\n\nComments?.\n\njana\n\n\n", "msg_date": "Wed, 11 Dec 2002 10:41:31 +0800", "msg_from": "Janardhan <jana-reddy@mediaring.com.sg>", "msg_from_op": true, "msg_subject": "Reusing Dead Tuples:" }, { "msg_contents": "Janardhan <jana-reddy@mediaring.com.sg> writes:\n> Does it breaks anythings by overwriting the dead tuples ?.\n\nYes. You cannot do that unless you've first removed index entries\npointing at the dead tuples --- and jumped through the same locking\nhoops that lazy vacuum does while removing index entries.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Dec 2002 23:35:35 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Reusing Dead Tuples: " }, { "msg_contents": "Tom Lane wrote:\n\n>Janardhan <jana-reddy@mediaring.com.sg> writes:\n> \n>\n>>Does it breaks anythings by overwriting the dead tuples ?.\n>> \n>>\n>\n>Yes. You cannot do that unless you've first removed index entries\n>pointing at the dead tuples --- and jumped through the same locking\n>hoops that lazy vacuum does while removing index entries.\n>\n>\t\t\tregards, tom lane\n>\n> \n>\nif i am not wrong while updating a tuple, we are also creating a new \nindex entry . so if the\ntuple is dead then the index entry pointing it also a dead index tuple. \nso even if dead index tuple is not\nremoved then also it should not break thing, since the dead index tuple \nwill not be used, am i correct?.\n\nwhat is reason why the dead heap tuples are maintained in a linked list \n?. since for every dead heap tuple there\nis a corresponding dead index tuple.\n\nRegards\njana\n\n\n\n\n\n\nTom Lane wrote:\n\nJanardhan <jana-reddy@mediaring.com.sg> writes:\n \n\nDoes it breaks anythings by overwriting the dead tuples ?.\n \n\n\nYes. You cannot do that unless you've first removed index entries\npointing at the dead tuples --- and jumped through the same locking\nhoops that lazy vacuum does while removing index entries.\n\n\t\t\tregards, tom lane\n\n \n\nif i am not wrong while  updating a tuple, we are also creating a  new\nindex entry . so  if the\ntuple is dead then the index entry pointing it also a dead index tuple.\nso even if dead index tuple is not \nremoved then also it should not break thing, since the dead index tuple\nwill not be used, am i correct?.\n\nwhat is reason why the dead heap tuples are maintained in a linked list\n?. since for every dead heap tuple there\nis a corresponding dead index tuple.\n\nRegards\njana", "msg_date": "Wed, 11 Dec 2002 13:47:14 +0800", "msg_from": "Janardhan <jana-reddy@mediaring.com.sg>", "msg_from_op": true, "msg_subject": "Re: Reusing Dead Tuples:" }, { "msg_contents": "Janardhan <jana-reddy@mediaring.com.sg> writes:\n> if i am not wrong while updating a tuple, we are also creating a new \n> index entry .\n\nYes.\n\n> so if the\n> tuple is dead then the index entry pointing it also a dead index tuple. \n\nYes.\n\n> so even if dead index tuple is not\n> removed then also it should not break thing, since the dead index tuple \n> will not be used, am i correct?.\n\nNo. A process running an indexscan will assume that the index tuple\naccurately describes the heap tuple it is pointing at. If the heap\ntuple is live then it will be returned as satisfying the indexscan.\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 11 Dec 2002 00:51:21 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Reusing Dead Tuples: " }, { "msg_contents": "Tom Lane wrote:\n\n>Janardhan <jana-reddy@mediaring.com.sg> writes:\n> \n>\n>>Does it breaks anythings by overwriting the dead tuples ?.\n>> \n>>\n>\n>Yes. You cannot do that unless you've first removed index entries\n>pointing at the dead tuples --- and jumped through the same locking\n>hoops that lazy vacuum does while removing index entries.\n>\n>\t\t\tregards, tom lane\n>\n> \n>\nDoes it breaks any other things if all the index entries pointing the \ndead tuple are removed before reusing the dead tuple?.\n\nRegards\njana\n\n\n\n\n\n\nTom Lane wrote:\n\nJanardhan <jana-reddy@mediaring.com.sg> writes:\n \n\nDoes it breaks anythings by overwriting the dead tuples ?.\n \n\n\nYes. You cannot do that unless you've first removed index entries\npointing at the dead tuples --- and jumped through the same locking\nhoops that lazy vacuum does while removing index entries.\n\n\t\t\tregards, tom lane\n\n \n\nDoes it breaks any other things if all the index entries pointing the\ndead tuple are  removed before reusing the dead tuple?.\n\nRegards\njana", "msg_date": "Fri, 13 Dec 2002 15:20:41 +0800", "msg_from": "Janardhan <jana-reddy@mediaring.com.sg>", "msg_from_op": true, "msg_subject": "Re: Reusing Dead Tuples:" }, { "msg_contents": "Janardhan <jana-reddy@mediaring.com.sg> writes:\n> Does it breaks any other things if all the index entries pointing the \n> dead tuple are removed before reusing the dead tuple?.\n\nPossibly you could make that work, but I think you'll find the\nefficiency advantage you were chasing to be totally gone. The locking\nscheme is heavily biased against you, and the index AMs don't offer an\nAPI designed for efficient retail index-tuple deletion.\n\nOf course that just says that you're swimming against the tide of\nprevious optimization efforts. But the thing you need to face up to\nis you are taking what had been background maintenance tasks (viz,\nVACUUM) and moving them into the foreground critical path. This *will*\nslow down your foreground applications.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 13 Dec 2002 02:37:21 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Reusing Dead Tuples: " }, { "msg_contents": "Tom Lane wrote:\n\n>Janardhan <jana-reddy@mediaring.com.sg> writes:\n> \n>\n>>Does it breaks any other things if all the index entries pointing the \n>>dead tuple are removed before reusing the dead tuple?.\n>> \n>>\n>\n>Possibly you could make that work, but I think you'll find the\n>efficiency advantage you were chasing to be totally gone. The locking\n>scheme is heavily biased against you, and the index AMs don't offer an\n>API designed for efficient retail index-tuple deletion.\n>\n>Of course that just says that you're swimming against the tide of\n>previous optimization efforts. But the thing you need to face up to\n>is you are taking what had been background maintenance tasks (viz,\n>VACUUM) and moving them into the foreground critical path. This *will*\n>slow down your foreground applications.\n>\n>\t\t\tregards, tom lane\n>\n> \n>\ntoday i could able to complete the patch and it is working only for \nb-tree. i have added a new method am_delete\nto the API and bt_delete to the B-tree index to delete a single entry. \n for the timebeing this works only with\nb-tree indexs.\n\nRegarding the complexity of deleting a tuple from b-tree , it is same \nor less then that of\ninserting a tuple into a B-tree( since delete does not require spliting \nthe page). The approach is slightly \ndifferent to that of lazy vacuum. lazy vacuum scan entire index table to \ndelete the dead entries.\nhere it search for the pariticilar entry similer to that of insert . \nhere locking may not have much impact. It locks only single buffer to \ndelete the index entry.\n\nRegarding the efficiency, if the entire Index table is in buffered then \nit does not require any\nadditional IO , only extra CPU is required to delete entries in index table.\nI am using postgres in a application where is there is heavy updates for \ngroup of tables(small size), before inserting\na single record in huge table. this entire thing constitue single \ntransaction. currently as time goes on the transaction\nprocessing speed decreases till the database is vacuumed.\n\nUsing this new patch i am hoping the trasaction processing time will \nremain constant irrespective of time. Only i need to do\nvaccum once i delete large number of entries from some of the tables.\n\nregards, jana\n\n\n\n\n\n\n\nTom Lane wrote:\n\nJanardhan <jana-reddy@mediaring.com.sg> writes:\n \n\nDoes it breaks any other things if all the index entries pointing the \ndead tuple are removed before reusing the dead tuple?.\n \n\n\nPossibly you could make that work, but I think you'll find the\nefficiency advantage you were chasing to be totally gone. The locking\nscheme is heavily biased against you, and the index AMs don't offer an\nAPI designed for efficient retail index-tuple deletion.\n\nOf course that just says that you're swimming against the tide of\nprevious optimization efforts. But the thing you need to face up to\nis you are taking what had been background maintenance tasks (viz,\nVACUUM) and moving them into the foreground critical path. This *will*\nslow down your foreground applications.\n\n\t\t\tregards, tom lane\n\n \n\ntoday i could able to complete the patch and  it is working only for\nb-tree.  i have added  a new  method am_delete \nto the  API and bt_delete to the B-tree index to delete a single entry.\n  for the timebeing this works only with\nb-tree indexs.\n\nRegarding the complexity of deleting a  tuple from b-tree , it is same\nor less then that of\ninserting a tuple into a B-tree( since delete does not require spliting\nthe page). The approach is  slightly  \ndifferent to that of lazy vacuum. lazy vacuum scan entire index table\nto delete the dead entries.\nhere it search for the pariticilar entry similer to that of insert .  \nhere locking may not have much impact. It locks only  single buffer to\ndelete  the index entry.\n\nRegarding the efficiency, if the entire Index table is in buffered then\nit does not require any \nadditional IO , only extra CPU is required to delete entries in index\ntable.\nI am using postgres in a application where is there is heavy updates\nfor group of tables(small size), before inserting\na single record in huge table. this entire thing constitue single\ntransaction. currently as time goes on the transaction \nprocessing speed decreases till the database is vacuumed. \n\nUsing this new patch i am hoping the trasaction processing time will\nremain constant irrespective of time. Only i need to do\nvaccum once i delete large number of entries from some of the tables.\n\nregards, jana", "msg_date": "Wed, 18 Dec 2002 16:23:15 +0800", "msg_from": "Janardhan <jana-reddy@mediaring.com.sg>", "msg_from_op": true, "msg_subject": "Re: Reusing Dead Tuples:" } ]
[ { "msg_contents": "\nJust wondering where I should put my modified tuning notes. I was planning \non making them section 3.7 in the Admin guide. Does that sound reasonable?\n\nThe current version can be seen at:\n\n http://www.rhyme.com.au/manuals/pgsql-7.3/postmaster-tuning-software.html\n\nI think it's important we get something on tuning into the manual - I'm not \nparticularly attached to where.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Thu, 12 Dec 2002 01:40:41 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "DB Tuning Notes - Where To?" }, { "msg_contents": "On Wed, 2002-12-11 at 09:40, Philip Warner wrote:\n> \n> Just wondering where I should put my modified tuning notes. I was planning \n> on making them section 3.7 in the Admin guide. Does that sound reasonable?\n> \n> The current version can be seen at:\n> \n> http://www.rhyme.com.au/manuals/pgsql-7.3/postmaster-tuning-software.html\n> \n> I think it's important we get something on tuning into the manual - I'm not \n> particularly attached to where.\n> \n\nI had thought the information would be tied to the relevant sections of\n3.4 Run-time Configuration. I'm not sure where the vacuum/analyze\ninformation would go in this scenario though, so a general software\ntuning section does seem appropriate. Do you see a 3.8 Tuning the Server\n(Hardware) section as well? \n\nRobert Treat\n\n\n", "msg_date": "11 Dec 2002 10:25:11 -0500", "msg_from": "Robert Treat <xzilla@users.sourceforge.net>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes - Where To?" }, { "msg_contents": "At 10:25 AM 11/12/2002 -0500, Robert Treat wrote:\n>Do you see a 3.8 Tuning the Server\n>(Hardware) section as well?\n\nHardware and/or OS. I think Bruce's tuning docs tend to address the \nhardware and environmental issues, so I was not planning to write anything \nmyself.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Thu, 12 Dec 2002 02:30:26 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes - Where To?" }, { "msg_contents": "Philip Warner wrote:\n> At 10:25 AM 11/12/2002 -0500, Robert Treat wrote:\n> >Do you see a 3.8 Tuning the Server\n> >(Hardware) section as well?\n> \n> Hardware and/or OS. I think Bruce's tuning docs tend to address the \n> hardware and environmental issues, so I was not planning to write anything \n> myself.\n\nI was unsure how _definiative_ the discussion was.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 11 Dec 2002 19:12:40 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes - Where To?" }, { "msg_contents": "At 07:12 PM 11/12/2002 -0500, Bruce Momjian wrote:\n>I was unsure how _definiative_ the discussion was.\n\nI am sure that the statements in the document are *not* totally accurate; \nbut my belief is that they are better than nothing and a good starting \npoint for tuning.\n\nFor example, most numbers are upper limits: the 'Changed' and 'Removed' \nfigures do not reflect actual free pages, they just reflect pages on which \none or more tuple was added or removed etc. The page may still be 90% full, \nand may never make it into the FSM depending on the setting of MAX_FSM_PAGES.\n\nThe choice of thresholds is arbitrary (as noted in the document), but also \nreflects a good starting point IMO.\n\nI think it is also important to put tuning notes in the main shipping \nmanuals; do people think it would be a good idea to add a disclaimer that:\n\n tuning is inherently database-specific, you may need to set\n thresholds much lower, or may find that these are too high\n - experiment with your database\n\n?\n\nIMO no tuning document can ever be definitive; it can only provide the user \nwith tools to understand the problem and manipulate the outcomes.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Thu, 12 Dec 2002 12:10:22 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes - Where To?" }, { "msg_contents": "At 12:10 PM 12/12/2002 +1100, Philip Warner wrote:\n>good starting point for tuning\n\nI think this probably sums it up.\n\nIMO it is grandiose to call it a tuning document; at best it is a \n'Misbehaviour Avoidance' document. We probably need something about the \nusual database-side tuning options: indexes, WAL, page sizes etc, and \nsomething else about environmental options (moving files, RAID etc).\n\nShould I change the section name to 'Routine Maintenance'?\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Thu, 12 Dec 2002 12:28:10 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes - Where To?" }, { "msg_contents": "Philip Warner wrote:\n> At 12:10 PM 12/12/2002 +1100, Philip Warner wrote:\n> >good starting point for tuning\n> \n> I think this probably sums it up.\n> \n> IMO it is grandiose to call it a tuning document; at best it is a \n> 'Misbehaviour Avoidance' document. We probably need something about the \n> usual database-side tuning options: indexes, WAL, page sizes etc, and \n> something else about environmental options (moving files, RAID etc).\n\nYep, that sounds like it. We should have that right in the docs next to\nthat tuning parameter, or somewhere in a separate section on freespace\nmap and point there. Also, this may improve over releases so we need to\ntrack the changes in the official release. If we can convey how the\nfree space map works, people will be able to understand how their\nworkload affects it.\n\n> Should I change the section name to 'Routine Maintenance'?\n\nWell, it isn't something you would play with regularly, like backups. \nIt is more like the disk space analysis section I added in 7.3.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 11 Dec 2002 20:43:22 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes - Where To?" }, { "msg_contents": "At 08:43 PM 11/12/2002 -0500, Bruce Momjian wrote:\n>Well, it isn't something you would play with regularly, like backups.\n\nHow about I call it 'Managing Server Resources' and put it between 'Runtime \nConfiguration' and 'Managing Kernel Resources'? ie. it becomes 3.5.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Thu, 12 Dec 2002 12:57:27 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes - Where To?" }, { "msg_contents": "Philip Warner <pjw@rhyme.com.au> writes:\n> Should I change the section name to 'Routine Maintenance'?\n\nNo, because we already have an admin-guide section titled more or less\nthat way. Also, \"tuning\" is not \"what you'd better do every week\";\nin my mind tuning activities will hold good till your database usage\nchanges.\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 12 Dec 2002 01:22:16 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes - Where To? " }, { "msg_contents": "At 01:22 AM 12/12/2002 -0500, Tom Lane wrote:\n>in my mind tuning activities will hold good till your database usage\n>changes.\n\nWhat about my later suggestion of 'Managing Server Resources', going before \n'Managing Kernel Resources'. Or perhaps, 'Tuning Server Resources'...\n\nThe document describes how to set the config items and vacuum/analyze \nfrequencies...so it should not be regularly performed.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Thu, 12 Dec 2002 17:33:45 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes - Where To? " }, { "msg_contents": "Philip Warner writes:\n\n> Just wondering where I should put my modified tuning notes. I was planning\n> on making them section 3.7 in the Admin guide. Does that sound reasonable?\n\nThe frequency of vacuum and analyze would seem to belong under Routine\nVacuuming in the Maintenance chapter. Setting max_fsm_relations belongs\nunder the entry in the big list of configuration parameters. Setting\nmax_fsm_pages also belongs in that list, but the parts that refer to the\nVACUUM output should be put near the Routine Vacuuming section. Create\nloads of clickable cross-references.\n\nBtw., please don't enshrine promises about future versions in the\ndocumentation. Either create a patch that makes the indicated changes\n(recommended, since the problem appears to be analyzed) or just document\nthe status quo.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Fri, 13 Dec 2002 20:55:33 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: DB Tuning Notes - Where To?" }, { "msg_contents": "At 08:55 PM 13/12/2002 +0100, Peter Eisentraut wrote:\n>The frequency of vacuum and analyze would seem to belong..\n>...max_fsm_relations belongs...\n>...Setting max_fsm_pages also belongs in that list...\n>...parts that refer to the VACUUM output should be put... near the Routine \n>Vacuuming\n\nNot sure I like this idea at all.\n\nThe stuff I have written is in a practical order and makes a complete whole \n- each section refers to information obtained in previous sections.\n\nI have no objection to putting links in the sections you listed, indicating \n'this is also discussed in XXX (Managing Server Resources)'...\n\n\n>----------------------------------------------------------------\n\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Sat, 14 Dec 2002 11:28:27 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: DB Tuning Notes - Where To?" } ]
[ { "msg_contents": "http://www.postgresql.org/idocs/ is down\n\nWarning: Unable to connect to PostgreSQL server: The Data Base System \nis shutting down in /usr/local/www/www/idocs/opendb.php on line 3\nUnable to access database\n-- \nDan Langille : http://www.langille.org/\n\n", "msg_date": "Wed, 11 Dec 2002 11:50:00 -0500", "msg_from": "\"Dan Langille\" <dan@langille.org>", "msg_from_op": true, "msg_subject": "http://www.postgresql.org/idocs/ is down" }, { "msg_contents": "Hi Dan,\n\nThanks for pointing this out.\n\nThe Admin guys are looking into it now. Hopefully it'll be fixed soon.\n\n:-/\n\nRegards and best wishes,\n\nJustin Clift\n\n\nDan Langille wrote:\n> http://www.postgresql.org/idocs/ is down\n> \n> Warning: Unable to connect to PostgreSQL server: The Data Base System \n> is shutting down in /usr/local/www/www/idocs/opendb.php on line 3\n> Unable to access database\n\n\n-- \n\"My grandfather once told me that there are two kinds of people: those\nwho work and those who take the credit. He told me to try to be in the\nfirst group; there was less competition there.\"\n- Indira Gandhi\n\n", "msg_date": "Thu, 12 Dec 2002 04:00:20 +1100", "msg_from": "Justin Clift <justin@postgresql.org>", "msg_from_op": false, "msg_subject": "Re: http://www.postgresql.org/idocs/ is down" }, { "msg_contents": "Hi Dan,\n\nThe database for the postgresql.org sites is back up again now.\n\nThanks for pointing it out.\n\n:-)\n\nRegards and best wishes,\n\nJustin Clift\n\n\nJustin Clift wrote:\n> Hi Dan,\n> \n> Thanks for pointing this out.\n> \n> The Admin guys are looking into it now. Hopefully it'll be fixed soon.\n> \n> :-/\n> \n> Regards and best wishes,\n> \n> Justin Clift\n> \n> \n> Dan Langille wrote:\n> \n>> http://www.postgresql.org/idocs/ is down\n>>\n>> Warning: Unable to connect to PostgreSQL server: The Data Base System \n>> is shutting down in /usr/local/www/www/idocs/opendb.php on line 3\n>> Unable to access database\n> \n> \n> \n\n\n-- \n\"My grandfather once told me that there are two kinds of people: those\nwho work and those who take the credit. He told me to try to be in the\nfirst group; there was less competition there.\"\n- Indira Gandhi\n\n", "msg_date": "Thu, 12 Dec 2002 04:50:08 +1100", "msg_from": "Justin Clift <justin@postgresql.org>", "msg_from_op": false, "msg_subject": "Re: http://www.postgresql.org/idocs/ is down" } ]
[ { "msg_contents": "On Tue, Dec 10, 2002 at 09:04:07PM +0100, Peter Eisentraut wrote:\n> Alvaro Herrera writes:\n> \n> > Would it work to make \\d tab-completable in a way that showed both the\n> > commands that are available and the objects they describe? e.g.\n> >\n> > \\d<tab> would show something like\n> > \\dt [tables]\t\\ds [sequences]\t\\dv [views] ...\n> \n> That won't work. The actual completion and the view of the alternatives\n> if the completion is ambiguous is driven by the same data.\n> \n\nSpeaking of psql and tab completion, I ran into a bogosity in ALTER\nyesterday, in ver 7.3. I was looking to change the ownership of a table,\nso I did:\n\nALTER TABLE foo <tab>\n ADD ALTER RENAME \n\nHmm, so maybe it's ALTER again? (I don't rember that syntax, but ...)\n\nALTER TABLE doo ALTER <tab>\n GROUP SCHEMA TABLE USER\n\nhuh? ALTER TABLE foo ALTER TABLE? \n\n<checks \\h ALTER and \\h ALTER TABLE>\n\nSeems ALTER TABLE foo <tab> should yield:\n\nADD DROP ALTER RENAME OWNER TO\n\nAnd ALTER <tab> (which currently gives GROUP SCHEMA TABLE USER)\nshould have:\n\nDATABASE GROUP TABLE TRIGGER USER\n\nAnd ALTER TABLE foo ALTER <tab>\nshould list column names of foo.\n\nDigging into the source, I see this case was there, but you couldn't\nreach it because of the match on ALTER.\n\nI've sent a patch to PATCHES to implement my proposed completions, with\nthe caveat that they're based on the \\h ALTER output, and manual testing,\nnot on reading the parser in detail. \n\nRoss\n\n", "msg_date": "Wed, 11 Dec 2002 11:59:40 -0600", "msg_from": "\"Ross J. Reedstrom\" <reedstrm@rice.edu>", "msg_from_op": true, "msg_subject": "psql's tab completions for ALTER command" }, { "msg_contents": "Table:\n\n Column | Type | Modifiers\n-------------------+-----------------------+---------------------------------------------------------------------\n imported_date | integer | not null default 0\n\n\nPG v7.2.1 (nice and clean):\n\nselect imported_date, \"timestamptz\"(imported_date) from server_accounts limit 1;\n imported_date | timestamptz\n---------------+------------------------\n 1037498593 | 2002-11-16 18:03:13-08\n(1 row)\n\n\nPG v7.3 (nasty and dirty):\n\nselect imported_date, \"timestamptz\"(imported_date) from server_accounts limit 1;\nERROR: Bad timestamp external representation '1027966107'\n\nselect imported_date, \"timestamptz\"( cast(imported_date as timestamp) ) from server_accounts limit 1;\nERROR: Cannot cast type integer to timestamp without time zone\n\nselect imported_date, \"timestamptz\"( cast(imported_date as timestamptz) ) from server_accounts limit 1;\nERROR: Cannot cast type integer to timestamp with time zone\n\nselect imported_date, ('Jan 1 1970'::DATE + (imported_date || ' seconds')::INTERVAL)::timestamp from server_accounts limit 1\n imported_date | timestamp\n---------------+---------------------\n 1027966107 | 2002-07-29 18:08:27\n\nThe last query works, but you must admit it is pretty nasty. Yes, I could create a from_epoch() function that takes care of this, but should this really be nessecary?\n\nIs there a better way to go about this in v7.3 without changing the column type?\n\n\n\n", "msg_date": "11 Dec 2002 10:33:44 -0800", "msg_from": "Mike Benoit <mikeb@netnation.com>", "msg_from_op": false, "msg_subject": "Frustration with date/times/epoch in v7.3." } ]
[ { "msg_contents": "I have worked with Shibashish Satpathy to add support for SCO Openserver\n5.0.4 using gcc in 7.3.1. The port was accomplished via a small change\nto template/sco. Seeing as it was an unsupported platform, this is a\nno-risk change, because now it _does_ work.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 11 Dec 2002 17:29:07 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "SCO Openserver supported in 7.3.1" }, { "msg_contents": "Bruce Momjian wrote:\n> I have worked with Shibashish Satpathy to add support for SCO Openserver\n> 5.0.4 using gcc in 7.3.1. The port was accomplished via a small change\n> to template/sco. Seeing as it was an unsupported platform, this is a\n> no-risk change, because now it _does_ work.\n\nLet me add something. I worked with Shibashish via Yahoo chat, and it\nwas a very efficient way to do it. I am not sure if we could have done\nit via email. Chat is good for most interative processes.\n\nI now have access to +10 regular hackers posters, and this makes things\neasier when we want to knock ideas around in a more interactive way.\n\nMy chat addresses are:\n\t\n\tAIM\tbmomjian\n\tICQ\t151255111\n\tYahoo\tbmomjian\n\tMSN\troot@candle.pha.pa.us\n\tIRC\tbmomjian at #postgresql via EFNet or OpenProjects\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 11 Dec 2002 17:40:51 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: SCO Openserver supported in 7.3.1" } ]
[ { "msg_contents": "This patch fixes minor bugs in dictionary generator in contrib/tsearch \n(contrib/tsearch/makedict/makedict.pl)\n\nThank you.\n\n-- \nTeodor Sigaev\nteodor@stack.net", "msg_date": "Thu, 12 Dec 2002 16:22:25 +0300", "msg_from": "Teodor Sigaev <teodor@stack.net>", "msg_from_op": true, "msg_subject": "Please, apply patch for 7.3.1 and current CVS " }, { "msg_contents": "\nYour patch has been added to the PostgreSQL unapplied patches list at:\n\n\thttp://momjian.postgresql.org/cgi-bin/pgpatches\n\nI will try to apply it within the next 48 hours.\n\n---------------------------------------------------------------------------\n\n\nTeodor Sigaev wrote:\n> This patch fixes minor bugs in dictionary generator in contrib/tsearch \n> (contrib/tsearch/makedict/makedict.pl)\n> \n> Thank you.\n> \n> -- \n> Teodor Sigaev\n> teodor@stack.net\n> \n\n[ application/gzip is not supported, skipping... ]\n\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 4: Don't 'kill -9' the postmaster\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Thu, 12 Dec 2002 10:45:56 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Please, apply patch for 7.3.1 and current CVS" }, { "msg_contents": "\nPatch applied. Thanks.\n\nBackpatched to 7.3.\n\n---------------------------------------------------------------------------\n\n\nTeodor Sigaev wrote:\n> This patch fixes minor bugs in dictionary generator in contrib/tsearch \n> (contrib/tsearch/makedict/makedict.pl)\n> \n> Thank you.\n> \n> -- \n> Teodor Sigaev\n> teodor@stack.net\n> \n\n[ application/gzip is not supported, skipping... ]\n\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 4: Don't 'kill -9' the postmaster\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 00:50:38 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Please, apply patch for 7.3.1 and current CVS" } ]
[ { "msg_contents": "Is there a way to get pg logging of plans to be produced in the terse \nformat like when a user types \"explain select * from foo where bar = x\"\n\nThe plan logging is very verbose. Having a lighter version of the \nlogging would be helpful in pinpointing troublesome queries without \nslogging through pages of explain info.\n\nAm I out to lunch? Is there a way to do this (I haven't found it yet, \nand I have tried several permutations of log parameters) If not, does \nanyone else see the value?\n\n\n\n", "msg_date": "Thu, 12 Dec 2002 10:20:25 -0500", "msg_from": "mlw <pgsql@mohawksoft.com>", "msg_from_op": true, "msg_subject": "Logging Feature " } ]
[ { "msg_contents": "\nI had the following behaviour reported by a pgAdmin user on PostgreSQL\n7.3 (reproduced here on 7.3rc1 as it's all I have right now):\n\ndave=# create function dave.testfunc() returns int4 as 'select 1 + 2'\nlanguage sql;\nCREATE FUNCTION\ndave=# select proacl from pg_proc where proname = 'testfunc';\n proacl\n--------\n\n(1 row)\n\ndave=# grant execute on function dave.testfunc() to webmstrz;\nGRANT\ndave=# select proacl from pg_proc where proname = 'testfunc';\n proacl\n----------------------------\n {=X,postgres=X,webmstrz=X}\n(1 row)\n\nI would not have expected public to now have execute permission. Any\nreason for this, or is it a bug?\n\nRegards, Dave.\n", "msg_date": "Thu, 12 Dec 2002 16:10:45 -0000", "msg_from": "\"Dave Page\" <dpage@vale-housing.co.uk>", "msg_from_op": true, "msg_subject": "Problem with function permissions" }, { "msg_contents": "\"Dave Page\" <dpage@vale-housing.co.uk> writes:\n> I would not have expected public to now have execute permission. Any\n> reason for this, or is it a bug?\n\nThe default permissions for functions grant execute to public; the\nsystem is just instantiating that default when you do an explicit grant.\n\nThe original implementation of function permissions didn't grant\nanything to public, but we concluded that that would break too many\napplications to be a reasonable default.\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 12 Dec 2002 11:48:24 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Problem with function permissions " } ]
[ { "msg_contents": "\n\n> -----Original Message-----\n> From: Tom Lane [mailto:tgl@sss.pgh.pa.us] \n> Sent: 12 December 2002 16:48\n> To: Dave Page\n> Cc: pgsql-hackers@postgresql.org\n> Subject: Re: [HACKERS] Problem with function permissions \n> \n> \n> \"Dave Page\" <dpage@vale-housing.co.uk> writes:\n> > I would not have expected public to now have execute \n> permission. Any \n> > reason for this, or is it a bug?\n> \n> The default permissions for functions grant execute to \n> public; the system is just instantiating that default when \n> you do an explicit grant.\n> \n> The original implementation of function permissions didn't \n> grant anything to public, but we concluded that that would \n> break too many applications to be a reasonable default.\n\nOK, thanks.\n\nRegards, Dave.\n", "msg_date": "Thu, 12 Dec 2002 16:52:24 -0000", "msg_from": "\"Dave Page\" <dpage@vale-housing.co.uk>", "msg_from_op": true, "msg_subject": "Re: Problem with function permissions " } ]
[ { "msg_contents": "I was bemused to notice that pg_dump is currently unable to dump the\nregression database. The domain regression test leaves an empty table\n(one with zero columns), which causes pg_dump to produce\n\n--\n-- TOC entry 172 (OID 675837)\n-- Name: domnotnull; Type: TABLE; Schema: public; Owner: postgres\n--\n\nCREATE TABLE domnotnull (\n);\n\n\nThis is rejected on restore:\n\nERROR: DefineRelation: please inherit from a relation or define an attribute\n\nI believe that the table would be correctly restored if we simply\nremoved that error check in DefineRelation. On the other hand, for\nordinary hand-entered CREATE TABLE commands it seems like a useful\nerror check.\n\nShould we remove this error check, thereby effectively making\nzero-column tables first-class citizens? If not, what shall we say to\npeople who complain that their dump is broken?\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 12 Dec 2002 16:08:06 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": true, "msg_subject": "Creating a zero-column table" }, { "msg_contents": "Tom Lane kirjutas R, 13.12.2002 kell 02:08:\n> I was bemused to notice that pg_dump is currently unable to dump the\n> regression database. The domain regression test leaves an empty table\n> (one with zero columns), which causes pg_dump to produce\n> \n> --\n> -- TOC entry 172 (OID 675837)\n> -- Name: domnotnull; Type: TABLE; Schema: public; Owner: postgres\n> --\n> \n> CREATE TABLE domnotnull (\n> );\n> \n> \n> This is rejected on restore:\n> \n> ERROR: DefineRelation: please inherit from a relation or define an attribute\n> \n> I believe that the table would be correctly restored if we simply\n> removed that error check in DefineRelation. On the other hand, for\n> ordinary hand-entered CREATE TABLE commands it seems like a useful\n> error check.\n> \n> Should we remove this error check, thereby effectively making\n> zero-column tables first-class citizens? \n\nI would vote for removing the check. I see no reason why one should not\nbe able to define a zero-column table. While we cant currently do\nanything useful (except select oid :) with it now, it does not mean that\nit would not serve as a valid base table for inheritance hierarchies in\nfuture. \n\nI'm probably going to propose an implicit zero-column base table for all\nuser defined tables (say any_table) so that one can get a list of all\ntuple ids in all tables by doing a simple \n\"select tableoid,oid from any_table\". This will of course not be very\nuseful for tables with no oids and where there is no index on oid.\n\n\n-- \nHannu Krosing <hannu@tm.ee>\n", "msg_date": "13 Dec 2002 02:21:00 +0500", "msg_from": "Hannu Krosing <hannu@tm.ee>", "msg_from_op": false, "msg_subject": "Re: Creating a zero-column table" }, { "msg_contents": "At 04:08 PM 12/12/2002 -0500, Tom Lane wrote:\n>Should we remove this error check, thereby effectively making\n>zero-column tables first-class citizens?\n\nIt's a bit daft, but I suspect it's the way to go. There has to be a \nnon-zero chance that a future version of pg_dump may want to add attributes \nto tables in a piecemeal fashion (cyclic FKs? Yuk).\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Fri, 13 Dec 2002 14:10:00 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": false, "msg_subject": "Re: Creating a zero-column table" }, { "msg_contents": "Tom Lane writes:\n\n> Should we remove this error check, thereby effectively making\n> zero-column tables first-class citizens?\n\nYes.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Sat, 14 Dec 2002 01:27:29 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: Creating a zero-column table" }, { "msg_contents": "On Fri, 2002-12-13 at 19:27, Peter Eisentraut wrote:\n> Tom Lane writes:\n> \n> > Should we remove this error check, thereby effectively making\n> > zero-column tables first-class citizens?\n> \n> Yes.\n> \n\nI've never been one for actively going against the sql standards (which\nthis case would seem to do) so if it is done lets make sure we add\ndocumentation to point out that we aren't compliant on this issue.\n\nRobert Treat\n\n\n", "msg_date": "16 Dec 2002 09:37:12 -0500", "msg_from": "Robert Treat <xzilla@users.sourceforge.net>", "msg_from_op": false, "msg_subject": "Re: Creating a zero-column table" }, { "msg_contents": "Robert Treat writes:\n\n> I've never been one for actively going against the sql standards (which\n> this case would seem to do) so if it is done lets make sure we add\n> documentation to point out that we aren't compliant on this issue.\n\nThis could be a reasonable place to hook in the \"SQL flagger\"\n(non-conforming syntax checker), which a conforming SQL implementation\nneeds to have anyhow.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Tue, 17 Dec 2002 00:45:30 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: Creating a zero-column table" } ]
[ { "msg_contents": "While playing around with a preliminary version of the information schema\nI get a failure in the type_sanity regression test here:\n\n-- Look for basic types that don't have an array type.\n-- NOTE: as of 7.3, this check finds SET, smgr, and unknown.\n\nSELECT p1.oid, p1.typname\nFROM pg_type as p1\nWHERE p1.typtype in ('b','d') AND p1.typname NOT LIKE '\\\\_%' AND NOT\nEXISTS\n (SELECT 1 FROM pg_type as p2\n WHERE p2.typname = ('_' || p1.typname)::name AND\n p2.typelem = p1.oid);\n\nThe information schema contains a couple of domains, but apparently no\narrays are generated for those. How is that supposed to be fixed?\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Fri, 13 Dec 2002 00:08:42 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": true, "msg_subject": "Arrays for domain types" }, { "msg_contents": "Peter Eisentraut <peter_e@gmx.net> writes:\n> While playing around with a preliminary version of the information schema\n> I get a failure in the type_sanity regression test here:\n\n> SELECT p1.oid, p1.typname\n> FROM pg_type as p1\n> WHERE p1.typtype in ('b','d') AND p1.typname NOT LIKE '\\\\_%' AND NOT\n> EXISTS\n> (SELECT 1 FROM pg_type as p2\n> WHERE p2.typname = ('_' || p1.typname)::name AND\n> p2.typelem = p1.oid);\n\nFor the moment I'd suggest backing off the regression test from\n\t p1.typtype in ('b','d')\nto\n\t p1.typtype in ('b')\nProbably eventually CREATE DOMAIN ought to create an associated array\ntype, but it's not really your problem if it doesn't ...\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 13 Dec 2002 00:56:38 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Arrays for domain types " } ]
[ { "msg_contents": "Hello I'm an Italian student of computer science at\nUniversity of Rome \"La Sapienza\". I've to analyze some\ndaemons which run under root privileges with a tool\ndeveloped by my departement. This tool intercepts\ncritical syscalls, like Execve, and blocks illegal\ninvocation of that primitives (E.g. Execve(\"/bin/sh\"))\nperformed by a daemon which runs under root\nprivileges. This approach blocks buffer overflow\nattacks before they can complete (or I hope so). Now,\nthe problem is that postgres doesn' t run under root\nprivileges and that the tool intercepts only the\nsyscalls invoked by a process with root privileges. Is\npossible to force postgres to run under root\nprivileges? How can be done? I know my request is\nanomalous but i've to do this for my laboratory\nproject course. I should be very grateful if you'll\nanswer as soon as possible.\nThank you.\n\nDiego Talucci\n\n______________________________________________________________________\nYahoo! Foto: salva, ritocca, condividi e ordina stampe professionali\nhttp://it.yahoo.com/mail_it/foot/?http://it.photos.yahoo.com\n", "msg_date": "Fri, 13 Dec 2002 00:57:02 +0100 (CET)", "msg_from": "\"=?iso-8859-1?q?Diego=20T.?=\" <sebulba76it@yahoo.it>", "msg_from_op": true, "msg_subject": "Postgres syscalls" }, { "msg_contents": "\nOn Fri, 13 Dec 2002, [iso-8859-1] Diego T. wrote:\n\n> Hello I'm an Italian student of computer science at\n> University of Rome \"La Sapienza\". I've to analyze some\n> daemons which run under root privileges with a tool\n> developed by my departement. This tool intercepts\n> critical syscalls, like Execve, and blocks illegal\n> invocation of that primitives (E.g. Execve(\"/bin/sh\"))\n> performed by a daemon which runs under root\n> privileges. This approach blocks buffer overflow\n> attacks before they can complete (or I hope so). Now,\n> the problem is that postgres doesn' t run under root\n> privileges and that the tool intercepts only the\n> syscalls invoked by a process with root privileges. Is\n> possible to force postgres to run under root\n> privileges? How can be done? I know my request is\n> anomalous but i've to do this for my laboratory\n> project course. I should be very grateful if you'll\n> answer as soon as possible.\n\nYou could probably just hack out the checks in main/main.c\nand recompile, but postgres does call system and such to do\nthings (like create databases) so I'm not sure it'd be terribly\nuseful for you.\n\n\n", "msg_date": "Thu, 12 Dec 2002 16:09:51 -0800 (PST)", "msg_from": "Stephan Szabo <sszabo@megazone23.bigpanda.com>", "msg_from_op": false, "msg_subject": "Re: Postgres syscalls" }, { "msg_contents": "On 12 Dec 2002 at 16:09, Stephan Szabo wrote:\n\n> \n> On Fri, 13 Dec 2002, [iso-8859-1] Diego T. wrote:\n> \n> > Hello I'm an Italian student of computer science at\n> > University of Rome \"La Sapienza\". I've to analyze some\n> > daemons which run under root privileges with a tool\n> > developed by my departement. This tool intercepts\n> > critical syscalls, like Execve, and blocks illegal\n> > invocation of that primitives (E.g. Execve(\"/bin/sh\"))\n> > performed by a daemon which runs under root\n> > privileges. This approach blocks buffer overflow\n> > attacks before they can complete (or I hope so). Now,\n\nOK..\n\n> > the problem is that postgres doesn' t run under root\n> > privileges and that the tool intercepts only the\n> > syscalls invoked by a process with root privileges. Is\n> > possible to force postgres to run under root\n\n> You could probably just hack out the checks in main/main.c\n> and recompile, but postgres does call system and such to do\n> things (like create databases) so I'm not sure it'd be terribly\n> useful for you.\n\nI agree. Not running root is a god idea from secutiry point of view. That way \nany buffer overflow attacks would be half dead as it is.\n\nSecondly In my understanding, buffer overflow attacks can be stopped very \neffectivelyif compiler has stack smashing patches. ( Or is it kernel as well?)\n\nAnd do look at strace. I feel you are shooting at same target..\n\nHTH\n\n\nBye\n Shridhar\n\n--\nGrinnell's Law of Labor Laxity:\tAt all times, for any task, you have not got \nenough done today.\n\n", "msg_date": "Fri, 13 Dec 2002 12:31:58 +0530", "msg_from": "\"Shridhar Daithankar\" <shridhar_daithankar@persistent.co.in>", "msg_from_op": false, "msg_subject": "Re: Postgres syscalls" } ]
[ { "msg_contents": "Hello,\n Using Postgresql 7.3 (CVS REL7_3_STABLE today), I received the following\nerror:\n\ndropsites=> delete from cart_stores;\nERROR: heap_update: (am)invalid tid\n\nThis came from a database that was dumped from 7.2.1 using 7.2.1's pg_dump\nand imported into 7.3. I was able to delete the rows individually with a\nwhere clause ( WHERE id = ... ). The problem attempting to delete all rows\npersisted even when there was only 1 row left, even though this last row\ndeleted fine when I specified it explicitly with a where clause. I was\nable to delete the contents of other tables in the db without incident. I\nwas also able to reproduce this on a second box, also CVS REL7_3_STABLE\nfrom about a week ago. This problem is not exhibited under 7.2.1.\n\nSo, the question is, how do I go about tracking this down? Has anyone seen\nthis? It was suggested by one person that memory may be the issue, but the\n fact that it is exhibited on two boxen makes that seem unlikely.\n\nFYI:\n Machine 1:\nPIII 600 w/ 256MB\nFreeBSD 4.6 STABLE\n Machine 2:\nDual PIII 600 w/ 512MB\nFreeBSD 4.7 STABLE\n\nThanx in advance.\n -Wade Klaver\n", "msg_date": "Thu, 12 Dec 2002 16:16:55 -0800", "msg_from": "wade <wade@wavefire.com>", "msg_from_op": true, "msg_subject": "invalid tid errors." } ]
[ { "msg_contents": "At 04:08 PM 12/12/2002 -0500, Tom Lane wrote:\n>Should we remove this error check, thereby effectively making\n>zero-column tables first-class citizens?\n\nI should wait 2 minutes before hitting 'send'.\n\nThe other option is to disallow the steps that resulted in the zero-column \ntable in the first place. This is probably more rational.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Fri, 13 Dec 2002 14:11:05 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: Creating a zero-column table" }, { "msg_contents": "Philip Warner <pjw@rhyme.com.au> writes:\n> At 04:08 PM 12/12/2002 -0500, Tom Lane wrote:\n>> Should we remove this error check, thereby effectively making\n>> zero-column tables first-class citizens?\n\n> The other option is to disallow the steps that resulted in the zero-column \n> table in the first place. This is probably more rational.\n\nThat was where we were a couple months ago: DROP COLUMN would refuse to\nlet you drop the last column of a table. But that has much worse\nproblems than whether we are able to deliver a \"helpful\" error message.\nFor example:\n\tAmy does CREATE TABLE foo(f1 beths_type);\n\tBeth now cannot drop her type beths_type.\nIn most circles this would be called a denial of service.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 13 Dec 2002 00:31:07 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Creating a zero-column table " }, { "msg_contents": "At 12:31 AM 13/12/2002 -0500, Tom Lane wrote:\n> Amy does CREATE TABLE foo(f1 beths_type);\n> Beth now cannot drop her type beths_type.\n>In most circles this would be called a denial of service.\n\nSeems like a feature - if beth made the type public, she has to deal with \nfame. I don't see it as different from preventing a dropdb when it is being \naccessed, or unmounting a disk while it is being accessed. But I resume \nthis argument has already been had...\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Fri, 13 Dec 2002 18:31:38 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: Creating a zero-column table " }, { "msg_contents": "Philip Warner <pjw@rhyme.com.au> writes:\n> At 12:31 AM 13/12/2002 -0500, Tom Lane wrote:\n>> Amy does CREATE TABLE foo(f1 beths_type);\n>> Beth now cannot drop her type beths_type.\n>> In most circles this would be called a denial of service.\n\n> Seems like a feature - if beth made the type public, she has to deal with \n> fame.\n\nBut in every other context, Beth has the unconditional right to drop her\ntype: if it's not the only column in Amy's table, Beth can drop her type\nand Amy's column along with it.\n\nBasically, the no-zero-column-tables restriction for deletion was\nremoved because it creates more weird corner cases than it prevents.\nI still agree with that decision. What we're seeing here is that the\ncorresponding restriction during table creation also creates weird\ncorner cases.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 13 Dec 2002 02:56:38 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Creating a zero-column table " }, { "msg_contents": "At 02:56 AM 13/12/2002 -0500, Tom Lane wrote:\n>if it's not the only column in Amy's table, Beth can drop her type\n>and Amy's column along with it.\n\nI keep forgetting PG's inheritance features. In a non-inheritance system, I \nwould vote for forcing a one column table to be dropped. For PG, I think \nyou are right.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Fri, 13 Dec 2002 19:43:37 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": true, "msg_subject": "Re: Creating a zero-column table " }, { "msg_contents": "On Fri, 2002-12-13 at 03:43, Philip Warner wrote:\n> At 02:56 AM 13/12/2002 -0500, Tom Lane wrote:\n> >if it's not the only column in Amy's table, Beth can drop her type\n> >and Amy's column along with it.\n> \n> I keep forgetting PG's inheritance features. In a non-inheritance system, I \n> would vote for forcing a one column table to be dropped. For PG, I think \n> you are right.\n> \n\nJust out of curiosity, do any of the SQL specs deal with 0 column\ntables? I can't recall any dbms supporting a create table command that\ndidn't require at least 1 column.\n\nRobert Treat\n\n\n", "msg_date": "13 Dec 2002 09:44:26 -0500", "msg_from": "Robert Treat <xzilla@users.sourceforge.net>", "msg_from_op": false, "msg_subject": "Re: Creating a zero-column table" }, { "msg_contents": "Robert Treat <xzilla@users.sourceforge.net> writes:\n> Just out of curiosity, do any of the SQL specs deal with 0 column\n> tables? I can't recall any dbms supporting a create table command that\n> didn't require at least 1 column.\n\nActually, in SQL92 11.17 <drop column definition> I find\n\n 3) C shall be a column of T and C shall not be the only column of\n T.\n\nSQL99 has the same rule. So they indeed prefer the no-drop wart to the\ntable-with-zero-columns wart. I have to disagree with that conclusion\nthough ...\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 13 Dec 2002 09:59:39 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Creating a zero-column table " } ]
[ { "msg_contents": "\nOK, so what do we do with 7.3.1. Increment major or minor?\n\n---------------------------------------------------------------------------\n\npgman wrote:\n> Tom Lane wrote:\n> > Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> > > Tom Lane wrote:\n> > >> Wrong --- if you need a recompile then it's not binary-compatible, so\n> > >> it should be a major version bump.\n> > \n> > > But the previous poster said only API changes were reasons to bump the\n> > > major, right?\n> > \n> > Yes. He meant *binary* API changes, though, ie, anything that would\n> > break extant executables originally linked to the prior version of the\n> > shared library.\n> \n> Just for clarification --- don't most/all our releases make a binary\n> change that needs are compile? Actually, you are saying a recompile of\n> the client, right? Yes, we do those rarely, and in 7.3, for the NOTIFY\n> structure change. What we do often is want old binaries to use our new\n> libraries. That isn't a major bump requirement, right?\n> \n> -- \n> Bruce Momjian | http://candle.pha.pa.us\n> pgman@candle.pha.pa.us | (610) 359-1001\n> + If your life is a hard drive, | 13 Roberts Road\n> + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 00:07:07 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> OK, so what do we do with 7.3.1. Increment major or minor?\n\nMajor. I thought you did it already?\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 13 Dec 2002 00:31:34 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken? " }, { "msg_contents": "Tom Lane wrote:\n> Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> > OK, so what do we do with 7.3.1. Increment major or minor?\n> \n> Major. I thought you did it already?\n\nI did only minor, which I knew was safe. Do folks realize this will\nrequire recompile of applications by 7.3 users moving to 7.3.1? That\nseems very drastic, and there have been very few problem reports about\nthe NOTIFY change.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 00:34:58 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "On Fri, Dec 13, 2002 at 12:34:58AM -0500, Bruce Momjian wrote:\n> \n> I did only minor, which I knew was safe. Do folks realize this will\n> require recompile of applications by 7.3 users moving to 7.3.1? That\n> seems very drastic, and there have been very few problem reports about\n> the NOTIFY change.\n\nIn my case, I only know it broke some test cases. But still, isn't it\nbetter to err on the safe side and at least give the user a clue that\nthe upgrade isn't necessarily supposed to work with his existing code?\n\n\nJeroen\n\n", "msg_date": "Fri, 13 Dec 2002 12:37:53 +0100", "msg_from": "\"Jeroen T. Vermeulen\" <jtv@xs4all.nl>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "On Fri, 2002-12-13 at 05:34, Bruce Momjian wrote:\n> Tom Lane wrote:\n> > Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> > > OK, so what do we do with 7.3.1. Increment major or minor?\n> > \n> > Major. I thought you did it already?\n> \n> I did only minor, which I knew was safe. Do folks realize this will\n> require recompile of applications by 7.3 users moving to 7.3.1? That\n> seems very drastic, and there have been very few problem reports about\n> the NOTIFY change.\n\nIf the ABI is different, they need to recompile but don't have any\nindication of it. This is bad.\n\nIf the major number changes, they can keep the old library around for\nthe benefits of applications that have not yet been recompiled, while\nnewly compiled applications can use the new library\n\nSo please change it.\n\n-- \nOliver Elphick Oliver.Elphick@lfix.co.uk\nIsle of Wight, UK http://www.lfix.co.uk/oliver\nGPG: 1024D/3E1D0C1C: CA12 09E0 E8D5 8870 5839 932A 614D 4C34 3E1D 0C1C\n ========================================\n \"For thou art my hope, O Lord GOD; thou art my trust \n from my youth.\" Psalms 71:5 \n\n", "msg_date": "13 Dec 2002 11:54:28 +0000", "msg_from": "Oliver Elphick <olly@lfix.co.uk>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "Oliver Elphick wrote:\n> On Fri, 2002-12-13 at 05:34, Bruce Momjian wrote:\n> > Tom Lane wrote:\n> > > Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> > > > OK, so what do we do with 7.3.1. Increment major or minor?\n> > > \n> > > Major. I thought you did it already?\n> > \n> > I did only minor, which I knew was safe. Do folks realize this will\n> > require recompile of applications by 7.3 users moving to 7.3.1? That\n> > seems very drastic, and there have been very few problem reports about\n> > the NOTIFY change.\n> \n> If the ABI is different, they need to recompile but don't have any\n> indication of it. This is bad.\n> \n> If the major number changes, they can keep the old library around for\n> the benefits of applications that have not yet been recompiled, while\n> newly compiled applications can use the new library\n> \n> So please change it.\n\nThe point is that we are changing it for 7.3.1, so though 7.3 libpq is\nalmost identical to 7.3.1 libpq, we are going to bump the major and\nforce recompile. The binary API change was from 7.2 to 7.3, not 7.3 to\n7.3.1. Do people still want a major bump in 7.3.1?\n\nClearly we should have bumped the major in 7.3, but we didn't.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 13:29:53 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> The point is that we are changing it for 7.3.1, so though 7.3 libpq is\n> almost identical to 7.3.1 libpq, we are going to bump the major and\n> force recompile. The binary API change was from 7.2 to 7.3, not 7.3 to\n> 7.3.1. Do people still want a major bump in 7.3.1?\n\nYes. We already discussed this: the consensus was that it's better to\ninconvenience the early adopters of 7.3 than leave a time bomb sitting\nthere throughout the 7.3.* cycle.\n\nBTW, if we do make a major bump, that would suggest we ought to get\n7.3.1 out as soon as possible, so as to minimize the uptake of 7.3.\nI have been thinking that a mid-Dec release of 7.3.1 would be a good\nplan anyway, since we already know of several moderately serious bugs\nin 7.3.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 13 Dec 2002 13:34:55 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken? " }, { "msg_contents": "Tom Lane wrote:\n> Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> > The point is that we are changing it for 7.3.1, so though 7.3 libpq is\n> > almost identical to 7.3.1 libpq, we are going to bump the major and\n> > force recompile. The binary API change was from 7.2 to 7.3, not 7.3 to\n> > 7.3.1. Do people still want a major bump in 7.3.1?\n> \n> Yes. We already discussed this: the consensus was that it's better to\n> inconvenience the early adopters of 7.3 than leave a time bomb sitting\n> there throughout the 7.3.* cycle.\n> \n> BTW, if we do make a major bump, that would suggest we ought to get\n> 7.3.1 out as soon as possible, so as to minimize the uptake of 7.3.\n> I have been thinking that a mid-Dec release of 7.3.1 would be a good\n> plan anyway, since we already know of several moderately serious bugs\n> in 7.3.\n\nOK, let me see if I understand the ramifications.\n\nIf you install 7.3.1 _on_top_of 7.3, both major versions will exist, and\nyou your old binaries will continue to work. However, if you delete the\nold libraries, then install, anything compiled against 7.3 will not work\nuntil it is recompiled.\n\nAlso, any new linking against a 7.3.1 that has both major version\nnumbers will use the newer major? Is that right?\n\n(I honestly never understood all this before.)\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 14:13:21 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "On Fri, 2002-12-13 at 19:13, Bruce Momjian wrote:\n> OK, let me see if I understand the ramifications.\n> \n> If you install 7.3.1 _on_top_of 7.3, both major versions will exist, and\n> you your old binaries will continue to work. However, if you delete the\n> old libraries, then install, anything compiled against 7.3 will not work\n> until it is recompiled.\n\nYes. You will have libpq.so.3.0 in 7.3.1; and you have libpq.so.2.2\nfrom 7.3 (and also from 7.2.x, though in fact they are different). If\nyou have installed 7.3.1 on top of 7.3, you will have libpq.so.3\n(symlinked to libpq.so.3.0) from 7.3.1, and libpq.so.2 (symlinked to\nlibpq.so.2.2) from an earlier release.\n\n> \n> Also, any new linking against a 7.3.1 that has both major version\n> numbers will use the newer major? Is that right?\n\n7.3.1 will only have the new major version number; the old one will have\ncome from 7.3 or earlier. The library chosen by the linker is the one\nlinked to libpq.so.\n\n\n-- \nOliver Elphick <olly@lfix.co.uk>\nLFIX Limited\n\n", "msg_date": "13 Dec 2002 21:06:47 +0000", "msg_from": "Oliver Elphick <olly@lfix.co.uk>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "On Fri, Dec 13, 2002 at 09:06:47PM +0000, Oliver Elphick wrote:\n> \n> Yes. You will have libpq.so.3.0 in 7.3.1; and you have libpq.so.2.2\n> from 7.3 (and also from 7.2.x, though in fact they are different). If\n> you have installed 7.3.1 on top of 7.3, you will have libpq.so.3\n> (symlinked to libpq.so.3.0) from 7.3.1, and libpq.so.2 (symlinked to\n> libpq.so.2.2) from an earlier release.\n\nDoesn't that mean that a user upgrading from 7.3 to 7.3.1 will have two\nlibraries, libpq.so.2.2 and libpq.so.3, that both adhere to the 7.3 ABI?\n\nPerhaps 7.3.1 could include a true, 7.2-style libpq.so.2.2 to overwrite\nany 7.3-style version that the original 7.3 may have installed under that\nname?\n\n\nJeroen\n\n", "msg_date": "Sat, 14 Dec 2002 02:33:06 +0100", "msg_from": "\"Jeroen T. Vermeulen\" <jtv@xs4all.nl>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "Jeroen T. Vermeulen wrote:\n> On Fri, Dec 13, 2002 at 09:06:47PM +0000, Oliver Elphick wrote:\n> > \n> > Yes. You will have libpq.so.3.0 in 7.3.1; and you have libpq.so.2.2\n> > from 7.3 (and also from 7.2.x, though in fact they are different). If\n> > you have installed 7.3.1 on top of 7.3, you will have libpq.so.3\n> > (symlinked to libpq.so.3.0) from 7.3.1, and libpq.so.2 (symlinked to\n> > libpq.so.2.2) from an earlier release.\n> \n> Doesn't that mean that a user upgrading from 7.3 to 7.3.1 will have two\n> libraries, libpq.so.2.2 and libpq.so.3, that both adhere to the 7.3 ABI?\n> \n> Perhaps 7.3.1 could include a true, 7.2-style libpq.so.2.2 to overwrite\n> any 7.3-style version that the original 7.3 may have installed under that\n> name?\n\nThat's an interesting idea, but then 7.3 binaries would link to\nlibpq.so.2.2, which is bad.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 21:22:50 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> Jeroen T. Vermeulen wrote:\n>> Perhaps 7.3.1 could include a true, 7.2-style libpq.so.2.2 to overwrite\n>> any 7.3-style version that the original 7.3 may have installed under that\n>> name?\n\n> That's an interesting idea, but then 7.3 binaries would link to\n> libpq.so.2.2, which is bad.\n\nThere are a number of problems with that, but the biggie is that the old\nbinary API is inherently dependent on NAMEDATALEN. We can *not* ship a\nlibrary that is both 100% backwards compatible and 100% compatible to\na standard-issue 7.3 server.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 13 Dec 2002 23:50:52 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken? " }, { "msg_contents": "Bruce Momjian wrote:\n> Tom Lane wrote:\n> > Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> > > OK, so what do we do with 7.3.1. Increment major or minor?\n> > \n> > Major. I thought you did it already?\n> \n> I did only minor, which I knew was safe. Do folks realize this will\n> require recompile of applications by 7.3 users moving to 7.3.1? That\n> seems very drastic, and there have been very few problem reports about\n> the NOTIFY change.\n\nKeep in mind that people who are running applications compiled against\nthe older (7.2) library have the option of keeping the older library\naround. They can recompile their applications to link against the\nnewer library at their leisure.\n\nSomeone who has recompiled their applications against the 7.3\nlibraries is already safe -- they can install 7.3.1 and keep the old\n7.3 libraries around at the same time if they don't want to recompile\ntheir applications again. They just have to be careful to not use any\nprecompiled applications that were linked against the 7.2 series.\n\nThe people who are running 7.2 have the option of keeping their 7.2\nlibraries around while upgrading to 7.3.1. Any applications that are\ncompiled against the 7.2 libraries will still work.\n\nThe only people who are in trouble are the people who upgraded from\n7.2 to 7.3 but did not recompile some of their applications. They\nhave three options: install 7.3.1 and recompile their applications,\ninstall 7.3.1 and reinstall the 7.2 shared libraries, or recompile\ntheir applications against 7.3 (and then upgrade to 7.3.1 at their\nleisure). Obviously the last option is the worst of the lot: if\nthey're going to recompile their applications, they may as well do so\nagainst the newer (7.3.1) libraries with a higher major number so\nthere is no confusion about which libraries are really compatible with\nwhich.\n\n\n-- \nKevin Brown\t\t\t\t\t kevin@sysexperts.com\n", "msg_date": "Sat, 14 Dec 2002 05:13:26 -0800", "msg_from": "Kevin Brown <kevin@sysexperts.com>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "\nOK, I have updated the libpq major number in 7.3.X, and updated major\nand minor in HEAD. Do I need to increment the other interfaces that\n_use_ libpq, like ecpg? I think so.\n\n---------------------------------------------------------------------------\n\nOliver Elphick wrote:\n> On Fri, 2002-12-13 at 19:13, Bruce Momjian wrote:\n> > OK, let me see if I understand the ramifications.\n> > \n> > If you install 7.3.1 _on_top_of 7.3, both major versions will exist, and\n> > you your old binaries will continue to work. However, if you delete the\n> > old libraries, then install, anything compiled against 7.3 will not work\n> > until it is recompiled.\n> \n> Yes. You will have libpq.so.3.0 in 7.3.1; and you have libpq.so.2.2\n> from 7.3 (and also from 7.2.x, though in fact they are different). If\n> you have installed 7.3.1 on top of 7.3, you will have libpq.so.3\n> (symlinked to libpq.so.3.0) from 7.3.1, and libpq.so.2 (symlinked to\n> libpq.so.2.2) from an earlier release.\n> \n> > \n> > Also, any new linking against a 7.3.1 that has both major version\n> > numbers will use the newer major? Is that right?\n> \n> 7.3.1 will only have the new major version number; the old one will have\n> come from 7.3 or earlier. The library chosen by the linker is the one\n> linked to libpq.so.\n> \n> \n> -- \n> Oliver Elphick <olly@lfix.co.uk>\n> LFIX Limited\n> \n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Sat, 14 Dec 2002 13:59:33 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "\nI figured out why I forgot to update the minor number for 7.3. The old\nRELEASE_CHANGES file had:\n\n bump interface version numbers\n o src/interfaces/libpq/libpq.rc\n o src/include/pg_config.h.win32\n\nI had forgotten to explicitly list the Makefile changes.\n\nThe new list is:\n\n bump interface version numbers\n o src/interfaces/*/Makefile\n o src/interfaces/libpq/libpq.rc\n o src/include/pg_config.h.win32\n\nOf course, incrementing the minor number may not have even helped us\nbecause we needed a major increase, which I didn't understand at the\ntime.\n\n---------------------------------------------------------------------------\n\nOliver Elphick wrote:\n> On Fri, 2002-12-13 at 19:13, Bruce Momjian wrote:\n> > OK, let me see if I understand the ramifications.\n> > \n> > If you install 7.3.1 _on_top_of 7.3, both major versions will exist, and\n> > you your old binaries will continue to work. However, if you delete the\n> > old libraries, then install, anything compiled against 7.3 will not work\n> > until it is recompiled.\n> \n> Yes. You will have libpq.so.3.0 in 7.3.1; and you have libpq.so.2.2\n> from 7.3 (and also from 7.2.x, though in fact they are different). If\n> you have installed 7.3.1 on top of 7.3, you will have libpq.so.3\n> (symlinked to libpq.so.3.0) from 7.3.1, and libpq.so.2 (symlinked to\n> libpq.so.2.2) from an earlier release.\n> \n> > \n> > Also, any new linking against a 7.3.1 that has both major version\n> > numbers will use the newer major? Is that right?\n> \n> 7.3.1 will only have the new major version number; the old one will have\n> come from 7.3 or earlier. The library chosen by the linker is the one\n> linked to libpq.so.\n> \n> \n> -- \n> Oliver Elphick <olly@lfix.co.uk>\n> LFIX Limited\n> \n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Sat, 14 Dec 2002 14:47:55 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "On Sat, 2002-12-14 at 18:59, Bruce Momjian wrote:\n> OK, I have updated the libpq major number in 7.3.X, and updated major\n> and minor in HEAD. Do I need to increment the other interfaces that\n> _use_ libpq, like ecpg? I think so.\n\nI don't think so.\n\n$ ldd /usr/lib/postgresql/lib/libecpg.so\n\tlibpq.so.2 => /usr/lib/libpq.so.2 (0x40019000)\n\tlibc.so.6 => /lib/libc.so.6 (0x4002e000)\n\tlibssl.so.0.9.6 => /usr/lib/i686/libssl.so.0.9.6 (0x40141000)\n\tlibcrypto.so.0.9.6 => /usr/lib/i686/libcrypto.so.0.9.6 (0x4016e000)\n\tlibkrb5.so.17 => /usr/lib/libkrb5.so.17 (0x40226000)\n\tlibcrypt.so.1 => /lib/libcrypt.so.1 (0x4025c000)\n\tlibresolv.so.2 => /lib/libresolv.so.2 (0x40289000)\n\tlibnsl.so.1 => /lib/libnsl.so.1 (0x4029a000)\n\t/lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x80000000)\n\tlibdl.so.2 => /lib/libdl.so.2 (0x402ad000)\n\tlibcom_err.so.1 => /usr/lib/libcom_err.so.1 (0x402b0000)\n\tlibasn1.so.5 => /usr/lib/libasn1.so.5 (0x402b2000)\n\tlibroken.so.9 => /usr/lib/libroken.so.9 (0x402d2000)\n\tlibdb3.so.3 => /usr/lib/libdb3.so.3 (0x402e3000)\n\n\nHere libecpg will look for libpq.so.2. When 7.3.1 is released, this\nlibecpg will be replaced by one that looks for libpq.so.3. But I think\nthat, unless the API of libecpg changes, its version should stay the\nsame.\n\nIf you change it with libpq, you must also change it with all the other\nlibraries it links in, like libkrb5 and libdb3. That is clearly\nimpracticable.\n\n-- \nOliver Elphick Oliver.Elphick@lfix.co.uk\nIsle of Wight, UK http://www.lfix.co.uk/oliver\nGPG: 1024D/3E1D0C1C: CA12 09E0 E8D5 8870 5839 932A 614D 4C34 3E1D 0C1C\n ========================================\n \"But I will hope continually, and will yet praise thee \n more and more.\" Psalms 71:14 \n\n", "msg_date": "14 Dec 2002 20:20:22 +0000", "msg_from": "Oliver Elphick <olly@lfix.co.uk>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "Bruce Momjian wrote:\n> \n> OK, I have updated the libpq major number in 7.3.X, and updated major\n> and minor in HEAD. Do I need to increment the other interfaces that\n> _use_ libpq, like ecpg? I think so.\n\nYou'll only need to increment the major number of those other\ninterfaces that are no longer binary-compatible with their immediate\npredecessors.\n\nThat is, if the version of libecpg that ships with 7.3.1 is\nincompatible with the version of libecpg that ships with 7.3, then you\nshould bump its major number (the one that shipped with 7.3 should\nhave had its major number bumped if it was incompatible with the one\nthat shipped with 7.2.3). But its binary interface should be\nindependent of the binary interface of libpg.\n\nWhen you build a shared library, you link it against any shared\nlibraries it depends on. The resulting dependency list is contained\nwithin the shared library the same way it's contained in an\napplication. You can run ldd on shared libraries just as you can on\napplications.\n\n\nIt might be helpful to think of the major number as being the same as\na protocol description number. You change the protocol description\nnumber when you make changes to the protocol that would make a server\nusing the new version of the protocol incompatible with a client using\nthe old version of the protocol.\n\nIn fact, there's really very little conceptual difference between a\nprotocol and an API: both are a means for two entities to communicate\nwith one another. Change the API and you have to do something to make\nit clear that the API has changed. Same deal with a protocol.\n\n\nI don't know if what I'm saying here makes much sense to you, but I\nhope it helps...\n\n\n\n-- \nKevin Brown\t\t\t\t\t kevin@sysexperts.com\n", "msg_date": "Sun, 15 Dec 2002 00:44:40 -0800", "msg_from": "Kevin Brown <kevin@sysexperts.com>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "Peter Eisentraut <peter_e@gmx.net> writes:\n> Bruce Momjian writes:\n>> Do I need to increment the other interfaces that\n>> _use_ libpq, like ecpg?\n\n> If and only if the libpq API is part of their documented API. For ecpg I\n> think this is not the case, but for libpq++ it would seem to be the case.\n\nHowever, an app linked against libpq++ would also be linked against\nlibpq, and so the incompatibility will be flagged by the linker anyway.\nI can see no need to bump libpq++'s own number.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 15 Dec 2002 11:21:07 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken? " }, { "msg_contents": "Bruce Momjian writes:\n\n> Do I need to increment the other interfaces that\n> _use_ libpq, like ecpg?\n\nIf and only if the libpq API is part of their documented API. For ecpg I\nthink this is not the case, but for libpq++ it would seem to be the case.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Sun, 15 Dec 2002 17:21:48 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "On Sun, Dec 15, 2002 at 11:21:07AM -0500, Tom Lane wrote:\n> \n> However, an app linked against libpq++ would also be linked against\n> libpq, and so the incompatibility will be flagged by the linker anyway.\n> I can see no need to bump libpq++'s own number.\n\nPlus, of course, libpq++ being a C++ library will break compatibility at\nsome compiler upgrades--which isn't even necessarily visible to anyone\ncompiling a postgres release.\n\n\nJeroen\n\n", "msg_date": "Sun, 15 Dec 2002 23:08:49 +0100", "msg_from": "\"Jeroen T. Vermeulen\" <jtv@xs4all.nl>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "But it's something they should of already had to do. We're just paying\nlate for old sins. ;) \n\nGreg\n\n\nOn Thu, 2002-12-12 at 23:34, Bruce Momjian wrote:\n> Tom Lane wrote:\n> > Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> > > OK, so what do we do with 7.3.1. Increment major or minor?\n> > \n> > Major. I thought you did it already?\n> \n> I did only minor, which I knew was safe. Do folks realize this will\n> require recompile of applications by 7.3 users moving to 7.3.1? That\n> seems very drastic, and there have been very few problem reports about\n> the NOTIFY change.\n-- \nGreg Copeland <greg@copelandconsulting.net>\nCopeland Computer Consulting\n\n", "msg_date": "15 Dec 2002 21:13:29 -0600", "msg_from": "Greg Copeland <greg@CopelandConsulting.Net>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "Tom Lane wrote:\n> Peter Eisentraut <peter_e@gmx.net> writes:\n> > Bruce Momjian writes:\n> >> Do I need to increment the other interfaces that\n> >> _use_ libpq, like ecpg?\n> \n> > If and only if the libpq API is part of their documented API. For ecpg I\n> > think this is not the case, but for libpq++ it would seem to be the case.\n> \n> However, an app linked against libpq++ would also be linked against\n> libpq, and so the incompatibility will be flagged by the linker anyway.\n> I can see no need to bump libpq++'s own number.\n\nNew question --- didn't we change the externally visible PGNotify\nstructure in libpq-fe.h in 7.3, and as returned by PQnotifies:\n\n\tPGnotify *\n\tPQnotifies(PGconn *conn)\n\nmeaning if ecpg references PGnotify, should it have a new major number\ntoo, so actually, we did change the API in 7.3 and not just the binary\ncompatibility.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Mon, 16 Dec 2002 11:01:00 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> meaning if ecpg references PGnotify, should it have a new major number\n> too,\n\nNo, because PGnotify is *not part of ecpg's API*.\n\nThe ecpg library, when compiled, will show a dependency on the new major\nnumber for libpq. That does not mean that applications linked to ecpg\nneed to be recompiled. They don't (unless they depend directly on libpq\ntoo).\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 16 Dec 2002 11:12:31 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken? " }, { "msg_contents": "Tom Lane wrote:\n> Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> > meaning if ecpg references PGnotify, should it have a new major number\n> > too,\n> \n> No, because PGnotify is *not part of ecpg's API*.\n> \n> The ecpg library, when compiled, will show a dependency on the new major\n> number for libpq. That does not mean that applications linked to ecpg\n> need to be recompiled. They don't (unless they depend directly on libpq\n> too).\n\nOK, got it. That's the think I needed to understand.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Mon, 16 Dec 2002 11:15:27 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "On Mon, Dec 16, 2002 at 11:01:00AM -0500, Bruce Momjian wrote:\n> \n> New question --- didn't we change the externally visible PGNotify\n> structure in libpq-fe.h in 7.3, and as returned by PQnotifies:\n\nSpeaking of which, what if user relies on sizeof(PGnotify::relname)?\nThat code will recompile without any problems, but it won't actually\nwork as expected. So in a way, the change may require more than a\nsimple recompile.\n\n\nJeroen\n\n", "msg_date": "Mon, 16 Dec 2002 17:41:06 +0100", "msg_from": "\"Jeroen T. Vermeulen\" <jtv@xs4all.nl>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "On Mon, Dec 16, 2002 at 05:41:06PM +0100, Jeroen T. Vermeulen wrote:\n> \n> Speaking of which, what if user relies on sizeof(PGnotify::relname)?\n ^\n code\n\n\nJeroen\n\n", "msg_date": "Mon, 16 Dec 2002 22:09:24 +0100", "msg_from": "\"Jeroen T. Vermeulen\" <jtv@xs4all.nl>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken?" }, { "msg_contents": "Jeroen T. Vermeulen writes:\n > On Mon, Dec 16, 2002 at 05:41:06PM +0100, Jeroen T. Vermeulen wrote:\n > > Speaking of which, what if user relies on sizeof(PGnotify::relname)?\n > ^\n > code\n\nYes, a change in the size of relname makes this binary incompatible\nand the user code changes may not be just a simple recompile - It all\ndepends on what it being used for!\n\nLee.\n", "msg_date": "Tue, 17 Dec 2002 12:23:20 +0000", "msg_from": "Lee Kindness <lkindness@csl.co.uk>", "msg_from_op": false, "msg_subject": "Re: PQnotifies() in 7.3 broken?" } ]
[ { "msg_contents": "I wanted to outline some of the big items we are looking at for 7.4:\n\nWin32 Port:\n\n\tKatie Ward and Jan are working on contributing their Win32\n\tport for 7.4. They plan to have a patch available by the end of\n\tDecember.\n\nPoint-In-Time Recovery (PITR)\n\n\tJ. R. Nield did a PITR patch late in 7.3 development, and Patrick\n\tMacDonald from Red Hat is working on merging it into CVS and\n\tadding any missing pieces. Patrick, do you have an ETA on that?\n\nReplication\n\n\tI have talked to Darren Johnson and I believe 7.4 is the time to\n\tmerge the Postgres-R source tree into our main CVS. Most of the\n\treplication code will be in its own directory, with only minor\n\tchanges to our existing tree. They have single-master\n\treplication working now, so we may have that feature in some\n\tcapacity for 7.4. I know others are working on replication\n\tsolutions. This is probably the time to decide for certain if\n\tthis is the direction we want to go for replication. Most who\n\thave have studied Postgres-R feel it is the most promising\n\tmulti-master replication solution for reliably networked hosts.\n\nComments?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 01:22:27 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Big 7.4 items" }, { "msg_contents": "On 13 Dec 2002 at 1:22, Bruce Momjian wrote:\n> Replication\n> \n> \tI have talked to Darren Johnson and I believe 7.4 is the time to\n> \tmerge the Postgres-R source tree into our main CVS. Most of the\n> \treplication code will be in its own directory, with only minor\n> \tchanges to our existing tree. They have single-master\n> \treplication working now, so we may have that feature in some\n> \tcapacity for 7.4. I know others are working on replication\n> \tsolutions. This is probably the time to decide for certain if\n> \tthis is the direction we want to go for replication. Most who\n> \thave have studied Postgres-R feel it is the most promising\n> \tmulti-master replication solution for reliably networked hosts.\n> \n> Comments?\n\nSome.\n\n1) What kind of replication are we looking at? log file replay/syncnronous etc. \nIf it is real time, like usogres( I hope I am in line with things here), that \nwould be real good .Choice is always good..\n\n2 If we are going to have replication, can we have built in load balancing? Is \nit a good idea to have it in postgresql or a separate application would be way \nto go?\n\nAnd where are nested transactions?\n\n\n\nBye\n Shridhar\n\n--\nBooker's Law:\tAn ounce of application is worth a ton of abstraction.\n\n", "msg_date": "Fri, 13 Dec 2002 12:35:05 +0530", "msg_from": "\"Shridhar Daithankar\" <shridhar_daithankar@persistent.co.in>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "On Fri, 2002-12-13 at 06:22, Bruce Momjian wrote:\n> I wanted to outline some of the big items we are looking at for 7.4:\n> Point-In-Time Recovery (PITR)\n> \n> \tJ. R. Nield did a PITR patch late in 7.3 development, and Patrick\n> \tMacDonald from Red Hat is working on merging it into CVS and\n> \tadding any missing pieces. Patrick, do you have an ETA on that?\n\nHow hard would it be to extend PITR for master-slave (hot backup)\nrepliaction, which should then amount to continuously shipping logs to\nslave and doing nonstop PITR there :)\n\nIt will never be usable for multi-master replication, but somehow it\nfeels that for master-slave replication simple log replay would be most\nsimple and robust solution.\n\n-- \nHannu Krosing <hannu@tm.ee>\n", "msg_date": "13 Dec 2002 10:53:14 +0000", "msg_from": "Hannu Krosing <hannu@tm.ee>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Bruce Momjian wrote:\n> I wanted to outline some of the big items we are looking at for 7.4:\n> \n> Win32 Port:\n> \n> \tKatie Ward and Jan are working on contributing their Win32\n> \tport for 7.4. They plan to have a patch available by the end of\n> \tDecember.\n> \n> Point-In-Time Recovery (PITR)\n> \n> \tJ. R. Nield did a PITR patch late in 7.3 development, and Patrick\n> \tMacDonald from Red Hat is working on merging it into CVS and\n> \tadding any missing pieces. Patrick, do you have an ETA on that?\n> \n> Replication\n> \n> \tI have talked to Darren Johnson and I believe 7.4 is the time to\n> \tmerge the Postgres-R source tree into our main CVS. Most of the\n> \treplication code will be in its own directory, with only minor\n> \tchanges to our existing tree. They have single-master\n> \treplication working now, so we may have that feature in some\n> \tcapacity for 7.4. I know others are working on replication\n> \tsolutions. This is probably the time to decide for certain if\n> \tthis is the direction we want to go for replication. Most who\n> \thave have studied Postgres-R feel it is the most promising\n> \tmulti-master replication solution for reliably networked hosts.\n> \n> Comments?\n\nWhat about distributed TX support:\n\nhttp://groups.google.com/groups?hl=en&lr=&ie=UTF-8&oe=UTF-8&threadm=20021106111554.69ae1dcd.pgsql%40snaga.org&rnum=2&prev=/groups%3Fq%3DNAGAYASU%2BSatoshi%26ie%3DUTF-8%26oe%3DUTF-8%26hl%3Den\n\nMike Mascari\nmascarm@mascari.com\n\n", "msg_date": "Fri, 13 Dec 2002 07:55:25 -0500", "msg_from": "Mike Mascari <mascarm@mascari.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Bruce Momjian wrote:\n> Win32 Port:\n> \n> \tKatie Ward and Jan are working on contributing their Win32\n> \tport for 7.4. They plan to have a patch available by the end of\n> \tDecember.\n\nI have .Net Studio available to me, so if you need help in merging or testing \nor whatever, let me know.\n\n> Point-In-Time Recovery (PITR)\n> \n> \tJ. R. Nield did a PITR patch late in 7.3 development, and Patrick\n> \tMacDonald from Red Hat is working on merging it into CVS and\n> \tadding any missing pieces. Patrick, do you have an ETA on that?\n\nAs Hannu asked (and related to your question below), is there any thought of \nextending this to allow simple log based replication? In many important \nscenarios that would be more than adequate, and simpler to set up.\n\n> Replication\n> \n> \tI have talked to Darren Johnson and I believe 7.4 is the time to\n> \tmerge the Postgres-R source tree into our main CVS. Most of the\n> \treplication code will be in its own directory, with only minor\n> \tchanges to our existing tree. They have single-master\n> \treplication working now, so we may have that feature in some\n> \tcapacity for 7.4. I know others are working on replication\n> \tsolutions. This is probably the time to decide for certain if\n> \tthis is the direction we want to go for replication. Most who\n> \thave have studied Postgres-R feel it is the most promising\n> \tmulti-master replication solution for reliably networked hosts.\n\nI'd question if we would want the one-and-only builtin replication method to \nbe dependent on an external communication library (Spread). I would like to \nsee Postgres-R merged, but I'd also like to see a simple log-based option.\n\n> Comments?\n> \n\nI'd also second Mike Mascari's question -- whatever happened to the person \nworking on two-phase commit? Is that likely to be done for 7.4? Did he ever \nsend in a patch?\n\nJoe\n\n", "msg_date": "Fri, 13 Dec 2002 07:08:54 -0800", "msg_from": "Joe Conway <mail@joeconway.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Shridhar Daithankar wrote:\n> And where are nested transactions?\n\nI didn't mention nested transactions because it didn't seem to be a\n_big_ item like the others.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 11:51:12 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "On Friday 13 December 2002 17:51, Bruce Momjian wrote:\n> Shridhar Daithankar wrote:\n> > And where are nested transactions?\n>\n> I didn't mention nested transactions because it didn't seem to be a\n> _big_ item like the others.\n\nThis is big item\n\nregards\nHaris Peco\n", "msg_date": "Fri, 13 Dec 2002 18:05:25 +0100", "msg_from": "snpe <snpe@snpe.co.yu>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Shridhar Daithankar wrote:\n> 1) What kind of replication are we looking at? log file\n> replay/synchronous etc. If it is real time, like usogres( I\n> hope I am in line with things here), that would be real good.\n> Choice is always good.\n\nGood. This is the discussion we need. Let me quote the TODO list\nreplication section first:\n\n * Add replication of distributed databases [replication]\n o automatic failover\n o load balancing\n o master/slave replication\n o multi-master replication\n o partition data across servers\n o sample implementation in contrib/rserv\n o queries across databases or servers (two-phase commit)\n o allow replication over unreliable or non-persistent links\n o http://gborg.postgresql.org/project/pgreplication/projdisplay.php\n\nOK, the first thing is that there isn't any one replication solution\nthat will behave optimally in all situations. \n\nNow, let me describe Postgres-R and then the other replication\nsolutions. Postgres-R is multi-master, meaning you can send SELECT and\nUPDATE/DELETE queries to any of the servers in the cluster, and get the\nsame result. It is also synchronous, meaning it doesn't update the\nlocal copy until it is sure the other nodes agree to the change. It\nallows failover, because if one node goes down, the others keep going.\n\nNow, let me contrast:\n\nrserv and dbmirror do master/slave. There is no mechanism to allow you\nto do updates on the slave, and have them propagate to the master. You\ncan, however, send SELECT queries to the slave, and in fact that's how\nusogres does load balancing.\n\nTwo-phase commit is probably the most popular commercial replication\nsolution. While it works for multi-master, it suffers from poor\nperformance and doesn't handle cases where one node disappears very\nwell.\n\nAnother replication need is for asynchronous replication, most\ntraditionally for traveling salesmen who need to update their databases\nperiodically. The only solution I know for that is PeerDirect's\nPostgreSQL commercial offering at http://www.peerdirect.com. It is\npossible PITR may help with this, but we need to handle propagating\nchanges made by the salesmen back up into the server, and to do that, we\nwill need a mechanism to handle conflicts that occur when two people\nupdate the same records. This is always a problem for asynchronous\nreplication.\n\n> 2 If we are going to have replication, can we have built in load\n> balancing? Is it a good idea to have it in postgresql or a\n> separate application would be way to go?\n\nWell, because Postgres-R is multi-master, it has automatic load\nbalancing. You can have your users point to whatever node you want. \nYou can implement this \"pointing\" by using dns IP address cycling, or\nhave a router that auto-load balances, though you would need to keep a\ndb session on the same node, of course.\n\nSo, in summary, I think we will eventually have two directions for\nreplication. One is Postgres-R for multi-master, synchronous\nreplication, and PITR, for asynchronous replication. I don't think\nthere is any value to use PITR for synchronous replication because by\ndefinition, you don't _store_ the changes for later use because it is\nsynchronous. In synchronous, you communicate your changes to all the\nnodes involved, then commit them.\n\nI will describe the use of 'spread' and the Postgres-R internal issues\nin my next email.\n\n--\n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 12:31:40 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Hannu Krosing wrote:\n> On Fri, 2002-12-13 at 06:22, Bruce Momjian wrote:\n> > I wanted to outline some of the big items we are looking at for 7.4:\n> > Point-In-Time Recovery (PITR)\n> > \n> > \tJ. R. Nield did a PITR patch late in 7.3 development, and Patrick\n> > \tMacDonald from Red Hat is working on merging it into CVS and\n> > \tadding any missing pieces. Patrick, do you have an ETA on that?\n> \n> How hard would it be to extend PITR for master-slave (hot backup)\n> repliaction, which should then amount to continuously shipping logs to\n> slave and doing nonstop PITR there :)\n> \n> It will never be usable for multi-master replication, but somehow it\n> feels that for master-slave replication simple log replay would be most\n> simple and robust solution.\n\nExactly. See my previous email. We eventually have two replication\nsolutions: one, Postgres-R for multi-master, and PITR used for for\nasynchonous master/slave.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 12:32:52 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Mike Mascari wrote:\n> What about distributed TX support:\n> \n> http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&oe=UTF-8&threadm=20021106111554.69ae1dcd.pgsql%40snaga.org&rnum=2&prev=/groups%3Fq%3DNAGAYASU%2BSatoshi%26ie%3DUTF-8%26oe%3DUTF-8%26hl%3Den\n\nOK, yes, that is Satoshi's 2-phase commit implementation. I will\naddress 2-phase commit vs Postgres-R in my next email about spread.\n\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 12:34:37 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Okay. But please keep in mind that a 2-phase commit implementation is used for more than just replication. Any distributed TX will require a 2PC protocol. As an example, for the DBLINK implementation to ultimately be transaction safe (at least amongst multiple PostgreSQL installations), the players in the distributed transaction must all be participants in a 2PC exchange. And a participant whose communications link is dropped needs to be able to recover by asking the coordinator whether or not to complete or abort the distributed TX. I am 100% ignorant of the distributed TX standard Tom referenced earlier, but I'd guess there might be an assumption of 2PC support in the implementation. In other words, I think we still need 2PC, regardless of the method of replication. And if Satoshi Nagayasu has an implementation ready, why not investigate its possibilities?\n\nMike Mascari\nmascarm@mascari.com\n\n----- Original Message ----- \nFrom: \"Bruce Momjian\" <pgman@candle.pha.pa.us>\n\n\n> Mike Mascari wrote:\n> > What about distributed TX support:\n\n> OK, yes, that is Satoshi's 2-phase commit implementation. I will\n> address 2-phase commit vs Postgres-R in my next email about spread.\n\n\n", "msg_date": "Fri, 13 Dec 2002 12:57:25 -0500", "msg_from": "\"Mike Mascari\" <mascarm@mascari.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Joe Conway wrote:\n> Bruce Momjian wrote:\n> > Win32 Port:\n> > \n> > \tKatie Ward and Jan are working on contributing their Win32\n> > \tport for 7.4. They plan to have a patch available by the end of\n> > \tDecember.\n> \n> I have .Net Studio available to me, so if you need help in merging or testing \n> or whatever, let me know.\n\nOK, Jan, let him know how he can help.\n\n> > Point-In-Time Recovery (PITR)\n> > \n> > \tJ. R. Nield did a PITR patch late in 7.3 development, and Patrick\n> > \tMacDonald from Red Hat is working on merging it into CVS and\n> > \tadding any missing pieces. Patrick, do you have an ETA on that?\n> \n> As Hannu asked (and related to your question below), is there any thought of \n> extending this to allow simple log based replication? In many important \n> scenarios that would be more than adequate, and simpler to set up.\n\nYes, see previous email.\n\n> I'd question if we would want the one-and-only builtin replication method to \n> be dependent on an external communication library (Spread). I would like to \n> see Postgres-R merged, but I'd also like to see a simple log-based option.\n\nOK, let me reiterate I think we will have two replication solutions in\nthe end --- on Postgres-R for multi-master/synchronous, and PITR for\nmaster/slave asynchronous replication.\n\nLet me address the Spread issue and two-phase commit. (Spread is an\nopen source piece of software used by Postgres-R.)\n\nIn two-phase commit, when one node is about to commit, it gets a lock\nfrom all the other nodes, does its commit, then releases the lock. (Of\ncourse, this is simplified.) It is called two-phase because it says to\nall the other nodes \"I am about to do something, is that OK?\", then when\ngets all OK's, it does the commit and says \"I did the commit\".\n\nPostgres-R uses a different mechanism. This method is shown on page 22\nand 24 and following of:\n\n\tftp://gborg.postgresql.org/pub/pgreplication/stable/PostgreSQLReplication.pdf.gz\n\nThe basic difference is that Spread groups all the write sets into a\nqueue who's ordering is the same on all the nodes. Instead of asking\nfor approval for a commit, a node puts its commit in the Spread queue,\nand then waits for its own commit to come back in the queue, meaning all\nthe other nodes saw its commit too.\n\nThe only tricky part is that while reading the other node's write sets\nbefore its own arrives, it has to check to see if any of these conflict\nwith its own write set. If it conflicts, it has to assume the earlier\nwrite set succeeded and its own failed. It also has to check the write\nset stream and apply only those changes that don't conflict.\n\nAs stated before in Postgres-R discussion, this mechanism hinges on\nbeing able to determine which write sets conflict because there is no\nexplicit \"I aborted\", only a stream of write sets, and each node has to\naccept the non-conflicting ones and reject the conflicting ones.\n\n> I'd also second Mike Mascari's question -- whatever happened to the person \n> working on two-phase commit? Is that likely to be done for 7.4? Did he ever \n> send in a patch?\n\nI have not seen a patch from him, but it is very likely he could have\none for 7.4. This is why it is good we discuss this now and figure out\nwhere we want to go for 7.4 so we can get started.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 13:20:09 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Mike Mascari wrote:\n> Okay. But please keep in mind that a 2-phase commit implementation\n> is used for more than just replication. Any distributed TX will\n> require a 2PC protocol. As an example, for the DBLINK implementation\n> to ultimately be transaction safe (at least amongst multiple\n> PostgreSQL installations), the players in the distributed\n> transaction must all be participants in a 2PC exchange. And a\n> participant whose communications link is dropped needs to be\n> able to recover by asking the coordinator whether or not to\n> complete or abort the distributed TX. I am 100% ignorant of the\n> distributed TX standard Tom referenced earlier, but I'd guess\n> there might be an assumption of 2PC support in the implementation.\n> In other words, I think we still need 2PC, regardless of the\n> method of replication. And if Satoshi Nagayasu has an implementation\n> ready, why not investigate its possibilities?\n\nThis is a good point. I don't want to push Postgres-R as our solution. \nRather, I have looked at both and like Postgres-R, but others need to\nlook at both and decide so we are all in agreement when we move forward.\n\n--\n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 13:22:53 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Bruce Momjian wrote:\n\n> OK, the first thing is that there isn't any one replication solution\n> that will behave optimally in all situations.\n\nRight\n\n> Now, let me describe Postgres-R and then the other replication\n> solutions. Postgres-R is multi-master, meaning you can send SELECT and\n> UPDATE/DELETE queries to any of the servers in the cluster, and get the\n> same result. It is also synchronous, meaning it doesn't update the\n> local copy until it is sure the other nodes agree to the change. It\n> allows failover, because if one node goes down, the others keep going.\n\nWrong\n\nIt is asynchronous without the need of 2 phase commit. It is group\ncommunication based and requires the group communication system to\nguarantee total order. The tricky part is, that the local transaction\nmust be on hold until the own commit message comes back without a prior\nlock conflict by a replication transaction. If such a lock confict\noccurs, the replication transaction wins and the local transaction rolls\nback.\n\n> \n> Now, let me contrast:\n> \n> rserv and dbmirror do master/slave. There is no mechanism to allow you\n> to do updates on the slave, and have them propagate to the master. You\n> can, however, send SELECT queries to the slave, and in fact that's how\n> usogres does load balancing.\n\nBut you cannot use the result of such a SELECT to update anything. So\nyou can only phase out complete read only transaction to the slaves.\nRequires support from the application since the load balancing system\ncannot know automatically what will be a read only transaction and what\nnot.\n\n> \n> Two-phase commit is probably the most popular commercial replication\n> solution. While it works for multi-master, it suffers from poor\n> performance and doesn't handle cases where one node disappears very\n> well.\n> \n> Another replication need is for asynchronous replication, most\n> traditionally for traveling salesmen who need to update their databases\n> periodically. The only solution I know for that is PeerDirect's\n> PostgreSQL commercial offering at http://www.peerdirect.com. It is\n> possible PITR may help with this, but we need to handle propagating\n> changes made by the salesmen back up into the server, and to do that, we\n> will need a mechanism to handle conflicts that occur when two people\n> update the same records. This is always a problem for asynchronous\n> replication.\n\nPITR doesn't help here at all, since PeerDirect's replication is trigger\nand control table based. What makes our replication system unique is\nthat it works bidirectional in a heterogenious world.\n\n> I will describe the use of 'spread' and the Postgres-R internal issues\n> in my next email.\n\nThe last time i was playing with spread (that was at Great Bridge in\nNorfolk), it was IMHO useless (for Postgres-R) because it sometimes\ndropped messages when the network load got too high. This occured\nwithout any indication, no error, nothing. This is not exactly what I\nunderstand as total order. I hope they have made some substantial\nprogress on that.\n\n\nJan\n\n-- \n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== JanWieck@Yahoo.com #\n", "msg_date": "Fri, 13 Dec 2002 13:36:59 -0500", "msg_from": "Jan Wieck <JanWieck@Yahoo.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "----- Original Message ----- \nFrom: \"Bruce Momjian\" <pgman@candle.pha.pa.us>\n\n> Mike Mascari wrote:\n> > Okay. But please keep in mind that a 2-phase commit implementation\n> > is used for more than just replication. \n> \n> This is a good point. I don't want to push Postgres-R as our solution. \n> Rather, I have looked at both and like Postgres-R, but others need to\n> look at both and decide so we are all in agreement when we move forward.\n> \n\nAfter having read your post regarding Spread, I see that it is an alternative to 2PC as a distributed TX protocol. If I understand you correctly, a DBLINK implementation built atop Spread would also be possible. Correct? The question then is, do other RDBMS expose a 2PC implementation which could not then be leveraged at a later time? For example imagine:\n\n1. 7.4 includes a native 2PC protocol with:\n\nCREATE DATABASE LINK accounting\nCONNECT TO accounting.acme.com:5432\nIDENTIFIED BY mascarm/mascarm;\n\nSELECT *\nFROM employees@accounting;\n\nINSERT INTO employees@accounting\nVALUES (1, 'Mike', 'Mascari');\n\nThat would be great, allowing PostgreSQL servers running in different departments to participate in a distributed tx. \n\n2. 7.5 includes a DBLINK which supports PostgreSQL participating in a heterogenous distributed transaction (with say, an Oracle database):\n\nCREATE DATABASE LINK finance\nCONNECT TO <oracle names entry>\nIDENTIFIED BY mascarm/mascarm\nUSING INTERFACE 'pg2oracle.so';\n\nINSERT INTO employees@finance\nVALUES (1, 'Mike', 'Mascari');\n\nI guess I'm basically asking:\n\n1) Is it necessary to *choose* between support for 2PC and Spread (Postgres-R) or can't we have both? Spread for Replication, 2PC for non-replicating distributed TX?\n\n2) Do major SQL DBMS vendors which support distributed options expose a callable interface into a 2PC protocol that would allow PostgreSQL to participate? I could check on this...\n\n3) Are there any standards (besides ODBC, which, the last time I looked just had COMMIT/ABORT APIs), that have been defined and adopted by the industry for distributed tx?\n\nAgain, I'd guess most people want:\n\n1) High performance Master/Master replication *and* (r.e. Postgres-R)\n2) Ability to participate in distrubuted tx's (r.e. 2PC?)\n\nMike Mascari\nmascarm@mascari.com\n\n\n\n", "msg_date": "Fri, 13 Dec 2002 13:51:17 -0500", "msg_from": "\"Mike Mascari\" <mascarm@mascari.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "On Fri, 2002-12-13 at 13:20, Bruce Momjian wrote:\n> Let me address the Spread issue and two-phase commit. (Spread is an\n> open source piece of software used by Postgres-R.)\n\nNote that while Spread is open source in the sense that \"the source is\navailable\", it's license is significantly more restrictive than\nPostgreSQL's:\n\n http://www.spread.org/license/\n\nJust FYI...\n\nCheers,\n\nNeil\n-- \nNeil Conway <neilc@samurai.com> || PGP Key ID: DB3C29FC\n\n\n\n", "msg_date": "13 Dec 2002 14:01:30 -0500", "msg_from": "Neil Conway <neilc@samurai.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Neil Conway wrote:\n> On Fri, 2002-12-13 at 13:20, Bruce Momjian wrote:\n> > Let me address the Spread issue and two-phase commit. (Spread is an\n> > open source piece of software used by Postgres-R.)\n> \n> Note that while Spread is open source in the sense that \"the source is\n> available\", it's license is significantly more restrictive than\n> PostgreSQL's:\n> \n> http://www.spread.org/license/\n> \n\nInteresting. It looks like a modified version of the old BSD license\nwhere you are required to mention you are using Spread. I believe we\ncan get that reduced. (I think Darren already addressed this with\nthem.) We certainly are not going to accept software that requires all\nPostgreSQL user sites to mention Spread.\n\nThe whole \"mention\" aspect of the old BSD license was pretty ambiguous,\nand I assume this is similar.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 14:09:16 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "I wrote:\n> \n> I guess I'm basically asking:\n> \n> 1) Is it necessary to *choose* between support for 2PC and Spread (Postgres-R) or can't we have both? Spread for Replication, 2PC for non-replicating distributed TX?\n> \n> 2) Do major SQL DBMS vendors which support distributed options expose a callable interface into a 2PC protocol that would allow PostgreSQL to participate? I could check on this...\n>\n> 3) Are there any standards (besides ODBC, which, the last time I looked just had COMMIT/ABORT APIs), that have been defined and adopted by the industry for distributed tx?\n\nAnswer:\n\nThe Open Group's Open/XA C193 specificiation for API for distributed transactions:\n\nhttp://www.opengroup.org/public/pubs/catalog/c193.htm\n\nI couldn't find any draft copies on the web, but a good description at the Sybase site:\n\nhttp://manuals.sybase.com/onlinebooks/group-xs/xsg1111e/xatuxedo/@ebt-link;pt=61?target=%25N%13_446_START_RESTART_N%25\n\nThe standard is 2PC based.\n\nMike Mascari\nmascarm@mascari.com\n\n\n", "msg_date": "Fri, 13 Dec 2002 14:34:54 -0500", "msg_from": "\"Mike Mascari\" <mascarm@mascari.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Bruce Momjian wrote:\n> \n> Joe Conway wrote:\n> > Bruce Momjian wrote:\n> > > Win32 Port:\n> > >\n> > > Katie Ward and Jan are working on contributing their Win32\n> > > port for 7.4. They plan to have a patch available by the end of\n> > > December.\n> >\n> > I have .Net Studio available to me, so if you need help in merging or testing\n> > or whatever, let me know.\n> \n> OK, Jan, let him know how he can help.\n\nMy current plan is to comb out the Win32 port only from what we've done\nall together against 7.2.1. The result should be a clean patch that\napplied against 7.2.1 builds a native windows port.\n\n From there, this patch must be lifted up to 7.4.\n\nI have the original context diff now down from 160,000 lines to 80,000\nlines. I think I will have the clean diff against 7.2.1 somewhere next\nweek. That would IMHO be a good time for Tom to start complaining so\nthat we can work in the required changes during the 7.4 lifting. ;-)\n\n\nJan\n\n-- \n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== JanWieck@Yahoo.com #\n", "msg_date": "Fri, 13 Dec 2002 14:45:43 -0500", "msg_from": "Jan Wieck <JanWieck@Yahoo.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Jan Wieck wrote:\n> Bruce Momjian wrote:\n> \n> > OK, the first thing is that there isn't any one replication solution\n> > that will behave optimally in all situations.\n> \n> Right\n> \n> > Now, let me describe Postgres-R and then the other replication\n> > solutions. Postgres-R is multi-master, meaning you can send SELECT and\n> > UPDATE/DELETE queries to any of the servers in the cluster, and get the\n> > same result. It is also synchronous, meaning it doesn't update the\n> > local copy until it is sure the other nodes agree to the change. It\n> > allows failover, because if one node goes down, the others keep going.\n> \n> Wrong\n> \n> It is asynchronous without the need of 2 phase commit. It is group\n\nWell, Darren's PDF at:\n\n\tftp://gborg.postgresql.org/pub/pgreplication/stable/PostgreSQLReplication.pdf.gz\n\ncalls Postgres-R \"Type: Embedded, Peer-to-Peer, Sync\". I don't know\nenough about replication so I will let you fight it out with him. ;-)\n\n> communication based and requires the group communication system to\n> guarantee total order. The tricky part is, that the local transaction\n> must be on hold until the own commit message comes back without a prior\n> lock conflict by a replication transaction. If such a lock confict\n> occurs, the replication transaction wins and the local transaction rolls\n> back.\n\nYep, that's the tricky part.\n\n> > \n> > Now, let me contrast:\n> > \n> > rserv and dbmirror do master/slave. There is no mechanism to allow you\n> > to do updates on the slave, and have them propagate to the master. You\n> > can, however, send SELECT queries to the slave, and in fact that's how\n> > usogres does load balancing.\n> \n> But you cannot use the result of such a SELECT to update anything. So\n> you can only phase out complete read only transaction to the slaves.\n> Requires support from the application since the load balancing system\n> cannot know automatically what will be a read only transaction and what\n> not.\n\nGood point. It has to be a read-only session because you can't jump\nnodes during a session. That definately limits its usefulness.\n\n> > Two-phase commit is probably the most popular commercial replication\n> > solution. While it works for multi-master, it suffers from poor\n> > performance and doesn't handle cases where one node disappears very\n> > well.\n> > \n> > Another replication need is for asynchronous replication, most\n> > traditionally for traveling salesmen who need to update their databases\n> > periodically. The only solution I know for that is PeerDirect's\n> > PostgreSQL commercial offering at http://www.peerdirect.com. It is\n> > possible PITR may help with this, but we need to handle propagating\n> > changes made by the salesmen back up into the server, and to do that, we\n> > will need a mechanism to handle conflicts that occur when two people\n> > update the same records. This is always a problem for asynchronous\n> > replication.\n> \n> PITR doesn't help here at all, since PeerDirect's replication is trigger\n> and control table based. What makes our replication system unique is\n> that it works bidirectional in a heterogenious world.\n\nI was only suggesting that PITR _may_ help as an archive method for the\nchanges. PeerDirect stores those changes in a table?\n\n> > I will describe the use of 'spread' and the Postgres-R internal issues\n> > in my next email.\n> \n> The last time i was playing with spread (that was at Great Bridge in\n> Norfolk), it was IMHO useless (for Postgres-R) because it sometimes\n> dropped messages when the network load got too high. This occured\n> without any indication, no error, nothing. This is not exactly what I\n> understand as total order. I hope they have made some substantial\n> progress on that.\n\nThat's a serious problem, clearly. Hopefully it is either fixed or it\nwill get fixed. We can't use it without reliability.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 16:09:44 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "On Fri, 2002-12-13 at 13:36, Jan Wieck wrote:\n> But you cannot use the result of such a SELECT to update anything. So\n> you can only phase out complete read only transaction to the slaves.\n> Requires support from the application since the load balancing system\n> cannot know automatically what will be a read only transaction and what\n> not.\n\nInteresting -- SQL contains the concept of \"read only\" and \"read write\"\ntransactions (the default is RW). If we implemented that (which\nshouldn't be too difficult[1]), it might make differentiating between\nclasses of transactions a little easier. Client applications would still\nneed to be modified, but not nearly as much.\n\nDoes this sound like it's worth doing?\n\n[1] -- AFAICS, the only tricky implementation detail is deciding exactly\nwhich database operations are \"writes\". Does nextval() count, for\nexample?\n\nCheers,\n\nNeil\n-- \nNeil Conway <neilc@samurai.com> || PGP Key ID: DB3C29FC\n\n\n\n", "msg_date": "13 Dec 2002 18:38:33 -0500", "msg_from": "Neil Conway <neilc@samurai.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Neil Conway wrote:\n> On Fri, 2002-12-13 at 13:36, Jan Wieck wrote:\n> > But you cannot use the result of such a SELECT to update anything. So\n> > you can only phase out complete read only transaction to the slaves.\n> > Requires support from the application since the load balancing system\n> > cannot know automatically what will be a read only transaction and what\n> > not.\n> \n> Interesting -- SQL contains the concept of \"read only\" and \"read write\"\n> transactions (the default is RW). If we implemented that (which\n> shouldn't be too difficult[1]), it might make differentiating between\n> classes of transactions a little easier. Client applications would still\n> need to be modified, but not nearly as much.\n> \n> Does this sound like it's worth doing?\n> \n> [1] -- AFAICS, the only tricky implementation detail is deciding exactly\n> which database operations are \"writes\". Does nextval() count, for\n> example?\n\nYou can't migrate a session between nodes, so the entire _session_ has\nto be read-only.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 19:48:09 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "> This is a good point. I don't want to push Postgres-R as our solution.\n> Rather, I have looked at both and like Postgres-R, but others need to\n> look at both and decide so we are all in agreement when we move forward.\n\nI think in either way, it's clear that they need to be in the main CVS, in\norder for it to get up to speed.\n\nChris\n\n\n", "msg_date": "Sat, 14 Dec 2002 09:20:16 +0800", "msg_from": "\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "On Fri, 2002-12-13 at 20:20, Christopher Kings-Lynne wrote:\n> > This is a good point. I don't want to push Postgres-R as our solution.\n> > Rather, I have looked at both and like Postgres-R, but others need to\n> > look at both and decide so we are all in agreement when we move forward.\n> \n> I think in either way, it's clear that they need to be in the main CVS, in\n> order for it to get up to speed.\n\nWhy's that?\n\nCheers,\n\nNeil\n-- \nNeil Conway <neilc@samurai.com> || PGP Key ID: DB3C29FC\n\n\n\n", "msg_date": "13 Dec 2002 23:50:09 -0500", "msg_from": "Neil Conway <neilc@samurai.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "On Friday 13 December 2002 11:01 pm, you wrote:\n> Good. This is the discussion we need. Let me quote the TODO list\n> replication section first:\n>\n> * Add replication of distributed databases [replication]\n> o automatic failover\n\nVery good. We need that for HA.\n\n> o load balancing\n\n> o master/slave replication\n> o multi-master replication\n> o partition data across servers\n\nI am interested in this for multitude of reasons. Scalability is obviously one \nof them. But I was just wondering about some things(After going thr. all the \nmessages on this).\n\nOnce we have partitioning and replication, that effectively means database \ncache can span multiple machines and no longer restricted by shared memory. \nSo will it work on mosix now? Just a thought.\n\n\n> OK, the first thing is that there isn't any one replication solution\n> that will behave optimally in all situations.\n>\n> Now, let me describe Postgres-R and then the other replication\n> solutions. Postgres-R is multi-master, meaning you can send SELECT and\n> UPDATE/DELETE queries to any of the servers in the cluster, and get the\n> same result. It is also synchronous, meaning it doesn't update the\n> local copy until it is sure the other nodes agree to the change. It\n> allows failover, because if one node goes down, the others keep going.\n>\n> Now, let me contrast:\n>\n> rserv and dbmirror do master/slave. There is no mechanism to allow you\n> to do updates on the slave, and have them propagate to the master. You\n> can, however, send SELECT queries to the slave, and in fact that's how\n> usogres does load balancing.\n\nSeems like mirroring v/s striping to me. Can we have both combined in either \nfashion just like RAID.\n\nMost importantly will it be able to resize the cluster on the fly? Are we \nlooking at network management of database like Oracle does. (OK the tools are \nunwarranted in many situation but it has to offer it).\n\nMost importantly I would like to see this thing easy to setup from a one \npoint-of-administration view.\n\nSomething like declare a cluster of these n1 machines as database partitions \nand have these another n2 machine do a slave sync with them for handling \nloads. If these kind of command-line options are there, adding easy tools on \ntop of them should be a pop.\n\nAnd please, in place database upgrades. Otherwise it will be a huge pain to \nmaintain such a cluster over long period of times.\n\n\n> Another replication need is for asynchronous replication, most\n> traditionally for traveling salesmen who need to update their databases\n> periodically. The only solution I know for that is PeerDirect's\n> PostgreSQL commercial offering at http://www.peerdirect.com. It is\n> possible PITR may help with this, but we need to handle propagating\n> changes made by the salesmen back up into the server, and to do that, we\n> will need a mechanism to handle conflicts that occur when two people\n> update the same records. This is always a problem for asynchronous\n> replication.\n\nWe need not offer entire asynchronous replication all at once. We can have \nlevels of asynchronous replication like read only(Syncing essentially) and \nRead-write. Even if we get slave sync only at first, that will be huge plus. \n\n> > 2 If we are going to have replication, can we have built in load\n> > balancing? Is it a good idea to have it in postgresql or a\n> > separate application would be way to go?\n>\n> Well, because Postgres-R is multi-master, it has automatic load\n> balancing. You can have your users point to whatever node you want.\n> You can implement this \"pointing\" by using dns IP address cycling, or\n> have a router that auto-load balances, though you would need to keep a\n> db session on the same node, of course.\n\nUmm. W.r.t above point i.e. combining data partitioning and slave-sync, will \nit take a partitioned cluster as a single server or that cluster can take \ncare of itself in such situattions?\n\n>\n> So, in summary, I think we will eventually have two directions for\n> replication. One is Postgres-R for multi-master, synchronous\n> replication, and PITR, for asynchronous replication. I don't think\n\nI would put that as two options rather than directions. We need to be able to \ndeploy them both if required.\n\nImagine postgresql running over 500 machine cluster..;-)\n\n Shridhar\n", "msg_date": "Sat, 14 Dec 2002 11:09:30 +0530", "msg_from": "Shridhar Daithankar <shridhar_daithankar@persistent.co.in>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Bruce Momjian wrote:\n> Joe Conway wrote:\n<snip>\n>>>Point-In-Time Recovery (PITR)\n>>>\n>>>\tJ. R. Nield did a PITR patch late in 7.3 development, and Patrick\n>>>\tMacDonald from Red Hat is working on merging it into CVS and\n>>>\tadding any missing pieces. Patrick, do you have an ETA on that?\n>>\n>>As Hannu asked (and related to your question below), is there any thought of \n>>extending this to allow simple log based replication? In many important \n>>scenarios that would be more than adequate, and simpler to set up.\n<snip>\n\nFor PITR-log-based-replication, how much data would be required to be pushed out to each slave system in order to bring \nit up to date?\n\nI'm having visions of a 16MB WAL file being pushed out to slave systems in order to update them with a few rows of data...\n\n:-/\n\nRegards and best wishes,\n\nJustin Clift\n\n-- \n\"My grandfather once told me that there are two kinds of people: those\nwho work and those who take the credit. He told me to try to be in the\nfirst group; there was less competition there.\"\n- Indira Gandhi\n\n", "msg_date": "Sat, 14 Dec 2002 18:02:49 +1100", "msg_from": "Justin Clift <justin@postgresql.org>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "On 14 Dec 2002 at 18:02, Justin Clift wrote:\n> For PITR-log-based-replication, how much data would be required to be pushed out to each slave system in order to bring \n> it up to date?\n> \n> I'm having visions of a 16MB WAL file being pushed out to slave systems in order to update them with a few rows of data...\n\nI was under impression that data is pushed to slave after a checkpoint is \ncomplete. i.e. 16MB of WAL file has recycled.\n\nConversely a slave would contain accurate data upto last WAL checkpoint.\n\nI think tunable WAL size should be of some help in such scenario. Otherwise the \nsystem designer has to use async. replication. for granularity upto a \ntransaction.\n\n\nBye\n Shridhar\n\n--\nConference, n.:\tA special meeting in which the boss gathers subordinates to \nhear\twhat they have to say, so long as it doesn't conflict with what\the's \nalready decided to do.\n\n", "msg_date": "Sat, 14 Dec 2002 13:04:43 +0530", "msg_from": "\"Shridhar Daithankar\" <shridhar_daithankar@persistent.co.in>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "> > I think in either way, it's clear that they need to be in the main CVS, in\n> > order for it to get up to speed.\n>\n> Why's that?\n\nBecause until replication is in CVS, it won't be used, tested and improved\nand developed as fast...\n\nChris\n\n\n", "msg_date": "Sat, 14 Dec 2002 17:53:03 +0800 (WST)", "msg_from": "Christopher Kings-Lynne <chriskl@familyhealth.com.au>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "On Fri, 2002-12-13 at 04:53, Hannu Krosing wrote:\n> On Fri, 2002-12-13 at 06:22, Bruce Momjian wrote:\n> > I wanted to outline some of the big items we are looking at for 7.4:\n> > Point-In-Time Recovery (PITR)\n> > \n> > \tJ. R. Nield did a PITR patch late in 7.3 development, and Patrick\n> > \tMacDonald from Red Hat is working on merging it into CVS and\n> > \tadding any missing pieces. Patrick, do you have an ETA on that?\n> \n> How hard would it be to extend PITR for master-slave (hot backup)\n> repliaction, which should then amount to continuously shipping logs to\n> slave and doing nonstop PITR there :)\n> \n> It will never be usable for multi-master replication, but somehow it\n> feels that for master-slave replication simple log replay would be most\n> simple and robust solution.\n\nI'm curious, what would be the recovery strategy for PITR master-slave\nreplication should the master fail (assuming hot fail over/backup)? A\nsimple dump/restore? Are there/is there any facilities in PorstgreSQL\nfor PITR archival which prevents PITR logs from be recycled (or perhaps,\nsimply archived off)? What about PITR streaming to networked and/or\nremovable media?\n\n-- \nGreg Copeland <greg@copelandconsulting.net>\nCopeland Computer Consulting\n\n", "msg_date": "16 Dec 2002 07:56:25 -0600", "msg_from": "Greg Copeland <greg@CopelandConsulting.Net>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "On Monday 16 December 2002 07:26 pm, you wrote:\n> I'm curious, what would be the recovery strategy for PITR master-slave\n> replication should the master fail (assuming hot fail over/backup)? A\n> simple dump/restore? Are there/is there any facilities in PorstgreSQL\n> for PITR archival which prevents PITR logs from be recycled (or perhaps,\n> simply archived off)? What about PITR streaming to networked and/or\n> removable media?\n\nIn asynchrounous replication, WAL log records are fed to anoter host, which \nreplays those transactions to sync the data. This way it does not matter if \nWAL log is recycled as it is already replicated someplace else..\n\nHTH\n\n Shridhar\n", "msg_date": "Mon, 16 Dec 2002 19:32:58 +0530", "msg_from": "Shridhar Daithankar <shridhar_daithankar@persistent.co.in>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "I must of miscommunicated here as you're describing PITR replication. \nI'm asking about a master failing and the slaving picking up. Now, some\nn-time later, how do you recover your master system to be back in sync\nwith the slave. Obviously, I'm assuming some level of manual recovery. \nI'm wondering what the general approach would be.\n\nConsider that on the slave which is now the active server (master dead),\nit's possible that the slave's PITR's will be recycled before the master\ncan come back up. As such, unless there is a, an archival process for\nPITR or b, a method of streaming PITR's off for archival, the odds of\nusing PITR to recover the master (resync if you will) seem greatly\nreduced as you will be unable to replay PITR on the master for\nsynchronization.\n\nGreg\n\n\n\nOn Mon, 2002-12-16 at 08:02, Shridhar Daithankar wrote:\n> On Monday 16 December 2002 07:26 pm, you wrote:\n> > I'm curious, what would be the recovery strategy for PITR master-slave\n> > replication should the master fail (assuming hot fail over/backup)? A\n> > simple dump/restore? Are there/is there any facilities in PorstgreSQL\n> > for PITR archival which prevents PITR logs from be recycled (or perhaps,\n> > simply archived off)? What about PITR streaming to networked and/or\n> > removable media?\n> \n> In asynchrounous replication, WAL log records are fed to anoter host, which \n> replays those transactions to sync the data. This way it does not matter if \n> WAL log is recycled as it is already replicated someplace else..\n> \n> HTH\n> \n> Shridhar\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 4: Don't 'kill -9' the postmaster\n-- \nGreg Copeland <greg@copelandconsulting.net>\nCopeland Computer Consulting\n\n", "msg_date": "16 Dec 2002 08:13:13 -0600", "msg_from": "Greg Copeland <greg@CopelandConsulting.Net>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "On Monday 16 December 2002 07:43 pm, you wrote:\n> Consider that on the slave which is now the active server (master dead),\n> it's possible that the slave's PITR's will be recycled before the master\n> can come back up. As such, unless there is a, an archival process for\n> PITR or b, a method of streaming PITR's off for archival, the odds of\n> using PITR to recover the master (resync if you will) seem greatly\n> reduced as you will be unable to replay PITR on the master for\n> synchronization.\n\nI agree. Since we are talking about features in future release, I think it \nshould be added to TODO if not already there.\n\nI don't know about WAL numbering but AFAIU, it increments and old files are \nremoved once there are enough WAL files as specified in posgresql.conf. IIRC \nthere are some perl based replication project exist already which use this \nfeature.\n\n Shridhar\n\n", "msg_date": "Mon, 16 Dec 2002 19:50:09 +0530", "msg_from": "Shridhar Daithankar <shridhar_daithankar@persistent.co.in>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "On Mon, 2002-12-16 at 08:20, Shridhar Daithankar wrote:\n> On Monday 16 December 2002 07:43 pm, you wrote:\n> > Consider that on the slave which is now the active server (master dead),\n> > it's possible that the slave's PITR's will be recycled before the master\n> > can come back up. As such, unless there is a, an archival process for\n> > PITR or b, a method of streaming PITR's off for archival, the odds of\n> > using PITR to recover the master (resync if you will) seem greatly\n> > reduced as you will be unable to replay PITR on the master for\n> > synchronization.\n> \n> I agree. Since we are talking about features in future release, I think it \n> should be added to TODO if not already there.\n> \n> I don't know about WAL numbering but AFAIU, it increments and old files are \n> removed once there are enough WAL files as specified in posgresql.conf. IIRC \n> there are some perl based replication project exist already which use this \n> feature.\n> \n\nThe problem with this is that most people, AFAICT, are going to size WAL\nbased on their performance/sizing requirements and not based on\ntheoretical estimates which someone might make to allow for a window of\nfailure. That is, I don't believe increasing the number of WAL's is\ngoing to satisfactorily address the issue.\n\n\n-- \nGreg Copeland <greg@copelandconsulting.net>\nCopeland Computer Consulting\n\n", "msg_date": "16 Dec 2002 08:37:53 -0600", "msg_from": "Greg Copeland <greg@CopelandConsulting.Net>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "On Monday 16 December 2002 08:07 pm, you wrote:\n> On Mon, 2002-12-16 at 08:20, Shridhar Daithankar wrote:\n> > I don't know about WAL numbering but AFAIU, it increments and old files\n> > are removed once there are enough WAL files as specified in\n> > posgresql.conf. IIRC there are some perl based replication project exist\n> > already which use this feature.\n>\n> The problem with this is that most people, AFAICT, are going to size WAL\n> based on their performance/sizing requirements and not based on\n> theoretical estimates which someone might make to allow for a window of\n> failure. That is, I don't believe increasing the number of WAL's is\n> going to satisfactorily address the issue.\n\nSorry for not being clear. When I said, WAL numbering, I meant WAL naming \nconventions where numbers are used to mark WAL files. \n\nIt is not number of WAL files. It is entirely upto the installation and IIRC, \neven in replication project(Sorry I forgot the exact name), you can set \nnumber of WAL files that it can have.\n\n Shridhar\n\n", "msg_date": "Mon, 16 Dec 2002 20:12:29 +0530", "msg_from": "Shridhar Daithankar <shridhar_daithankar@persistent.co.in>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Shridhar Daithankar wrote:\n> On Monday 16 December 2002 08:07 pm, you wrote:\n> > On Mon, 2002-12-16 at 08:20, Shridhar Daithankar wrote:\n> > > I don't know about WAL numbering but AFAIU, it increments and old files\n> > > are removed once there are enough WAL files as specified in\n> > > posgresql.conf. IIRC there are some perl based replication project exist\n> > > already which use this feature.\n> >\n> > The problem with this is that most people, AFAICT, are going to size WAL\n> > based on their performance/sizing requirements and not based on\n> > theoretical estimates which someone might make to allow for a window of\n> > failure. That is, I don't believe increasing the number of WAL's is\n> > going to satisfactorily address the issue.\n> \n> Sorry for not being clear. When I said, WAL numbering, I meant WAL naming \n> conventions where numbers are used to mark WAL files. \n> \n> It is not number of WAL files. It is entirely upto the installation and IIRC, \n> even in replication project(Sorry I forgot the exact name), you can set \n> number of WAL files that it can have.\n\nBasically, PITR is going to have a way to archive off a log of database\nchanges, either from WAL or from somewhere else. At some point, there\nis going to have to be administrative action which says, \"I have a\nmaster down for three days. I am going to have to save my PITR logs for\nthat period.\" So, PITR will probably be used for recovery of a failed\nmaster, and such recover is going to have to require some administrative\naction _if_ the automatic expiration of PITR logs is shorter than the\nduration the master is down.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Mon, 16 Dec 2002 10:51:46 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Bruce Momjian wrote:\n> \n> I wanted to outline some of the big items we are looking at for 7.4:\n> \n> [snip]\n>\n> Point-In-Time Recovery (PITR)\n> \n> J. R. Nield did a PITR patch late in 7.3 development, and Patrick\n> MacDonald from Red Hat is working on merging it into CVS and\n> adding any missing pieces. Patrick, do you have an ETA on that?\n\nNeil Conway and I will be working on this starting the beginning\nof January. By the middle of January, we hope to have a handle on\nan ETA.\n\nCheers,\nPatrick\n--\nPatrick Macdonald\nRed Hat Database Development\n", "msg_date": "Mon, 16 Dec 2002 12:47:28 -0500", "msg_from": "Patrick Macdonald <patrickm@redhat.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Patrick Macdonald wrote:\n> Bruce Momjian wrote:\n> > \n> > I wanted to outline some of the big items we are looking at for 7.4:\n> > \n> > [snip]\n> >\n> > Point-In-Time Recovery (PITR)\n> > \n> > J. R. Nield did a PITR patch late in 7.3 development, and Patrick\n> > MacDonald from Red Hat is working on merging it into CVS and\n> > adding any missing pieces. Patrick, do you have an ETA on that?\n> \n> Neil Conway and I will be working on this starting the beginning\n> of January. By the middle of January, we hope to have a handle on\n> an ETA.\n\nEwe, that is later than I was hoping. I have put J.R's PITR patch up\nat:\n\n\tftp://candle.pha.pa.us/pub/postgresql/PITR_20020822_02.gz\n\n(I have tried to contact J.R. several times over the past few months,\nwith no reply.)\n\nJ.R felt it was ready to go. I would like to have an evaluation of the\npatch to know what it does and what is missing. I would like that\nsooner rather than later because:\n\n\to I want to avoid too much code drift\n\to I don't want to find there are major pieces missing and to \n\t not have enough time to implement them in 7.4\n\to It is a big feature so I would like sufficient testing before beta\n\nOK, I just talked to Patrick on the phone, and he says Neil Conway is\nworking on merging the code into 7.3, and adding missing pieces like\nlogging table creation. So, it seems PITR is moving forward. Neil, can\nyou comment on where you are with this, and what still needs to be done?\nDo we need to start looking at log archiving options? How is the PITR\nlog contents different from the WAL log contents, except of course no\npre-write page images?\n\nIf we need to discuss things, perhaps we can do it now and get folks\nworking on other pieces, or at least thinking about them.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Mon, 16 Dec 2002 13:38:37 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "On Mon, 2002-12-16 at 13:38, Bruce Momjian wrote:\n> OK, I just talked to Patrick on the phone, and he says Neil Conway is\n> working on merging the code into 7.3, and adding missing pieces like\n> logging table creation. So, it seems PITR is moving forward. Neil, can\n> you comment on where you are with this, and what still needs to be done?\n\nWell, I should have a preliminary merge of the old PITR patch with CVS\nHEAD done by Wednesday or Thursday. It took me a while to merge because\n(a) I've got final exams at university at the moment (b) I had to merge\nmost of it by hand, as much of the diff is a single hunk (!), for some\nreason.\n\nAs for the status of the code, I haven't really had a chance to evaluate\nit; as Patrick noted, I think we should be able to give you an ETA by\nthe middle of January or so (I'll be offline starting Thursday until the\nfirst week of January).\n\nCheers,\n\nNeil\n-- \nNeil Conway <neilc@samurai.com> || PGP Key ID: DB3C29FC\n\n\n\n", "msg_date": "16 Dec 2002 14:46:19 -0500", "msg_from": "Neil Conway <neilc@samurai.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Bruce Momjian wrote:\n> \n> Patrick Macdonald wrote:\n> > Bruce Momjian wrote:\n> > >\n> > > I wanted to outline some of the big items we are looking at for 7.4:\n> > >\n> > > [snip]\n> > >\n> > > Point-In-Time Recovery (PITR)\n> > >\n> > > J. R. Nield did a PITR patch late in 7.3 development, and Patrick\n> > > MacDonald from Red Hat is working on merging it into CVS and\n> > > adding any missing pieces. Patrick, do you have an ETA on that?\n> >\n> > Neil Conway and I will be working on this starting the beginning\n> > of January. By the middle of January, we hope to have a handle on\n> > an ETA.\n> \n> Ewe, that is later than I was hoping. I have put J.R's PITR patch up\n> at:\n> \n> ftp://candle.pha.pa.us/pub/postgresql/PITR_20020822_02.gz\n> \n> (I have tried to contact J.R. several times over the past few months,\n> with no reply.)\n> \n> J.R felt it was ready to go. I would like to have an evaluation of the\n> patch to know what it does and what is missing. I would like that\n> sooner rather than later because:\n> \n> o I want to avoid too much code drift\n> o I don't want to find there are major pieces missing and to\n> not have enough time to implement them in 7.4\n> o It is a big feature so I would like sufficient testing before beta\n> \n> OK, I just talked to Patrick on the phone, and he says Neil Conway is\n> working on merging the code into 7.3, and adding missing pieces like\n> logging table creation. So, it seems PITR is moving forward.\n\nWell, sort of. I stated that Neil was already working on merging the\npatch into the CVS tip. I also mentioned that there are missing \npieces but have no idea if Neil is currently working on them.\n\nCheers,\nPatrick\n--\nPatrick Macdonald\nRed Hat Database Development\n", "msg_date": "Mon, 16 Dec 2002 14:50:27 -0500", "msg_from": "Patrick Macdonald <patrickm@redhat.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "Patrick Macdonald wrote:\n> > OK, I just talked to Patrick on the phone, and he says Neil Conway is\n> > working on merging the code into 7.3, and adding missing pieces like\n> > logging table creation. So, it seems PITR is moving forward.\n> \n> Well, sort of. I stated that Neil was already working on merging the\n> patch into the CVS tip. I also mentioned that there are missing \n> pieces but have no idea if Neil is currently working on them.\n\nOh, OK. What I would like to do is find out what actually needs to be\ndone so we can get folks started on it. If we can get a 7.3 merge,\nmaybe we should get it into CVS and then list the items needing\nattention and folks can submit patches to implement those.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Mon, 16 Dec 2002 14:52:04 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "The file \"ftp://candle.pha.pa.us/pub/postgresql/PITR_20020822_02.gz\" \ndo not\nhave read permissions to copy. please provide the read permissions to copy.\n\nRegards\njana\n\n>Patrick Macdonald wrote:\n> \n>\n>>Bruce Momjian wrote:\n>> \n>>\n>>>I wanted to outline some of the big items we are looking at for 7.4:\n>>>\n>>>[snip]\n>>>\n>>>Point-In-Time Recovery (PITR)\n>>>\n>>> J. R. Nield did a PITR patch late in 7.3 development, and Patrick\n>>> MacDonald from Red Hat is working on merging it into CVS and\n>>> adding any missing pieces. Patrick, do you have an ETA on that?\n>>> \n>>>\n>>Neil Conway and I will be working on this starting the beginning\n>>of January. By the middle of January, we hope to have a handle on\n>>an ETA.\n>> \n>>\n>\n>Ewe, that is later than I was hoping. I have put J.R's PITR patch up\n>at:\n>\n>\tftp://candle.pha.pa.us/pub/postgresql/PITR_20020822_02.gz\n>\n>(I have tried to contact J.R. several times over the past few months,\n>with no reply.)\n>\n>J.R felt it was ready to go. I would like to have an evaluation of the\n>patch to know what it does and what is missing. I would like that\n>sooner rather than later because:\n>\n>\to I want to avoid too much code drift\n>\to I don't want to find there are major pieces missing and to \n>\t not have enough time to implement them in 7.4\n>\to It is a big feature so I would like sufficient testing before beta\n>\n>OK, I just talked to Patrick on the phone, and he says Neil Conway is\n>working on merging the code into 7.3, and adding missing pieces like\n>logging table creation. So, it seems PITR is moving forward. Neil, can\n>you comment on where you are with this, and what still needs to be done?\n>Do we need to start looking at log archiving options? How is the PITR\n>log contents different from the WAL log contents, except of course no\n>pre-write page images?\n>\n>If we need to discuss things, perhaps we can do it now and get folks\n>working on other pieces, or at least thinking about them.\n>\n> \n>\n\n\n\n\n\n\n\nThe  file  \"ftp://candle.pha.pa.us/pub/postgresql/PITR_20020822_02.gz\"\ndo not\nhave read permissions to copy. please provide the read permissions to\ncopy.\n\nRegards\njana\n\nPatrick Macdonald wrote:\n \n\nBruce Momjian wrote:\n \n\nI wanted to outline some of the big items we are looking at for 7.4:\n\n[snip]\n\nPoint-In-Time Recovery (PITR)\n\n J. R. Nield did a PITR patch late in 7.3 development, and Patrick\n MacDonald from Red Hat is working on merging it into CVS and\n adding any missing pieces. Patrick, do you have an ETA on that?\n \n\nNeil Conway and I will be working on this starting the beginning\nof January. By the middle of January, we hope to have a handle on\nan ETA.\n \n\n\nEwe, that is later than I was hoping. I have put J.R's PITR patch up\nat:\n\n\tftp://candle.pha.pa.us/pub/postgresql/PITR_20020822_02.gz\n\n(I have tried to contact J.R. several times over the past few months,\nwith no reply.)\n\nJ.R felt it was ready to go. I would like to have an evaluation of the\npatch to know what it does and what is missing. I would like that\nsooner rather than later because:\n\n\to I want to avoid too much code drift\n\to I don't want to find there are major pieces missing and to \n\t not have enough time to implement them in 7.4\n\to It is a big feature so I would like sufficient testing before beta\n\nOK, I just talked to Patrick on the phone, and he says Neil Conway is\nworking on merging the code into 7.3, and adding missing pieces like\nlogging table creation. So, it seems PITR is moving forward. Neil, can\nyou comment on where you are with this, and what still needs to be done?\nDo we need to start looking at log archiving options? How is the PITR\nlog contents different from the WAL log contents, except of course no\npre-write page images?\n\nIf we need to discuss things, perhaps we can do it now and get folks\nworking on other pieces, or at least thinking about them.", "msg_date": "Tue, 17 Dec 2002 10:35:49 +0800", "msg_from": "Janardhan <jana-reddy@mediaring.com.sg>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "\nOops, sorry. Permissions fixed.\n\n---------------------------------------------------------------------------\n\nJanardhan wrote:\n> The file \"ftp://candle.pha.pa.us/pub/postgresql/PITR_20020822_02.gz\" \n> do not\n> have read permissions to copy. please provide the read permissions to copy.\n> \n> Regards\n> jana\n> \n> >Patrick Macdonald wrote:\n> > \n> >\n> >>Bruce Momjian wrote:\n> >> \n> >>\n> >>>I wanted to outline some of the big items we are looking at for 7.4:\n> >>>\n> >>>[snip]\n> >>>\n> >>>Point-In-Time Recovery (PITR)\n> >>>\n> >>> J. R. Nield did a PITR patch late in 7.3 development, and Patrick\n> >>> MacDonald from Red Hat is working on merging it into CVS and\n> >>> adding any missing pieces. Patrick, do you have an ETA on that?\n> >>> \n> >>>\n> >>Neil Conway and I will be working on this starting the beginning\n> >>of January. By the middle of January, we hope to have a handle on\n> >>an ETA.\n> >> \n> >>\n> >\n> >Ewe, that is later than I was hoping. I have put J.R's PITR patch up\n> >at:\n> >\n> >\tftp://candle.pha.pa.us/pub/postgresql/PITR_20020822_02.gz\n> >\n> >(I have tried to contact J.R. several times over the past few months,\n> >with no reply.)\n> >\n> >J.R felt it was ready to go. I would like to have an evaluation of the\n> >patch to know what it does and what is missing. I would like that\n> >sooner rather than later because:\n> >\n> >\to I want to avoid too much code drift\n> >\to I don't want to find there are major pieces missing and to \n> >\t not have enough time to implement them in 7.4\n> >\to It is a big feature so I would like sufficient testing before beta\n> >\n> >OK, I just talked to Patrick on the phone, and he says Neil Conway is\n> >working on merging the code into 7.3, and adding missing pieces like\n> >logging table creation. So, it seems PITR is moving forward. Neil, can\n> >you comment on where you are with this, and what still needs to be done?\n> >Do we need to start looking at log archiving options? How is the PITR\n> >log contents different from the WAL log contents, except of course no\n> >pre-write page images?\n> >\n> >If we need to discuss things, perhaps we can do it now and get folks\n> >working on other pieces, or at least thinking about them.\n> >\n> > \n> >\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Mon, 16 Dec 2002 22:26:54 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "So if this gets added to the 7.3 branch, will there be documentation \naccompanying it?\n\n-tfo\n\nIn article <200212161838.gBGIcb717436@candle.pha.pa.us>,\n pgman@candle.pha.pa.us (Bruce Momjian) wrote:\n\n> OK, I just talked to Patrick on the phone, and he says Neil Conway is\n> working on merging the code into 7.3, and adding missing pieces like\n> logging table creation. So, it seems PITR is moving forward.\n", "msg_date": "Tue, 17 Dec 2002 11:15:41 -0600", "msg_from": "Thomas O'Connell <tfo@monsterlabs.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "\nI meant he is merging it into HEAD, not the 7.3 CVS. Sorry for the\nconfusion.\n\n---------------------------------------------------------------------------\n\nThomas O'Connell wrote:\n> So if this gets added to the 7.3 branch, will there be documentation \n> accompanying it?\n> \n> -tfo\n> \n> In article <200212161838.gBGIcb717436@candle.pha.pa.us>,\n> pgman@candle.pha.pa.us (Bruce Momjian) wrote:\n> \n> > OK, I just talked to Patrick on the phone, and he says Neil Conway is\n> > working on merging the code into 7.3, and adding missing pieces like\n> > logging table creation. So, it seems PITR is moving forward.\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 4: Don't 'kill -9' the postmaster\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 12:21:19 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" } ]
[ { "msg_contents": "\n\n> -----Original Message-----\n> From: \tSteve King \n> Sent:\t12 December 2002 11:45\n> To:\tpgsql-hackers@postgresql.org\n> Subject:\tDuplicate oids!\n> \n> Forgive me if this is a previous question but I cannot find any\n> information on it in any of the mailing lists.\n> \n> I have a postgres database that contains a table with two identical\n> records including the oid.\n> It seems as though one insert statement (intending one record to be\n> inserted) has caused two identical records to be inserted.\n> The insert statement was done via the c++ library.\n> \n> Does anyone know anything about this?\n> \n> My info can be supplied if this is not a known problem!\n> \n> P.S. I am running Postgres 7.2\n", "msg_date": "Fri, 13 Dec 2002 09:27:08 -0000", "msg_from": "Steve King <steve.king@ecmsys.co.uk>", "msg_from_op": true, "msg_subject": "FW: Duplicate oids!" }, { "msg_contents": "On Fri, 2002-12-13 at 09:27, Steve King wrote:\n> > -----Original Message-----\n> > From: \tSteve King \n> > Sent:\t12 December 2002 11:45\n> > To:\tpgsql-hackers@postgresql.org\n> > Subject:\tDuplicate oids!\n> > \n> > Forgive me if this is a previous question but I cannot find any\n> > information on it in any of the mailing lists.\n> > \n> > I have a postgres database that contains a table with two identical\n> > records including the oid.\n\nWhat about ctid's, are they also the same ? \n\nAre the tuples on the same page ?\n\n> > It seems as though one insert statement (intending one record to be\n> > inserted) has caused two identical records to be inserted.\n> > The insert statement was done via the c++ library.\n> > \n> > Does anyone know anything about this?\n> > \n> > My info can be supplied if this is not a known problem!\n> > \n> > P.S. I am running Postgres 7.2\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 6: Have you searched our list archives?\n> \n> http://archives.postgresql.org\n-- \nHannu Krosing <hannu@tm.ee>\n", "msg_date": "13 Dec 2002 11:53:35 +0000", "msg_from": "Hannu Krosing <hannu@tm.ee>", "msg_from_op": false, "msg_subject": "Re: FW: Duplicate oids!" } ]
[ { "msg_contents": "The ctids are different however vaccum is run on this table and the record\nis updated.\nThe machineid is a SERIAL and so should also never be duplicated.\n\n ctid | oid | machineid\n--------+---------+-----------\n (7,18) | 9646238 | 12\n (7,10) | 9646238 | 12\n\n\nAny help as usual is greatly appreciated.\n-----Original Message-----\nFrom: Hannu Krosing [mailto:hannu@tm.ee]\nSent: 13 December 2002 11:54\nTo: Steve King\nCc: 'pgsql-hackers@postgresql.org'\nSubject: Re: [HACKERS] FW: Duplicate oids!\n\n\nOn Fri, 2002-12-13 at 09:27, Steve King wrote:\n> > -----Original Message-----\n> > From: \tSteve King \n> > Sent:\t12 December 2002 11:45\n> > To:\tpgsql-hackers@postgresql.org\n> > Subject:\tDuplicate oids!\n> > \n> > Forgive me if this is a previous question but I cannot find any\n> > information on it in any of the mailing lists.\n> > \n> > I have a postgres database that contains a table with two identical\n> > records including the oid.\n\nWhat about ctid's, are they also the same ? \n\nAre the tuples on the same page ?\n\n> > It seems as though one insert statement (intending one record to be\n> > inserted) has caused two identical records to be inserted.\n> > The insert statement was done via the c++ library.\n> > \n> > Does anyone know anything about this?\n> > \n> > My info can be supplied if this is not a known problem!\n> > \n> > P.S. I am running Postgres 7.2\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 6: Have you searched our list archives?\n> \n> http://archives.postgresql.org\n-- \nHannu Krosing <hannu@tm.ee>\n\n---------------------------(end of broadcast)---------------------------\nTIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org\n", "msg_date": "Fri, 13 Dec 2002 11:56:51 -0000", "msg_from": "Steve King <steve.king@ecmsys.co.uk>", "msg_from_op": true, "msg_subject": "Re: FW: Duplicate oids!" }, { "msg_contents": "Steve King <steve.king@ecmsys.co.uk> writes:\n> The ctids are different however vaccum is run on this table and the record\n> is updated.\n\nIt would be useful to look at xmin,xmax,cmin,cmax of these tuples too.\n\nActually, if you don't mind grabbing a copy of pg_filedump --- see\nhttp://sources.redhat.com/rhdb/tools.html --- then a dump of page 7\nof that relation would be good to look at. I am wondering about the\nstates of the infomask bits for these tuples...\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 13 Dec 2002 09:43:19 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: FW: Duplicate oids! " }, { "msg_contents": "On Fri, Dec 13, 2002 at 09:43:19AM -0500, Tom Lane wrote:\n\n> Actually, if you don't mind grabbing a copy of pg_filedump --- see\n> http://sources.redhat.com/rhdb/tools.html\n\nHas this been updated for 7.3? Last time I looked it only did 7.2, and\nthe site shows an old date. If it hasn't, are there plans to update it\nsometime soon? It would be very useful to me right now...\n\n-- \nAlvaro Herrera (<alvherre[a]dcc.uchile.cl>)\n\"Para tener mas hay que desear menos\"\n", "msg_date": "Fri, 13 Dec 2002 12:33:27 -0300", "msg_from": "Alvaro Herrera <alvherre@dcc.uchile.cl>", "msg_from_op": false, "msg_subject": "Re: FW: Duplicate oids!" }, { "msg_contents": "Alvaro Herrera <alvherre@dcc.uchile.cl> writes:\n> On Fri, Dec 13, 2002 at 09:43:19AM -0500, Tom Lane wrote:\n>> Actually, if you don't mind grabbing a copy of pg_filedump --- see\n>> http://sources.redhat.com/rhdb/tools.html\n\n> Has this been updated for 7.3? Last time I looked it only did 7.2, and\n> the site shows an old date. If it hasn't, are there plans to update it\n> sometime soon? It would be very useful to me right now...\n\nAFAIK it has not been updated yet. Patrick, do you have any near-term\nplans to do so? If not, perhaps Alvaro would like to do the legwork ;-)\n\nI believe it should be possible to make a single version of pg_filedump\nthat understands both the 7.2 and 7.3 page layouts (the version field in\nthe page header would work for telling what you're looking at).\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 13 Dec 2002 10:40:48 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: FW: Duplicate oids! " }, { "msg_contents": "Tom Lane wrote:\n> \n> Alvaro Herrera <alvherre@dcc.uchile.cl> writes:\n> > On Fri, Dec 13, 2002 at 09:43:19AM -0500, Tom Lane wrote:\n> >> Actually, if you don't mind grabbing a copy of pg_filedump --- see\n> >> http://sources.redhat.com/rhdb/tools.html\n> \n> > Has this been updated for 7.3? Last time I looked it only did 7.2, and\n> > the site shows an old date. If it hasn't, are there plans to update it\n> > sometime soon? It would be very useful to me right now...\n> \n> AFAIK it has not been updated yet. Patrick, do you have any near-term\n> plans to do so? If not, perhaps Alvaro would like to do the legwork ;-)\n\nYes, it's on my list of things to do. Look for an updated version\nby middle of the week (once all the RHDB 2.1 work is finished).\n \n> I believe it should be possible to make a single version of pg_filedump\n> that understands both the 7.2 and 7.3 page layouts (the version field in\n> the page header would work for telling what you're looking at)\n\nCorrect. The tool will be updated to understand the different\npage layouts/formats. Two tools would be a pain...\n\nCheers,\nPatrick\n--\nPatrick Macdonald\nRed Hat Database Development\n", "msg_date": "Mon, 16 Dec 2002 12:59:03 -0500", "msg_from": "Patrick Macdonald <patrickm@redhat.com>", "msg_from_op": false, "msg_subject": "Re: FW: Duplicate oids!" }, { "msg_contents": "Patrick Macdonald wrote:\n> \n> Tom Lane wrote:\n> >\n> > Alvaro Herrera <alvherre@dcc.uchile.cl> writes:\n> > > On Fri, Dec 13, 2002 at 09:43:19AM -0500, Tom Lane wrote:\n> > >> Actually, if you don't mind grabbing a copy of pg_filedump --- see\n> > >> http://sources.redhat.com/rhdb/tools.html\n> >\n> > > Has this been updated for 7.3? Last time I looked it only did 7.2, and\n> > > the site shows an old date. If it hasn't, are there plans to update it\n> > > sometime soon? It would be very useful to me right now...\n> >\n> > AFAIK it has not been updated yet. Patrick, do you have any near-term\n> > plans to do so? If not, perhaps Alvaro would like to do the legwork ;-)\n> \n> Yes, it's on my list of things to do. Look for an updated version\n> by middle of the week (once all the RHDB 2.1 work is finished).\n\nI've updated the pg_filedump utility for PostgreSQL 7.3. The new \nversion, 1.1, requires a PostgreSQL 7.3 source tree to build and \ncan be used against RHDB 2.x/1.x and PostgreSQL 7.3/7.2/7.1\ninstallations. \n\nAll questions and comments about the tool should be directed to \nrhdb@sources.redhat.com, not this list.\n\nThe pg_filedump utility can be found at the Red Hat Database Project\nsite (http://sources.redhat.com/rhdb).\n\nCheers,\nPatrick\n", "msg_date": "Wed, 18 Dec 2002 15:24:46 -0500", "msg_from": "Patrick Macdonald <patrickm@redhat.com>", "msg_from_op": false, "msg_subject": "Re: FW: Duplicate oids!" } ]
[ { "msg_contents": "> \n> How hard would it be to extend PITR for master-slave (hot backup)\n> repliaction, which should then amount to continuously shipping logs to\n> slave and doing nonstop PITR there :)\n\nI have not looked at the PITR patch yet, but it might be\npossible to use the same PITR format to queue/log writesetswith postgres-r, so we can have multi-master replication\nand PITR from the same mechanism.\n\nDarren\n\n", "msg_date": "Fri, 13 Dec 2002 9:13:26 -0500", "msg_from": "<darren@up.hrcoxmail.com>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "darren@up.hrcoxmail.com wrote:\n> >\n> > How hard would it be to extend PITR for master-slave (hot backup)\n> > repliaction, which should then amount to continuously shipping logs to\n> > slave and doing nonstop PITR there :)\n> \n> I have not looked at the PITR patch yet, but it might be possible\n> to use the same PITR format to queue/log writesetswith postgres-r,\n> so we can have multi-master replication and PITR from the same\n> mechanism.\n\nYes, we do need a method to send write sets to the various nodes, and\nPITR may be a help in getting those write sets. However, it should be\nclear that we really aren't archiving-replaying them like you would\nthink for PITR. We are only grabbing stuff from the PITR to send to\nother nodes. We may also be able to use PITR to bring nodes back up to\ndate if they have fallen out of communication.\n\n--\n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 12:42:22 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" } ]
[ { "msg_contents": "CVSROOT:\t/cvsroot\nModule name:\tpgsql-server\nChanges by:\ttgl@postgresql.org\t02/12/13 15:35:57\n\nModified files:\n\tsrc/test/regress: resultmap \nAdded files:\n\tsrc/test/regress/expected: geometry_1.out \nRemoved files:\n\tsrc/test/regress/expected: geometry-positive-zeros.out \n\nLog message:\n\tAdmit defeat on the prospect of keeping straight exactly which platforms\n\tproduce which output in the geometry test, even with the problem narrowed\n\tdown to only whether they print minus zero or not. Instead, use\n\tpg_regress' locale-variant mechanism to automatically consider the test\n\tto pass if it matches either supplied comparison file. geometry_1.out\n\treplaces the former geometry-positive-zeros.out.\n\n", "msg_date": "Fri, 13 Dec 2002 15:35:57 -0500 (EST)", "msg_from": "tgl@postgresql.org (Tom Lane)", "msg_from_op": true, "msg_subject": "pgsql-server/src/test/regress resultmap expect ..." }, { "msg_contents": "Tom Lane wrote:\n> CVSROOT:\t/cvsroot\n> Module name:\tpgsql-server\n> Changes by:\ttgl@postgresql.org\t02/12/13 15:35:57\n> \n> Modified files:\n> \tsrc/test/regress: resultmap \n> Added files:\n> \tsrc/test/regress/expected: geometry_1.out \n> Removed files:\n> \tsrc/test/regress/expected: geometry-positive-zeros.out \n> \n> Log message:\n> \tAdmit defeat on the prospect of keeping straight exactly which platforms\n> \tproduce which output in the geometry test, even with the problem narrowed\n> \tdown to only whether they print minus zero or not. Instead, use\n> \tpg_regress' locale-variant mechanism to automatically consider the test\n> \tto pass if it matches either supplied comparison file. geometry_1.out\n> \treplaces the former geometry-positive-zeros.out.\n\nOne issue is how are we going to know when a results file _isn't_ needed\nanymore, because we will not be matching regression files against\nplatforms anymore? Do we need a comment somewhere in the file?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 15:44:27 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: [COMMITTERS] pgsql-server/src/test/regress resultmap expect ..." } ]
[ { "msg_contents": "> > Note that while Spread is open source in the sense that \"the source is\n> > available\", it's license is significantly more restrictive than\n> > PostgreSQL's:\n> > \n> > http://www.spread.org/license/\n> > \n> \n> Interesting. It looks like a modified version of the old BSD license\n> where you are required to mention you are using Spread. I believe we\n> can get that reduced. (I think Darren already addressed this with\n> them.) We certainly are not going to accept software that requires all\n> PostgreSQL user sites to mention Spread.\n> \n\nI dont think this is the case. We don't redistribute spread\nfrom the pg-replication site. There are links to the down\nload area. I don't think this should be any different if\npostgres-r is merged with the main postgresql tree. If \nSpread is the group communication we choose to use for \npostgresql replication, then I would think some Spread \ninformation would be in order on the advocacy site, and \nin any set up documentation for replication.\n\nI have spoken to Yair Amir from the Spread camp on \nseveral occasions, and they are very excited about the\nreplication project. I sure it won't be an issue, but \nI will forward this message to him.\n\nDarren\n\n\n", "msg_date": "Fri, 13 Dec 2002 15:42:45 -0500", "msg_from": "<darren@up.hrcoxmail.com>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "darren@up.hrcoxmail.com wrote:\n> > > Note that while Spread is open source in the sense that \"the source is\n> > > available\", it's license is significantly more restrictive than\n> > > PostgreSQL's:\n> > > \n> > > http://www.spread.org/license/\n> > > \n> > \n> > Interesting. It looks like a modified version of the old BSD license\n> > where you are required to mention you are using Spread. I believe we\n> > can get that reduced. (I think Darren already addressed this with\n> > them.) We certainly are not going to accept software that requires all\n> > PostgreSQL user sites to mention Spread.\n> > \n> \n> I dont think this is the case. We don't redistribute spread\n> from the pg-replication site. There are links to the down\n> load area. I don't think this should be any different if\n> postgres-r is merged with the main postgresql tree. If \n> Spread is the group communication we choose to use for \n> postgresql replication, then I would think some Spread \n> information would be in order on the advocacy site, and \n> in any set up documentation for replication.\n\nYes, the question is whether we will ship the spread code inside our\ntarball? I doubt we would ever have replication running by default, but\nwe may want to ship a binary that was replication-enabled. I am\nespecially thinking of commerical vendors. Can you imagine Red Hat DB\nbeing required to mention Spread on their web page? I don't think that\nwill fly.\n\nOf course we will want to mention Spread on our web site and in our\ndocumentation, but we don't want to be forced to, and we don't want that\nburden to \"spread\" out to other users.\n\n> I have spoken to Yair Amir from the Spread camp on \n> several occasions, and they are very excited about the\n> replication project. I sure it won't be an issue, but \n> I will forward this message to him.\n\nGood.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 15:51:52 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" } ]
[ { "msg_contents": "On RedHat 8.0 (upgraded from Red Hat 7.2 with PostgreSQL 7.2.3), \"make\ncheck\" gives the following error:\n\n\"/home/postgres/postgresql-7.3/src/test/regress/./tmp_check/install//usr/loc\nal/pgsql/bin/pg_encoding: relocation error:\n/home/postgres/postgresql-7.3/src/test/regress/./tmp_check/install//usr/loca\nl/pgsql/bin/pg_encoding: undefined symbol: pg_char_to_encoding\ncreatedb: \"SQL_ASCII\" is not a valid encoding name\npg_regress: createdb failed\"\n\n./configure line was:\nCFLAGS='-O2 -pipe' ./configure --enable-integer-datetimes --with-perl\n--with-python --with-tcl --enable-syslog\n\nOn a different machine, I have successfully compiled and installed running\nRedHat 8.0 (but without an existing PostgreSQL installation) and passed all\ntests.\n\nAny help with this problem would be much appreciated.\n\nThanks,\n\tMurthy\n\n-----Original Message-----\nFrom: Bruce Momjian [mailto:pgman@candle.pha.pa.us]\nSent: Monday, December 09, 2002 20:09\nTo: Shibashish\nCc: pgsql-admin@postgresql.org; pgsql-bugs@postgresql.org;\npgsql-hackers@postgresql.org\nSubject: Re: [ADMIN] PostgreSQL 7.3 Installation on SCO\n\n\n\nIt should have worked, but edit Makefile.shlib and remove that offending\nexport from the link line. That may fix it.\n\n---------------------------------------------------------------------------\n\nShibashish wrote:\n> Dear Sir,\n> \n> I use SCO Open Server 5.0.5 on an intel box. Although I have installed and\n> used PostgreSQL on Linux, setting it on SCO has not been successful :$\n> I have downloaded the latest version ie \"Postgresql-7.3\" I have also tried\n> installing postgresql-7.1 and postgresql-7.2.3, but never succeeded.\n> \n> I also installed \"ant\" package for using java. I have \"tcl8.0,\" \"tk8.0,\"\n> \"itclsh3.0\" and \"itkwish3.0\" installed in my system.\n> \"gcc version 2.7.2.1\"\n> \"Java 2 SDK, Standard Edition, v. 1.2.1\"\n> \"GNU Make 3.80\"\n> \n> My \"configure\" command was as following ...\n> ./configure --prefix=/data/pgsql --with-perl --with-tcl\n> --with-tclconfig=/data/tcl/lib/ --with-tkconfig=/data/tk/lib/\n> --with-java --without-readline --without-zlib\n> \n> The output has been attached as file \"configoutput.txt\"\n> \n> Then i give the \"make\" command. The compiling stops on an error and\n> exits after some time. The output has been attached as file\n> \"makeoutput.txt\"\n> \n> I'd be thankful to you if you can help me out sort the problem. I got your\n> mail-ids from the net and came to know that you are working on the similar\n> lines.\n> \n> Also, in the file doc/FAQ_SCO, does the section \"Compiling PostgreSQL 7.1\n> with the UDK\" apply to release 7.3 of Postgresql ? I am not using UDK.\n> \n> Waiting for a quick response from your end. kindly inform me if you have\n> already solved the problem, or whether any patch is available. Any\n> documentation or url will be highly helpful.\n> \n> Thanking You in anticipation.\n> \n> with regards\n> Shibashish\n> \n> shiba@it.iitb.ac.in\n> On Yahoo Messenger : shib_leo\n> \n> Software Engineer\n> IIT Bombay, India.\n\nContent-Description: \n\n[ Attachment, skipping... ]\n\nContent-Description: \n\n[ Attachment, skipping... ]\n\nContent-Description: \n\n[ Attachment, skipping... ]\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n\n---------------------------(end of broadcast)---------------------------\nTIP 6: Have you searched our list archives?\n\nhttp://archives.postgresql.org\n", "msg_date": "Fri, 13 Dec 2002 15:57:42 -0500", "msg_from": "Murthy Kambhampaty <murthy.kambhampaty@goeci.com>", "msg_from_op": true, "msg_subject": "PostgreSQL 7.3 installation on RedHat 8.0 fails" }, { "msg_contents": "Murthy Kambhampaty <murthy.kambhampaty@goeci.com> writes:\n> \"/home/postgres/postgresql-7.3/src/test/regress/./tmp_check/install//usr/loc\n> al/pgsql/bin/pg_encoding: relocation error:\n> /home/postgres/postgresql-7.3/src/test/regress/./tmp_check/install//usr/loca\n> l/pgsql/bin/pg_encoding: undefined symbol: pg_char_to_encoding\n\nThis usually means that the dynamic linker is picking up a version of\nlibpq.so that doesn't have MULTIBYTE support, but the program you are\ntrying to run requires MULTIBYTE. Fix: update your ldconfig settings\nso that the thing finds the version of libpq.so you just built, rather\nthan whatever old version it's finding.\n\npg_regress tries to get the system to notice the libpq.so that it\ninstalled in the tmp_check/install area, but I believe it's quite a\nways short of making that work correctly everywhere.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 13 Dec 2002 19:24:43 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: PostgreSQL 7.3 installation on RedHat 8.0 fails " } ]
[ { "msg_contents": "> It is asynchronous without the need of 2 phase commit. It is group\n> communication based and requires the group communication system to\n> guarantee total order. The tricky part is, that the local transaction\n> must be on hold until the own commit message comes back without a prior\n\nNo, It holds until it's own Writeset comes back. Commits \nand then send a commit message on the simple channel, so\ncommits don't wait for ordered writesets. \n\nRemember total order guarantees if no changes in front of \nthe local changes conflict, the local changes can commit.\n\n\n> lock conflict by a replication transaction. If such a lock confict\n> occurs, the replication transaction wins and the local transaction rolls\n> back.\n\nCorrect.\n\n> \n> The last time i was playing with spread (that was at Great Bridge in\n> Norfolk), it was IMHO useless (for Postgres-R) because it sometimes\n> dropped messages when the network load got too high. This occured\n> without any indication, no error, nothing. This is not exactly what I\n> understand as total order. I hope they have made some substantial\n> progress on that.\n> \n\nI remember the TCL tester you set up, and having problems,\nbut I don't recall investigating what the problems were.\nIf you still have the code I can try and reproduce the \nproblem, and investigate it on the spread list. \n\nDarren\n\n", "msg_date": "Fri, 13 Dec 2002 16:12:37 -0500", "msg_from": "<darren@up.hrcoxmail.com>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "darren@up.hrcoxmail.com wrote:\n> > It is asynchronous without the need of 2 phase commit. It is group\n> > communication based and requires the group communication system to\n> > guarantee total order. The tricky part is, that the local transaction\n> > must be on hold until the own commit message comes back without a prior\n> \n> No, It holds until it's own Writeset comes back. Commits \n> and then send a commit message on the simple channel, so\n> commits don't wait for ordered writesets. \n\nDarren, can you clarify this? Why does it send that message? How does\nit allow commits not to wait for ordered writesets?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 16:19:32 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "darren@up.hrcoxmail.com wrote:\n> \n> > It is asynchronous without the need of 2 phase commit. It is group\n> > communication based and requires the group communication system to\n> > guarantee total order. The tricky part is, that the local transaction\n> > must be on hold until the own commit message comes back without a prior\n> \n> No, It holds until it's own Writeset comes back. Commits\n> and then send a commit message on the simple channel, so\n> commits don't wait for ordered writesets.\n> \n> Remember total order guarantees if no changes in front of\n> the local changes conflict, the local changes can commit.\n\nRight, it's the writeset ... that get's sent just before you flip bits\nin the clog, then wait until it comes back and flip 'em. \n\n> >\n> > The last time i was playing with spread (that was at Great Bridge in\n> > Norfolk), it was IMHO useless (for Postgres-R) because it sometimes\n> > dropped messages when the network load got too high. This occured\n> > without any indication, no error, nothing. This is not exactly what I\n> > understand as total order. I hope they have made some substantial\n> > progress on that.\n> >\n> \n> I remember the TCL tester you set up, and having problems,\n> but I don't recall investigating what the problems were.\n> If you still have the code I can try and reproduce the\n> problem, and investigate it on the spread list.\n\nMaybe you heard about it, there was this funny conversation while\nwalking down the hallway:\n\n\"Did that German guy ever turn in his notebook?\"\n\n\"You mean THIS German guy?\"\n\n\"Yes, did he turn it in?\"\n\n\"He is here. Right BEHIND YOU!!!\"\n\n\"Hummmpf ... er ...\"\n\n\nThe stuff was on that notebook. Sorry.\n\n\nJan\n\n-- \n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== JanWieck@Yahoo.com #\n", "msg_date": "Fri, 13 Dec 2002 16:31:51 -0500", "msg_from": "Jan Wieck <JanWieck@Yahoo.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "darren@up.hrcoxmail.com wrote:\n>>It is asynchronous without the need of 2 phase commit. It is group\n>>communication based and requires the group communication system to\n>>guarantee total order. The tricky part is, that the local transaction\n>>must be on hold until the own commit message comes back without a prior\n> \n> \n> No, It holds until it's own Writeset comes back. Commits \n> and then send a commit message on the simple channel, so\n> commits don't wait for ordered writesets. \n> \n> Remember total order guarantees if no changes in front of \n> the local changes conflict, the local changes can commit.\n\nDo people have to be careful about how they use sequences, as they don't normally roll back?\n\nRegards and best wishes,\n\nJustin Clift\n\n<snip>\n> Darren\n\n-- \n\"My grandfather once told me that there are two kinds of people: those\nwho work and those who take the credit. He told me to try to be in the\nfirst group; there was less competition there.\"\n- Indira Gandhi\n\n", "msg_date": "Sat, 14 Dec 2002 17:55:05 +1100", "msg_from": "Justin Clift <justin@postgresql.org>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" } ]
[ { "msg_contents": "> > It is asynchronous without the need of 2 phase commit. It is group\n> \n> Well, Darren's PDF at:\n> \n> \tftp://gborg.postgresql.org/pub/pgreplication/stable/PostgreSQLReplication.pdf.gz\n> \n> calls Postgres-R \"Type: Embedded, Peer-to-Peer, Sync\". I don't know\n> enough about replication so I will let you fight it out with him. ;-)\n> \n\n\nIf were still defining synchronous as pre commit, then\npostgres-r is synchronous.\n\nDarren\n\n", "msg_date": "Fri, 13 Dec 2002 16:18:01 -0500", "msg_from": "<darren@up.hrcoxmail.com>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" } ]
[ { "msg_contents": "> \n> \n> Darren, can you clarify this? Why does it send that message? How does\n> it allow commits not to wait for ordered writesets?\n> \n\nThere are two channels. One for total order writesets \n(changes to the DB). The other is simple order for\naborts, commits, joins (systems joining the replica), etc.\nThe simple channel is necessary, because we don't want to\nwait for total ordered changes to get an abort message and\nso forth. In some cases you might get an abort or a commit\nmessage before you get the writeset it refers to.\n\nLets say we have systems A, B and C. Each one has some\nchanges and sends a writeset to the group communication\nsystem (GSC). The total order dictates WS(A), WS(B), and\nWS(C) and the writes sets are recieved in that order at\neach system. Now C gets WS(A) no conflict, gets WS(B) no\nconflict, and receives WS(C). Now C can commit WS(C) even \nbefore the commit messages C(A) or C(B), because there is no\nconflict. \n\nHope that helps,\n\nDarren\n\n", "msg_date": "Fri, 13 Dec 2002 16:33:34 -0500", "msg_from": "<darren@up.hrcoxmail.com>", "msg_from_op": true, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "darren@up.hrcoxmail.com wrote:\n> \n> >\n> >\n> > Darren, can you clarify this? Why does it send that message? How does\n> > it allow commits not to wait for ordered writesets?\n> >\n> \n> There are two channels. One for total order writesets\n> (changes to the DB). The other is simple order for\n> aborts, commits, joins (systems joining the replica), etc.\n> The simple channel is necessary, because we don't want to\n> wait for total ordered changes to get an abort message and\n> so forth. In some cases you might get an abort or a commit\n> message before you get the writeset it refers to.\n> \n> Lets say we have systems A, B and C. Each one has some\n> changes and sends a writeset to the group communication\n> system (GSC). The total order dictates WS(A), WS(B), and\n> WS(C) and the writes sets are recieved in that order at\n> each system. Now C gets WS(A) no conflict, gets WS(B) no\n> conflict, and receives WS(C). Now C can commit WS(C) even\n> before the commit messages C(A) or C(B), because there is no\n> conflict.\n\nAnd that is IMHO not synchronous. C does not have to wait for A and B to\nfinish the same tasks. If now at this very moment two new transactions\nquery system A and system C (assuming A has not yet committed WS(C)\nwhile C has), they will get different data back (thanks to non-blocking\nreads). I think this is pretty asynchronous. \n\nIt doesn't lead to inconsistencies, because the transaction on A cannot\ndo something that is in conflict with the changes made by WS(C), since\nit's WS(A)2 will come back after WS(C) arrived at A and thus WS(C)\narriving at A will cause WS(A)2 to rollback (WS used synonymous to Xact\nin this context).\n\n> \n> Hope that helps,\n> \n> Darren\n\nHope this doesn't add too much confusion :-)\n\nJan\n\n-- \n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== JanWieck@Yahoo.com #\n", "msg_date": "Fri, 13 Dec 2002 16:53:52 -0500", "msg_from": "Jan Wieck <JanWieck@Yahoo.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "darren@up.hrcoxmail.com wrote:\n> > \n> > \n> > Darren, can you clarify this? Why does it send that message? How does\n> > it allow commits not to wait for ordered writesets?\n> > \n> \n> There are two channels. One for total order writesets \n> (changes to the DB). The other is simple order for\n> aborts, commits, joins (systems joining the replica), etc.\n> The simple channel is necessary, because we don't want to\n> wait for total ordered changes to get an abort message and\n> so forth. In some cases you might get an abort or a commit\n> message before you get the writeset it refers to.\n> \n> Lets say we have systems A, B and C. Each one has some\n> changes and sends a writeset to the group communication\n> system (GSC). The total order dictates WS(A), WS(B), and\n> WS(C) and the writes sets are recieved in that order at\n> each system. Now C gets WS(A) no conflict, gets WS(B) no\n> conflict, and receives WS(C). Now C can commit WS(C) even \n> before the commit messages C(A) or C(B), because there is no\n> conflict. \n\nOh, so C doesn't apply A's changes until it see A's commit, but it can\ncontinue with its own changes because there is no conflict?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Fri, 13 Dec 2002 17:11:22 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": ">\n>\n>>\n>>Lets say we have systems A, B and C. Each one has some\n>>changes and sends a writeset to the group communication\n>>system (GSC). The total order dictates WS(A), WS(B), and\n>>WS(C) and the writes sets are recieved in that order at\n>>each system. Now C gets WS(A) no conflict, gets WS(B) no\n>>conflict, and receives WS(C). Now C can commit WS(C) even\n>>before the commit messages C(A) or C(B), because there is no\n>>conflict.\n>>\n>\n>And that is IMHO not synchronous. C does not have to wait for A and B to\n>finish the same tasks. If now at this very moment two new transactions\n>query system A and system C (assuming A has not yet committed WS(C)\n>while C has), they will get different data back (thanks to non-blocking\n>reads). I think this is pretty asynchronous. \n>\n\nSo if we hold WS(C) until we receive commit messages for WS(A) and \nWS(B), will that meet\nyour synchronous expectations, or do all the systems need to commit the \nWS in the same order\nand at the same exact time.\n\n>\n>\n>It doesn't lead to inconsistencies, because the transaction on A cannot\n>do something that is in conflict with the changes made by WS(C), since\n>it's WS(A)2 will come back after WS(C) arrived at A and thus WS(C)\n>arriving at A will cause WS(A)2 to rollback (WS used synonymous to Xact\n>in this context).\n>\nRight\n\n>\n>Hope this doesn't add too much confusion :-)\n>\nNo, however I guess I need to adjust my slides to include your \ndefinition of synchronous\nreplication. ;-)\n\nDarren\n\n>\n\n\n", "msg_date": "Fri, 13 Dec 2002 20:28:24 -0500", "msg_from": "Darren Johnson <darren@up.hrcoxmail.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items" }, { "msg_contents": "For live replication could I propose that we consider the systems A,B, and C\nconnected to each other independantly (i.e. A has links to B and C, B has\nlinks to A and C, and C has links to A and B), and that replication is\nhandled by the node receiving the write based transaction.\n\nIf we consider a write transaction that arrives at A (called WT(A)), system\nA will then send WT(A) to systems B and C via it's direct connections.\nSystem A will receive back either an OK response if there are not conflicts,\na NOT_OK response if there are conflicts, or no response if the system is\nunavailable.\n\nIf system A receives a NOT_OK response from any other node it begins the\nprocess of rolling back the transaction from all nodes which previously\nissued an OK, and the transaction returns a failure code to the client which\nsubmitted WT(A). The other systems (B and C) would track recent transactions\nand there would be a specified timeout after which the transaction is\nconsidered safe and could not be rolled out.\n\nAny system not returning an OK or NOT_OK state is assumed to be down, and\nerror messages are logged to state that the transaction could not be sent to\nthe system due it it's unavailablility, and any monitoring system would\nalter the administrator that a replicant is faulty.\n\nThere would also need to be code developed to ensure that a system could be\nbrought into sync with the current state of other systems within the group\nin order to allow new databases to be added, and faulty databases to be\nre-entered to the group. This code could also be used for non-realtime\nreplication to allow databases to be syncronised with the live master.\n\nThis would give a multi-master solution whereby a write transaction to any\none node would guarentee that all available replicants would also hold the\ndata once it is completed, and would also provide the code to handle\nscenarios where non-realtime data replication is required.\n\nThis system assumes that a majority of transactions will be sucessful (which\nshould be the case for a well designed system).\n\nComments?\n\nAl.\n\n\n\n\n\n\n----- Original Message -----\nFrom: \"Darren Johnson\" <darren@up.hrcoxmail.com>\nTo: \"Jan Wieck\" <JanWieck@Yahoo.com>\nCc: \"Bruce Momjian\" <pgman@candle.pha.pa.us>;\n<shridhar_daithankar@persistent.co.in>; \"PostgreSQL-development\"\n<pgsql-hackers@postgresql.org>\nSent: Saturday, December 14, 2002 1:28 AM\nSubject: [mail] Re: [HACKERS] Big 7.4 items\n\n\n> >\n> >\n> >>\n> >>Lets say we have systems A, B and C. Each one has some\n> >>changes and sends a writeset to the group communication\n> >>system (GSC). The total order dictates WS(A), WS(B), and\n> >>WS(C) and the writes sets are recieved in that order at\n> >>each system. Now C gets WS(A) no conflict, gets WS(B) no\n> >>conflict, and receives WS(C). Now C can commit WS(C) even\n> >>before the commit messages C(A) or C(B), because there is no\n> >>conflict.\n> >>\n> >\n> >And that is IMHO not synchronous. C does not have to wait for A and B to\n> >finish the same tasks. If now at this very moment two new transactions\n> >query system A and system C (assuming A has not yet committed WS(C)\n> >while C has), they will get different data back (thanks to non-blocking\n> >reads). I think this is pretty asynchronous.\n> >\n>\n> So if we hold WS(C) until we receive commit messages for WS(A) and\n> WS(B), will that meet\n> your synchronous expectations, or do all the systems need to commit the\n> WS in the same order\n> and at the same exact time.\n>\n> >\n> >\n> >It doesn't lead to inconsistencies, because the transaction on A cannot\n> >do something that is in conflict with the changes made by WS(C), since\n> >it's WS(A)2 will come back after WS(C) arrived at A and thus WS(C)\n> >arriving at A will cause WS(A)2 to rollback (WS used synonymous to Xact\n> >in this context).\n> >\n> Right\n>\n> >\n> >Hope this doesn't add too much confusion :-)\n> >\n> No, however I guess I need to adjust my slides to include your\n> definition of synchronous\n> replication. ;-)\n>\n> Darren\n>\n> >\n>\n>\n>\n> ---------------------------(end of broadcast)---------------------------\n> TIP 6: Have you searched our list archives?\n>\n> http://archives.postgresql.org\n>\n\n\n", "msg_date": "Sat, 14 Dec 2002 12:51:13 -0000", "msg_from": "\"Al Sutton\" <al@alsutton.com>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items - Replication" }, { "msg_contents": "\nThis sounds like two-phase commit. While it will work, it is probably\nslower than Postgres-R's method.\n\n---------------------------------------------------------------------------\n\nAl Sutton wrote:\n> For live replication could I propose that we consider the systems A,B, and C\n> connected to each other independantly (i.e. A has links to B and C, B has\n> links to A and C, and C has links to A and B), and that replication is\n> handled by the node receiving the write based transaction.\n> \n> If we consider a write transaction that arrives at A (called WT(A)), system\n> A will then send WT(A) to systems B and C via it's direct connections.\n> System A will receive back either an OK response if there are not conflicts,\n> a NOT_OK response if there are conflicts, or no response if the system is\n> unavailable.\n> \n> If system A receives a NOT_OK response from any other node it begins the\n> process of rolling back the transaction from all nodes which previously\n> issued an OK, and the transaction returns a failure code to the client which\n> submitted WT(A). The other systems (B and C) would track recent transactions\n> and there would be a specified timeout after which the transaction is\n> considered safe and could not be rolled out.\n> \n> Any system not returning an OK or NOT_OK state is assumed to be down, and\n> error messages are logged to state that the transaction could not be sent to\n> the system due it it's unavailablility, and any monitoring system would\n> alter the administrator that a replicant is faulty.\n> \n> There would also need to be code developed to ensure that a system could be\n> brought into sync with the current state of other systems within the group\n> in order to allow new databases to be added, and faulty databases to be\n> re-entered to the group. This code could also be used for non-realtime\n> replication to allow databases to be syncronised with the live master.\n> \n> This would give a multi-master solution whereby a write transaction to any\n> one node would guarentee that all available replicants would also hold the\n> data once it is completed, and would also provide the code to handle\n> scenarios where non-realtime data replication is required.\n> \n> This system assumes that a majority of transactions will be sucessful (which\n> should be the case for a well designed system).\n> \n> Comments?\n> \n> Al.\n> \n> \n> \n> \n> \n> \n> ----- Original Message -----\n> From: \"Darren Johnson\" <darren@up.hrcoxmail.com>\n> To: \"Jan Wieck\" <JanWieck@Yahoo.com>\n> Cc: \"Bruce Momjian\" <pgman@candle.pha.pa.us>;\n> <shridhar_daithankar@persistent.co.in>; \"PostgreSQL-development\"\n> <pgsql-hackers@postgresql.org>\n> Sent: Saturday, December 14, 2002 1:28 AM\n> Subject: [mail] Re: [HACKERS] Big 7.4 items\n> \n> \n> > >\n> > >\n> > >>\n> > >>Lets say we have systems A, B and C. Each one has some\n> > >>changes and sends a writeset to the group communication\n> > >>system (GSC). The total order dictates WS(A), WS(B), and\n> > >>WS(C) and the writes sets are recieved in that order at\n> > >>each system. Now C gets WS(A) no conflict, gets WS(B) no\n> > >>conflict, and receives WS(C). Now C can commit WS(C) even\n> > >>before the commit messages C(A) or C(B), because there is no\n> > >>conflict.\n> > >>\n> > >\n> > >And that is IMHO not synchronous. C does not have to wait for A and B to\n> > >finish the same tasks. If now at this very moment two new transactions\n> > >query system A and system C (assuming A has not yet committed WS(C)\n> > >while C has), they will get different data back (thanks to non-blocking\n> > >reads). I think this is pretty asynchronous.\n> > >\n> >\n> > So if we hold WS(C) until we receive commit messages for WS(A) and\n> > WS(B), will that meet\n> > your synchronous expectations, or do all the systems need to commit the\n> > WS in the same order\n> > and at the same exact time.\n> >\n> > >\n> > >\n> > >It doesn't lead to inconsistencies, because the transaction on A cannot\n> > >do something that is in conflict with the changes made by WS(C), since\n> > >it's WS(A)2 will come back after WS(C) arrived at A and thus WS(C)\n> > >arriving at A will cause WS(A)2 to rollback (WS used synonymous to Xact\n> > >in this context).\n> > >\n> > Right\n> >\n> > >\n> > >Hope this doesn't add too much confusion :-)\n> > >\n> > No, however I guess I need to adjust my slides to include your\n> > definition of synchronous\n> > replication. ;-)\n> >\n> > Darren\n> >\n> > >\n> >\n> >\n> >\n> > ---------------------------(end of broadcast)---------------------------\n> > TIP 6: Have you searched our list archives?\n> >\n> > http://archives.postgresql.org\n> >\n> \n> \n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 6: Have you searched our list archives?\n> \n> http://archives.postgresql.org\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Sat, 14 Dec 2002 11:59:28 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items - Replication" }, { "msg_contents": "\n\n--En cette belle journée de samedi 14 décembre 2002 11:59 -0500,\n-- Bruce Momjian écrivait avec ses petits doigts :\n> \n> This sounds like two-phase commit. While it will work, it is probably\n> slower than Postgres-R's method.\n\nWhat exactly is Postgres-R's method ?\n\n-- \nMathieu Arnold\n", "msg_date": "Sat, 14 Dec 2002 18:03:18 +0100", "msg_from": "Mathieu Arnold <mat@mat.cc>", "msg_from_op": false, "msg_subject": "Re: Big 7.4 items - Replication" }, { "msg_contents": "I see it as very difficult to avoid a two stage process because there will\nbe the following two parts to any transaction;\n\n1) All databases must agree upon the acceptability of a transaction before\nthe client can be informed of it's success. 2) All databases must be\ninformed as to whether or not the transaction was accepted by the entire\nreplicant set, and thus whether it should be written to the database.\n\nIf stage1 is missed then the client application may be informed of a\nsucessful transaction which may fail when it is replicated to other\ndatabases.\n\nIf stage 2 is missed then databases may become out of sync because they have\naccepted transactions that were rejected by other databases.\n\n From reading the PDF on Postgres-R I can see that either one of two things\nwill occur;\n\na) There will be a central point of synchronization where conflicts will be\ntested and delt with. This is not desirable because it will leave the\nsynchronization and replication processing load concentrated in one place\nwhich will limit scaleability as well as leaving a single point of failure.\n\nor\n\nb) The Group Communication blob will consist of a number of processes which\nneed to talk to all of the others to interrogate them for changes which may\nconflict with the current write that being handled and then issue the\ntransaction response. This is basically the two phase commit solution with\nphases moved into the group communication process.\n\nI can see the possibility of using solution b and having less group\ncommunication processes than databases as attempt to simplify things, but\nthis would mean the loss of a number of databases if the machine running the\ngroup communication process for the set of databases is lost.\n\nAl.\n\n----- Original Message -----\nFrom: \"Bruce Momjian\" <pgman@candle.pha.pa.us>\nTo: \"Al Sutton\" <al@alsutton.com>\nCc: \"Darren Johnson\" <darren@up.hrcoxmail.com>; \"Jan Wieck\"\n<JanWieck@Yahoo.com>; <shridhar_daithankar@persistent.co.in>;\n\"PostgreSQL-development\" <pgsql-hackers@postgresql.org>\nSent: Saturday, December 14, 2002 4:59 PM\nSubject: [mail] Re: [HACKERS] Big 7.4 items - Replication\n\n\n>\n> This sounds like two-phase commit. While it will work, it is probably\n> slower than Postgres-R's method.\n>\n> --------------------------------------------------------------------------\n-\n>\n> Al Sutton wrote:\n> > For live replication could I propose that we consider the systems A,B,\nand C\n> > connected to each other independantly (i.e. A has links to B and C, B\nhas\n> > links to A and C, and C has links to A and B), and that replication is\n> > handled by the node receiving the write based transaction.\n> >\n> > If we consider a write transaction that arrives at A (called WT(A)),\nsystem\n> > A will then send WT(A) to systems B and C via it's direct connections.\n> > System A will receive back either an OK response if there are not\nconflicts,\n> > a NOT_OK response if there are conflicts, or no response if the system\nis\n> > unavailable.\n> >\n> > If system A receives a NOT_OK response from any other node it begins the\n> > process of rolling back the transaction from all nodes which previously\n> > issued an OK, and the transaction returns a failure code to the client\nwhich\n> > submitted WT(A). The other systems (B and C) would track recent\ntransactions\n> > and there would be a specified timeout after which the transaction is\n> > considered safe and could not be rolled out.\n> >\n> > Any system not returning an OK or NOT_OK state is assumed to be down,\nand\n> > error messages are logged to state that the transaction could not be\nsent to\n> > the system due it it's unavailablility, and any monitoring system would\n> > alter the administrator that a replicant is faulty.\n> >\n> > There would also need to be code developed to ensure that a system could\nbe\n> > brought into sync with the current state of other systems within the\ngroup\n> > in order to allow new databases to be added, and faulty databases to be\n> > re-entered to the group. This code could also be used for non-realtime\n> > replication to allow databases to be syncronised with the live master.\n> >\n> > This would give a multi-master solution whereby a write transaction to\nany\n> > one node would guarentee that all available replicants would also hold\nthe\n> > data once it is completed, and would also provide the code to handle\n> > scenarios where non-realtime data replication is required.\n> >\n> > This system assumes that a majority of transactions will be sucessful\n(which\n> > should be the case for a well designed system).\n> >\n> > Comments?\n> >\n> > Al.\n> >\n> >\n> >\n> >\n> >\n> >\n> > ----- Original Message -----\n> > From: \"Darren Johnson\" <darren@up.hrcoxmail.com>\n> > To: \"Jan Wieck\" <JanWieck@Yahoo.com>\n> > Cc: \"Bruce Momjian\" <pgman@candle.pha.pa.us>;\n> > <shridhar_daithankar@persistent.co.in>; \"PostgreSQL-development\"\n> > <pgsql-hackers@postgresql.org>\n> > Sent: Saturday, December 14, 2002 1:28 AM\n> > Subject: [mail] Re: [HACKERS] Big 7.4 items\n> >\n> >\n> > > >\n> > > >\n> > > >>\n> > > >>Lets say we have systems A, B and C. Each one has some\n> > > >>changes and sends a writeset to the group communication\n> > > >>system (GSC). The total order dictates WS(A), WS(B), and\n> > > >>WS(C) and the writes sets are recieved in that order at\n> > > >>each system. Now C gets WS(A) no conflict, gets WS(B) no\n> > > >>conflict, and receives WS(C). Now C can commit WS(C) even\n> > > >>before the commit messages C(A) or C(B), because there is no\n> > > >>conflict.\n> > > >>\n> > > >\n> > > >And that is IMHO not synchronous. C does not have to wait for A and B\nto\n> > > >finish the same tasks. If now at this very moment two new\ntransactions\n> > > >query system A and system C (assuming A has not yet committed WS(C)\n> > > >while C has), they will get different data back (thanks to\nnon-blocking\n> > > >reads). I think this is pretty asynchronous.\n> > > >\n> > >\n> > > So if we hold WS(C) until we receive commit messages for WS(A) and\n> > > WS(B), will that meet\n> > > your synchronous expectations, or do all the systems need to commit\nthe\n> > > WS in the same order\n> > > and at the same exact time.\n> > >\n> > > >\n> > > >\n> > > >It doesn't lead to inconsistencies, because the transaction on A\ncannot\n> > > >do something that is in conflict with the changes made by WS(C),\nsince\n> > > >it's WS(A)2 will come back after WS(C) arrived at A and thus WS(C)\n> > > >arriving at A will cause WS(A)2 to rollback (WS used synonymous to\nXact\n> > > >in this context).\n> > > >\n> > > Right\n> > >\n> > > >\n> > > >Hope this doesn't add too much confusion :-)\n> > > >\n> > > No, however I guess I need to adjust my slides to include your\n> > > definition of synchronous\n> > > replication. ;-)\n> > >\n> > > Darren\n> > >\n> > > >\n> > >\n> > >\n> > >\n> > > ---------------------------(end of\nbroadcast)---------------------------\n> > > TIP 6: Have you searched our list archives?\n> > >\n> > > http://archives.postgresql.org\n> > >\n> >\n> >\n> >\n> > ---------------------------(end of broadcast)---------------------------\n> > TIP 6: Have you searched our list archives?\n> >\n> > http://archives.postgresql.org\n> >\n>\n> --\n> Bruce Momjian | http://candle.pha.pa.us\n> pgman@candle.pha.pa.us | (610) 359-1001\n> + If your life is a hard drive, | 13 Roberts Road\n> + Christ can be your backup. | Newtown Square, Pennsylvania\n19073\n>\n\n\n", "msg_date": "Sat, 14 Dec 2002 18:18:10 -0000", "msg_from": "\"Al Sutton\" <al@alsutton.com>", "msg_from_op": false, "msg_subject": "Re: [mail] Re: Big 7.4 items - Replication" }, { "msg_contents": ">\n>\n>\n>b) The Group Communication blob will consist of a number of processes which\n>need to talk to all of the others to interrogate them for changes which may\n>conflict with the current write that being handled and then issue the\n>transaction response. This is basically the two phase commit solution with\n>phases moved into the group communication process.\n>\n>I can see the possibility of using solution b and having less group\n>communication processes than databases as attempt to simplify things, but\n>this would mean the loss of a number of databases if the machine running the\n>group communication process for the set of databases is lost.\n>\nThe group communication system doesn't just run on one system. For \npostgres-r using spread\nthere is actually a spread daemon that runs on each database server. It \nhas nothing to do with\ndetecting the conflicts. Its job is to deliver messages in a total \norder for writesets or simple order\nfor commits, aborts, joins, etc. \n\nThe detection of conflicts will be done at the database level, by a \nbackend processes. The basic\nconcept is \"if all databases get the writesets (changes) in the exact \nsame order, apply them in a\nconsistent order, avoid conflicts, then one copy serialization is \nachieved. (one copy of the database\nreplicated across all databases in the replica)\n\nI hope that explains the group communication system's responsibility.\n\nDarren\n\n\n>\n\n\n", "msg_date": "Sat, 14 Dec 2002 13:48:06 -0500", "msg_from": "Darren Johnson <darren@up.hrcoxmail.com>", "msg_from_op": false, "msg_subject": "Re: [mail] Re: Big 7.4 items - Replication" }, { "msg_contents": "Many thanks for the explanation. Could you explain to me where the order or\nthe writeset for the following scenario;\n\nIf a tranasction takes 50ms to reach one database from another, for a\nspecific data element (called X), the following timeline occurs\n\nat 0ms, T1(X) is written to system A.\nat 10ms, T2(X) is written to system B.\n\nWhere T1(X) and T2(X) conflict.\n\nMy concern is that if the Group Communication Daemon (gcd) is operating on\neach database, a successful result for T1(X) will returned to the client\ntalking to database A because T2(X) has not reached it, and thus no conflict\nis known about, and a sucessful result is returned to the client submitting\nT2(X) to database B because it is not aware of T1(X). This would mean that\nthe two clients beleive bothe T1(X) and T2(X) completed succesfully, yet\nthey can not due to the conflict.\n\nThanks,\n\nAl.\n\n----- Original Message -----\nFrom: \"Darren Johnson\" <darren@up.hrcoxmail.com>\nTo: \"Al Sutton\" <al@alsutton.com>\nCc: \"Bruce Momjian\" <pgman@candle.pha.pa.us>; \"Jan Wieck\"\n<JanWieck@Yahoo.com>; <shridhar_daithankar@persistent.co.in>;\n\"PostgreSQL-development\" <pgsql-hackers@postgresql.org>\nSent: Saturday, December 14, 2002 6:48 PM\nSubject: Re: [mail] Re: [HACKERS] Big 7.4 items - Replication\n\n\n> >\n> >\n> >\n> >b) The Group Communication blob will consist of a number of processes\nwhich\n> >need to talk to all of the others to interrogate them for changes which\nmay\n> >conflict with the current write that being handled and then issue the\n> >transaction response. This is basically the two phase commit solution\nwith\n> >phases moved into the group communication process.\n> >\n> >I can see the possibility of using solution b and having less group\n> >communication processes than databases as attempt to simplify things, but\n> >this would mean the loss of a number of databases if the machine running\nthe\n> >group communication process for the set of databases is lost.\n> >\n> The group communication system doesn't just run on one system. For\n> postgres-r using spread\n> there is actually a spread daemon that runs on each database server. It\n> has nothing to do with\n> detecting the conflicts. Its job is to deliver messages in a total\n> order for writesets or simple order\n> for commits, aborts, joins, etc.\n>\n> The detection of conflicts will be done at the database level, by a\n> backend processes. The basic\n> concept is \"if all databases get the writesets (changes) in the exact\n> same order, apply them in a\n> consistent order, avoid conflicts, then one copy serialization is\n> achieved. (one copy of the database\n> replicated across all databases in the replica)\n>\n> I hope that explains the group communication system's responsibility.\n>\n> Darren\n>\n>\n> >\n>\n>\n>\n> ---------------------------(end of broadcast)---------------------------\n> TIP 5: Have you checked our extensive FAQ?\n>\n> http://www.postgresql.org/users-lounge/docs/faq.html\n\n\n", "msg_date": "Sun, 15 Dec 2002 10:16:22 -0000", "msg_from": "\"Al Sutton\" <al@alsutton.com>", "msg_from_op": false, "msg_subject": "Re: [mail] Re: Big 7.4 items - Replication" }, { "msg_contents": "Another concern I have with multi-master systems is what happens if the \nnetwork splits in 2 so that 2 master systems are taking commits for 2 \nseparate sets of clients. It seems to me that to re-sync the 2 databases \nupon the network healing would be a very complex task or impossible task.\n\nOn Sunday 15 December 2002 04:16 am, Al Sutton wrote:\n> Many thanks for the explanation. Could you explain to me where the order or\n> the writeset for the following scenario;\n>\n> If a tranasction takes 50ms to reach one database from another, for a\n> specific data element (called X), the following timeline occurs\n>\n> at 0ms, T1(X) is written to system A.\n> at 10ms, T2(X) is written to system B.\n>\n> Where T1(X) and T2(X) conflict.\n>\n> My concern is that if the Group Communication Daemon (gcd) is operating on\n> each database, a successful result for T1(X) will returned to the client\n> talking to database A because T2(X) has not reached it, and thus no\n> conflict is known about, and a sucessful result is returned to the client\n> submitting T2(X) to database B because it is not aware of T1(X). This would\n> mean that the two clients beleive bothe T1(X) and T2(X) completed\n> succesfully, yet they can not due to the conflict.\n>\n> Thanks,\n>\n> Al.\n>\n> ----- Original Message -----\n> From: \"Darren Johnson\" <darren@up.hrcoxmail.com>\n> To: \"Al Sutton\" <al@alsutton.com>\n> Cc: \"Bruce Momjian\" <pgman@candle.pha.pa.us>; \"Jan Wieck\"\n> <JanWieck@Yahoo.com>; <shridhar_daithankar@persistent.co.in>;\n> \"PostgreSQL-development\" <pgsql-hackers@postgresql.org>\n> Sent: Saturday, December 14, 2002 6:48 PM\n> Subject: Re: [mail] Re: [HACKERS] Big 7.4 items - Replication\n>\n> > >b) The Group Communication blob will consist of a number of processes\n>\n> which\n>\n> > >need to talk to all of the others to interrogate them for changes which\n>\n> may\n>\n> > >conflict with the current write that being handled and then issue the\n> > >transaction response. This is basically the two phase commit solution\n>\n> with\n>\n> > >phases moved into the group communication process.\n> > >\n> > >I can see the possibility of using solution b and having less group\n> > >communication processes than databases as attempt to simplify things,\n> > > but this would mean the loss of a number of databases if the machine\n> > > running\n>\n> the\n>\n> > >group communication process for the set of databases is lost.\n> >\n> > The group communication system doesn't just run on one system. For\n> > postgres-r using spread\n> > there is actually a spread daemon that runs on each database server. It\n> > has nothing to do with\n> > detecting the conflicts. Its job is to deliver messages in a total\n> > order for writesets or simple order\n> > for commits, aborts, joins, etc.\n> >\n> > The detection of conflicts will be done at the database level, by a\n> > backend processes. The basic\n> > concept is \"if all databases get the writesets (changes) in the exact\n> > same order, apply them in a\n> > consistent order, avoid conflicts, then one copy serialization is\n> > achieved. (one copy of the database\n> > replicated across all databases in the replica)\n> >\n> > I hope that explains the group communication system's responsibility.\n> >\n> > Darren\n> >\n> >\n> >\n> >\n> >\n> >\n> > ---------------------------(end of broadcast)---------------------------\n> > TIP 5: Have you checked our extensive FAQ?\n> >\n> > http://www.postgresql.org/users-lounge/docs/faq.html\n>\n> ---------------------------(end of broadcast)---------------------------\n> TIP 6: Have you searched our list archives?\n>\n> http://archives.postgresql.org\n\n", "msg_date": "Sun, 15 Dec 2002 08:29:58 -0600", "msg_from": "David Walker <pgsql@grax.com>", "msg_from_op": false, "msg_subject": "Re: [MLIST] Re: [mail] Re: Big 7.4 items - Replication" }, { "msg_contents": "David,\n\nThis can be resolved by requiring that for any transaction to succeed the\nentrypoint database must receive acknowlegements from n/2 + 0.5 (rounded up\nto the nearest integer) databases where n is the total number in the\nreplicant set. The following cases are shown as an example;\n\nTotal Number of databases: 2\nNumber required to accept transaction: 2\n\nTotal Number of databases: 3\nNumber required to accept transaction: 2\n\nTotal Number of databases: 4\nNumber required to accept transaction: 3\n\nTotal Number of databases: 5\nNumber required to accept transaction: 3\n\nTotal Number of databases: 6\nNumber required to accept transaction: 4\n\nTotal Number of databases: 7\nNumber required to accept transaction: 4\n\nTotal Number of databases: 8\nNumber required to accept transaction: 5\n\nThis would prevent two replicant sub-sets forming, because it is impossible\nfor both sets to have over 50% of the databases.\n\nApplications could be able to detect when a database has dropped out of the\nreplicant set because the database could report a state of \"Unable to obtain\nmajority consesus\". This would allow applications differentiate between a\ndatabase out of the set where writing to other databases in the set could\nyield a sucessful result, and \"Unable to commit due to conflict\" where\ntrying other databases is pointless.\n\nAl\n\nExample\n----- Original Message -----\nFrom: \"David Walker\" <pgsql@grax.com>\nTo: \"Al Sutton\" <al@alsutton.com>; \"Darren Johnson\"\n<darren@up.hrcoxmail.com>\nCc: \"Bruce Momjian\" <pgman@candle.pha.pa.us>; \"Jan Wieck\"\n<JanWieck@Yahoo.com>; <shridhar_daithankar@persistent.co.in>;\n\"PostgreSQL-development\" <pgsql-hackers@postgresql.org>\nSent: Sunday, December 15, 2002 2:29 PM\nSubject: Re: [MLIST] Re: [mail] Re: [HACKERS] Big 7.4 items - Replication\n\n\n> Another concern I have with multi-master systems is what happens if the\n> network splits in 2 so that 2 master systems are taking commits for 2\n> separate sets of clients. It seems to me that to re-sync the 2 databases\n> upon the network healing would be a very complex task or impossible task.\n>\n> On Sunday 15 December 2002 04:16 am, Al Sutton wrote:\n> > Many thanks for the explanation. Could you explain to me where the order\nor\n> > the writeset for the following scenario;\n> >\n> > If a tranasction takes 50ms to reach one database from another, for a\n> > specific data element (called X), the following timeline occurs\n> >\n> > at 0ms, T1(X) is written to system A.\n> > at 10ms, T2(X) is written to system B.\n> >\n> > Where T1(X) and T2(X) conflict.\n> >\n> > My concern is that if the Group Communication Daemon (gcd) is operating\non\n> > each database, a successful result for T1(X) will returned to the\nclient\n> > talking to database A because T2(X) has not reached it, and thus no\n> > conflict is known about, and a sucessful result is returned to the\nclient\n> > submitting T2(X) to database B because it is not aware of T1(X). This\nwould\n> > mean that the two clients beleive bothe T1(X) and T2(X) completed\n> > succesfully, yet they can not due to the conflict.\n> >\n> > Thanks,\n> >\n> > Al.\n> >\n> > ----- Original Message -----\n> > From: \"Darren Johnson\" <darren@up.hrcoxmail.com>\n> > To: \"Al Sutton\" <al@alsutton.com>\n> > Cc: \"Bruce Momjian\" <pgman@candle.pha.pa.us>; \"Jan Wieck\"\n> > <JanWieck@Yahoo.com>; <shridhar_daithankar@persistent.co.in>;\n> > \"PostgreSQL-development\" <pgsql-hackers@postgresql.org>\n> > Sent: Saturday, December 14, 2002 6:48 PM\n> > Subject: Re: [mail] Re: [HACKERS] Big 7.4 items - Replication\n> >\n> > > >b) The Group Communication blob will consist of a number of processes\n> >\n> > which\n> >\n> > > >need to talk to all of the others to interrogate them for changes\nwhich\n> >\n> > may\n> >\n> > > >conflict with the current write that being handled and then issue the\n> > > >transaction response. This is basically the two phase commit solution\n> >\n> > with\n> >\n> > > >phases moved into the group communication process.\n> > > >\n> > > >I can see the possibility of using solution b and having less group\n> > > >communication processes than databases as attempt to simplify things,\n> > > > but this would mean the loss of a number of databases if the machine\n> > > > running\n> >\n> > the\n> >\n> > > >group communication process for the set of databases is lost.\n> > >\n> > > The group communication system doesn't just run on one system. For\n> > > postgres-r using spread\n> > > there is actually a spread daemon that runs on each database server.\nIt\n> > > has nothing to do with\n> > > detecting the conflicts. Its job is to deliver messages in a total\n> > > order for writesets or simple order\n> > > for commits, aborts, joins, etc.\n> > >\n> > > The detection of conflicts will be done at the database level, by a\n> > > backend processes. The basic\n> > > concept is \"if all databases get the writesets (changes) in the exact\n> > > same order, apply them in a\n> > > consistent order, avoid conflicts, then one copy serialization is\n> > > achieved. (one copy of the database\n> > > replicated across all databases in the replica)\n> > >\n> > > I hope that explains the group communication system's responsibility.\n> > >\n> > > Darren\n> > >\n> > >\n> > >\n> > >\n> > >\n> > >\n> > > ---------------------------(end of\nbroadcast)---------------------------\n> > > TIP 5: Have you checked our extensive FAQ?\n> > >\n> > > http://www.postgresql.org/users-lounge/docs/faq.html\n> >\n> > ---------------------------(end of broadcast)---------------------------\n> > TIP 6: Have you searched our list archives?\n> >\n> > http://archives.postgresql.org\n>\n>\n\n\n", "msg_date": "Sun, 15 Dec 2002 15:06:04 -0000", "msg_from": "\"Al Sutton\" <al@alsutton.com>", "msg_from_op": false, "msg_subject": "Re: [MLIST] Re: [mail] Re: Big 7.4 items - Replication" }, { "msg_contents": "Jonathan,\n\nHow do the group communication daemons on system A and B agree that T2 is\nafter T1?,\n\nAs I understand it the operation is performed locally before being passed on\nto the group for replication, when T2 arrives at system B, system B has no\nknowlege of T1 and so can perform T2 sucessfully.\n\nI am guessing that the System B performs T2 locally, sends it to the group\ncommunication daemon for ordering, and then receives it back from the group\ncommunication order queue after it's position in the order queue has been\ndecided before it is written to the database.\n\nThis would indicate to me that there is a single central point which decides\nthat T2 is after T1.\n\nIs this true?\n\nAl.\n\n----- Original Message -----\nFrom: \"Jonathan Stanton\" <jonathan@cnds.jhu.edu>\nTo: \"Al Sutton\" <al@alsutton.com>\nCc: \"Darren Johnson\" <darren@up.hrcoxmail.com>; \"Bruce Momjian\"\n<pgman@candle.pha.pa.us>; \"Jan Wieck\" <JanWieck@Yahoo.com>;\n<shridhar_daithankar@persistent.co.in>; \"PostgreSQL-development\"\n<pgsql-hackers@postgresql.org>\nSent: Sunday, December 15, 2002 5:00 PM\nSubject: Re: [mail] Re: [HACKERS] Big 7.4 items - Replication\n\n\n> The total order provided by the group communication daemons guarantees\n> that every member will see the tranactions/writesets in the same order.\n> So both A and B will see that T1 is ordered before T2 BEFORE writing\n> anything back to the client. So for both servers T1 will be completed\n> successfully, and T2 will be aborted because of conflicting writesets.\n>\n> Jonathan\n>\n> On Sun, Dec 15, 2002 at 10:16:22AM -0000, Al Sutton wrote:\n> > Many thanks for the explanation. Could you explain to me where the order\nor\n> > the writeset for the following scenario;\n> >\n> > If a tranasction takes 50ms to reach one database from another, for a\n> > specific data element (called X), the following timeline occurs\n> >\n> > at 0ms, T1(X) is written to system A.\n> > at 10ms, T2(X) is written to system B.\n> >\n> > Where T1(X) and T2(X) conflict.\n> >\n> > My concern is that if the Group Communication Daemon (gcd) is operating\non\n> > each database, a successful result for T1(X) will returned to the\nclient\n> > talking to database A because T2(X) has not reached it, and thus no\nconflict\n> > is known about, and a sucessful result is returned to the client\nsubmitting\n> > T2(X) to database B because it is not aware of T1(X). This would mean\nthat\n> > the two clients beleive bothe T1(X) and T2(X) completed succesfully, yet\n> > they can not due to the conflict.\n> >\n> > Thanks,\n> >\n> > Al.\n> >\n> > ----- Original Message -----\n> > From: \"Darren Johnson\" <darren@up.hrcoxmail.com>\n> > To: \"Al Sutton\" <al@alsutton.com>\n> > Cc: \"Bruce Momjian\" <pgman@candle.pha.pa.us>; \"Jan Wieck\"\n> > <JanWieck@Yahoo.com>; <shridhar_daithankar@persistent.co.in>;\n> > \"PostgreSQL-development\" <pgsql-hackers@postgresql.org>\n> > Sent: Saturday, December 14, 2002 6:48 PM\n> > Subject: Re: [mail] Re: [HACKERS] Big 7.4 items - Replication\n> >\n> >\n> > > >\n> > > >\n> > > >\n> > > >b) The Group Communication blob will consist of a number of processes\n> > which\n> > > >need to talk to all of the others to interrogate them for changes\nwhich\n> > may\n> > > >conflict with the current write that being handled and then issue the\n> > > >transaction response. This is basically the two phase commit solution\n> > with\n> > > >phases moved into the group communication process.\n> > > >\n> > > >I can see the possibility of using solution b and having less group\n> > > >communication processes than databases as attempt to simplify things,\nbut\n> > > >this would mean the loss of a number of databases if the machine\nrunning\n> > the\n> > > >group communication process for the set of databases is lost.\n> > > >\n> > > The group communication system doesn't just run on one system. For\n> > > postgres-r using spread\n> > > there is actually a spread daemon that runs on each database server.\nIt\n> > > has nothing to do with\n> > > detecting the conflicts. Its job is to deliver messages in a total\n> > > order for writesets or simple order\n> > > for commits, aborts, joins, etc.\n> > >\n> > > The detection of conflicts will be done at the database level, by a\n> > > backend processes. The basic\n> > > concept is \"if all databases get the writesets (changes) in the exact\n> > > same order, apply them in a\n> > > consistent order, avoid conflicts, then one copy serialization is\n> > > achieved. (one copy of the database\n> > > replicated across all databases in the replica)\n> > >\n> > > I hope that explains the group communication system's responsibility.\n> > >\n> > > Darren\n> > >\n> > >\n> > > >\n> > >\n> > >\n> > >\n> > > ---------------------------(end of\nbroadcast)---------------------------\n> > > TIP 5: Have you checked our extensive FAQ?\n> > >\n> > > http://www.postgresql.org/users-lounge/docs/faq.html\n> >\n> >\n> >\n> > ---------------------------(end of broadcast)---------------------------\n> > TIP 6: Have you searched our list archives?\n> >\n> > http://archives.postgresql.org\n>\n> --\n> -------------------------------------------------------\n> Jonathan R. Stanton jonathan@cs.jhu.edu\n> Dept. of Computer Science\n> Johns Hopkins University\n> -------------------------------------------------------\n>\n\n\n", "msg_date": "Sun, 15 Dec 2002 19:42:35 -0000", "msg_from": "\"Al Sutton\" <al@alsutton.com>", "msg_from_op": false, "msg_subject": "Re: [mail] Re: Big 7.4 items - Replication" }, { "msg_contents": "Jonathan,\n\nMany thanks for clarifying the situation some more. With token passing, I\nhave the following concerns;\n\n1) What happends if a server holding the token should die whilst it is in\nposession of the token.\n\n2) If I have n servers, and the time to pass the token between each server\nis x milliseconds, I may have to wait for upto m times x milliseconds in\norder for a transaction to be processed. If a server is limited to a single\ntransaction per posession of the token (in order to ensure no system hogs\nthe token), and the server develops a queue of length y, I will have to wait\nm times x times y for the transaction to be processed. Both scenarios I\nbeleive would not scale well beyond a small subset of servers with low\nnetwork latency between them.\n\nIf we consider the following situation I can illustrate why I'm still in\nfavour of a two phase commit;\n\nImagine, for example, credit card details about the status of an account\nreplicated in real time between databases in London, Moscow, Singapore,\nSyndey, and New York. If any server can talk to any other server with a\nguarenteed packet transfer time of 150ms a two phase commit could complete\nin 600ms as it's worst case (assuming that the two phases consist of\nrequest/response pairs, and that each server talks to all the others in\nparallel). A token passing system may have to wait for the token to pass\nthrough every other server before reaching the one that has the transaction\ncomitted to it, which could take about 750ms.\n\nIf you then expand the network to allow for a primary and disaster recover\ndatabase at each location the two phase commit still maintains it's 600ms\nresponse time, but the token passing system doubles to 1500ms.\n\nAllowing disjointed segments to continue executing is also a concern because\nany split in the replication group could effectively double the accepted\ncard limit for any card holder should they purchase items from various\nlocations around the globe.\n\nI can see an idea that the token may be passed to the system with the most\ntransactions in a wait state, but this would cause low volume databases to\nloose out on response times to higher volume ones, which is again,\nundesirable.\n\nAl.\n\n----- Original Message -----\nFrom: \"Jonathan Stanton\" <jonathan@cnds.jhu.edu>\nTo: \"Al Sutton\" <al@alsutton.com>\nCc: \"Darren Johnson\" <darren@up.hrcoxmail.com>; \"Bruce Momjian\"\n<pgman@candle.pha.pa.us>; \"Jan Wieck\" <JanWieck@Yahoo.com>;\n<shridhar_daithankar@persistent.co.in>; \"PostgreSQL-development\"\n<pgsql-hackers@postgresql.org>\nSent: Sunday, December 15, 2002 9:17 PM\nSubject: Re: [mail] Re: [HACKERS] Big 7.4 items - Replication\n\n\n> On Sun, Dec 15, 2002 at 07:42:35PM -0000, Al Sutton wrote:\n> > Jonathan,\n> >\n> > How do the group communication daemons on system A and B agree that T2\nis\n> > after T1?,\n>\n> Lets split this into two separate problems:\n>\n> 1) How do the daemons totally order a set of messages (abstract\n> messages)\n>\n> 2) How do database transactions get split into writesets that are sent\n> as messages through the group communication system.\n>\n> As to question 1, the set of daemons (usually one running on each\n> participating server) run a distributed ordering algorithm, as well as\n> distributed algorithms to provide message reliability, fault-detection,\n> and membership services. These are completely distributed algorithms, no\n> \"central\" controller node exists, so even if network partitions occur\n> the group communication system keeps running and providing ordering and\n> reliability guarantees to messages.\n>\n> A number of different algorithms exist as to how to provide a total\n> order on messages. Spread currently uses a token algorithm, that\n> involves passing a token between the daemons, and a counter attached to\n> each message, but other algorithms exist and we have implemneted some\n> other ones in our research. You can find lots of details in the papers\n> at www.cnds.jhu.edu/publications/ and www.spread.org.\n>\n> As to question 2, there are several different approaches to how to use\n> such a total order for actual database replication. They all use the gcs\n> total order to establish a single sequence of \"events\" that all the\n> databases see. Then each database can act on the events as they are\n> delivered by teh gcs and be guaranteed that no other database will see a\n> different order.\n>\n> In the postgres-R case, the action received from a client is performned\n> partially at the originating postgres server, the writesets are then\n> sent through the gcs to order them and determine conflicts. Once they\n> are delivered back, if no conflicts occured in the meantime, the\n> original transaction is completed and the result returned to the client.\n> If a conflict occured, the original transaction is rolled back and\n> aborted. and the abort is returned to the client.\n>\n> >\n> > As I understand it the operation is performed locally before being\npassed on\n> > to the group for replication, when T2 arrives at system B, system B has\nno\n> > knowlege of T1 and so can perform T2 sucessfully.\n> >\n> > I am guessing that the System B performs T2 locally, sends it to the\ngroup\n> > communication daemon for ordering, and then receives it back from the\ngroup\n> > communication order queue after it's position in the order queue has\nbeen\n> > decided before it is written to the database.\n>\n> If I understand the above correctly, yes, that is the same as I describe\n> above.\n>\n> >\n> > This would indicate to me that there is a single central point which\ndecides\n> > that T2 is after T1.\n>\n> No, there is a distributed algorithm that determins the order. The\n> distributed algorithm \"emulates\" a central controller who decides the\n> order, but no single controller actually exists.\n>\n> Jonathan\n>\n> > ----- Original Message -----\n> > From: \"Jonathan Stanton\" <jonathan@cnds.jhu.edu>\n> > To: \"Al Sutton\" <al@alsutton.com>\n> > Cc: \"Darren Johnson\" <darren@up.hrcoxmail.com>; \"Bruce Momjian\"\n> > <pgman@candle.pha.pa.us>; \"Jan Wieck\" <JanWieck@Yahoo.com>;\n> > <shridhar_daithankar@persistent.co.in>; \"PostgreSQL-development\"\n> > <pgsql-hackers@postgresql.org>\n> > Sent: Sunday, December 15, 2002 5:00 PM\n> > Subject: Re: [mail] Re: [HACKERS] Big 7.4 items - Replication\n> >\n> >\n> > > The total order provided by the group communication daemons guarantees\n> > > that every member will see the tranactions/writesets in the same\norder.\n> > > So both A and B will see that T1 is ordered before T2 BEFORE writing\n> > > anything back to the client. So for both servers T1 will be completed\n> > > successfully, and T2 will be aborted because of conflicting writesets.\n> > >\n> > > Jonathan\n> > >\n> > > On Sun, Dec 15, 2002 at 10:16:22AM -0000, Al Sutton wrote:\n> > > > Many thanks for the explanation. Could you explain to me where the\norder\n> > or\n> > > > the writeset for the following scenario;\n> > > >\n> > > > If a tranasction takes 50ms to reach one database from another, for\na\n> > > > specific data element (called X), the following timeline occurs\n> > > >\n> > > > at 0ms, T1(X) is written to system A.\n> > > > at 10ms, T2(X) is written to system B.\n> > > >\n> > > > Where T1(X) and T2(X) conflict.\n> > > >\n> > > > My concern is that if the Group Communication Daemon (gcd) is\noperating\n> > on\n> > > > each database, a successful result for T1(X) will returned to the\n> > client\n> > > > talking to database A because T2(X) has not reached it, and thus no\n> > conflict\n> > > > is known about, and a sucessful result is returned to the client\n> > submitting\n> > > > T2(X) to database B because it is not aware of T1(X). This would\nmean\n> > that\n> > > > the two clients beleive bothe T1(X) and T2(X) completed succesfully,\nyet\n> > > > they can not due to the conflict.\n> > > >\n> > > > Thanks,\n> > > >\n> > > > Al.\n> > > >\n> > > > ----- Original Message -----\n> > > > From: \"Darren Johnson\" <darren@up.hrcoxmail.com>\n> > > > To: \"Al Sutton\" <al@alsutton.com>\n> > > > Cc: \"Bruce Momjian\" <pgman@candle.pha.pa.us>; \"Jan Wieck\"\n> > > > <JanWieck@Yahoo.com>; <shridhar_daithankar@persistent.co.in>;\n> > > > \"PostgreSQL-development\" <pgsql-hackers@postgresql.org>\n> > > > Sent: Saturday, December 14, 2002 6:48 PM\n> > > > Subject: Re: [mail] Re: [HACKERS] Big 7.4 items - Replication\n> > > >\n> > > >\n> > > > > >\n> > > > > >\n> > > > > >\n> > > > > >b) The Group Communication blob will consist of a number of\nprocesses\n> > > > which\n> > > > > >need to talk to all of the others to interrogate them for changes\n> > which\n> > > > may\n> > > > > >conflict with the current write that being handled and then issue\nthe\n> > > > > >transaction response. This is basically the two phase commit\nsolution\n> > > > with\n> > > > > >phases moved into the group communication process.\n> > > > > >\n> > > > > >I can see the possibility of using solution b and having less\ngroup\n> > > > > >communication processes than databases as attempt to simplify\nthings,\n> > but\n> > > > > >this would mean the loss of a number of databases if the machine\n> > running\n> > > > the\n> > > > > >group communication process for the set of databases is lost.\n> > > > > >\n> > > > > The group communication system doesn't just run on one system.\nFor\n> > > > > postgres-r using spread\n> > > > > there is actually a spread daemon that runs on each database\nserver.\n> > It\n> > > > > has nothing to do with\n> > > > > detecting the conflicts. Its job is to deliver messages in a\ntotal\n> > > > > order for writesets or simple order\n> > > > > for commits, aborts, joins, etc.\n> > > > >\n> > > > > The detection of conflicts will be done at the database level, by\na\n> > > > > backend processes. The basic\n> > > > > concept is \"if all databases get the writesets (changes) in the\nexact\n> > > > > same order, apply them in a\n> > > > > consistent order, avoid conflicts, then one copy serialization is\n> > > > > achieved. (one copy of the database\n> > > > > replicated across all databases in the replica)\n> > > > >\n> > > > > I hope that explains the group communication system's\nresponsibility.\n> > > > >\n> > > > > Darren\n> > > > >\n> > > > >\n> > > > > >\n> > > > >\n> > > > >\n> > > > >\n> > > > > ---------------------------(end of\n> > broadcast)---------------------------\n> > > > > TIP 5: Have you checked our extensive FAQ?\n> > > > >\n> > > > > http://www.postgresql.org/users-lounge/docs/faq.html\n> > > >\n> > > >\n> > > >\n> > > > ---------------------------(end of\nbroadcast)---------------------------\n> > > > TIP 6: Have you searched our list archives?\n> > > >\n> > > > http://archives.postgresql.org\n> > >\n> > > --\n> > > -------------------------------------------------------\n> > > Jonathan R. Stanton jonathan@cs.jhu.edu\n> > > Dept. of Computer Science\n> > > Johns Hopkins University\n> > > -------------------------------------------------------\n> > >\n> >\n> >\n>\n> --\n> -------------------------------------------------------\n> Jonathan R. Stanton jonathan@cs.jhu.edu\n> Dept. of Computer Science\n> Johns Hopkins University\n> -------------------------------------------------------\n>\n\n\n", "msg_date": "Sun, 15 Dec 2002 22:15:40 -0000", "msg_from": "\"Al Sutton\" <al@alsutton.com>", "msg_from_op": false, "msg_subject": "Re: [mail] Re: Big 7.4 items - Replication" }, { "msg_contents": "Darren Johnson wrote:\n\n> The group communication system doesn't just run on one system. For\n> postgres-r using spread\n\nThe reason why group communication software is used is simply because\nthis software is designed with two goals in mind:\n\n1) optimize bandwidth usage\n\n2) make many-to-many communication easy\n\nNumber one is done by utilizing things like multicasting where\navailable.\n\nNumber two is done by using global scoped queues.\n\nI add this only to avoid reading that pushing some PITR log snippets via\nFTP or worse over a network would do the same. It did not in the past,\nit does not do right now and it will not do in the future.\n\n\nJan\n\n-- \n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== JanWieck@Yahoo.com #\n", "msg_date": "Sun, 15 Dec 2002 22:40:17 -0500", "msg_from": "Jan Wieck <JanWieck@Yahoo.com>", "msg_from_op": false, "msg_subject": "Re: [mail] Re: Big 7.4 items - Replication" } ]
[ { "msg_contents": "Thank you for a good workaround.\n\nEven BETTER would be to fix the aggregates so workarounds wouldn't have to\nbe found.\n\nThanks again,\n\nL.\nOn Fri, 13 Dec 2002, Josh Berkus wrote:\n\n> \n> \n> ---------- Forwarded Message ----------\n> \n> Subject: Re: [PERFORM] Odd Sort/Limit/Max Problem\n> Date: Fri, 13 Dec 2002 12:10:20 -0800 (PST)\n> From: Stephan Szabo <sszabo@megazone23.bigpanda.com>\n> To: Josh Berkus <josh@agliodbs.com>\n> Cc: <pgsql-performance@postgresql.org>\n> \n> On Fri, 13 Dec 2002, Josh Berkus wrote:\n> \n> > First, as expected, a regular aggregate is slow:\n> \n> > So we use the workaround standard for PostgreSQL:\n> >\n> > ... which is fast, but returns NULL, since nulls sort to the bottom! So we\n> > add IS NOT NULL:\n> >\n> > jwnet=> explain analyze select date_resolved from case_clients where\n> > date_resolved is not null order by date_resolved desc limit 1;\n> > NOTICE: QUERY PLAN:\n> >\n> > Limit (cost=0.00..4.06 rows=1 width=4) (actual time=219.63..219.64 rows=1\n> > loops=1)\n> > -> Index Scan Backward using idx_caseclients_resolved on case_clients\n> > (cost=0.00..163420.59 rows=40272 width=4) (actual time=219.62..219.62 rows=2\n> > loops=1)\n> > Total runtime: 219.76 msec\n> >\n> > Aieee! Almost as slow as the aggregate!\n> \n> I'd suggest trying a partial index on date_resolved where date_resolve is\n> not null. In my simple tests on about 200,000 rows of ints where 50% are\n> null that sort of index cut the runtime on my machine from 407.66 msec to\n> 0.15 msec.\n> \n> \n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org\n> \n> \n> -------------------------------------------------------\n> \n> \n\n-- \nLaurette Cisneros\nThe Database Group\n(510) 420-3137\nNextBus Information Systems, Inc.\nwww.nextbus.com\n----------------------------------\nThere's more to life than just SQL.\n\n", "msg_date": "Fri, 13 Dec 2002 16:00:45 -0800 (PST)", "msg_from": "Laurette Cisneros <laurette@nextbus.com>", "msg_from_op": true, "msg_subject": "Re: Fwd: Re: [PERFORM] Odd Sort/Limit/Max Problem" } ]
[ { "msg_contents": "A basic version of the SQL information schema is now available in newly\ninitdb'ed database installations. There's still a bunch of work to do to\ncreate all the views that the spec defines.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Sat, 14 Dec 2002 01:32:09 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": true, "msg_subject": "Information schema now available" }, { "msg_contents": "Peter Eisentraut kirjutas L, 14.12.2002 kell 05:32:\n> A basic version of the SQL information schema is now available in newly\n> initdb'ed database installations.\n\nCould you also post it somewhere as a plain SQL script for 7.3 ?\n\nIMHO this should become the default way for \\d, ODBC, JDBC, and other\nsimilar interfaces for getting at this information and making it\navailable for 7.3 would give the implementors of those a head start.\n\n> > There's still a bunch of work to do to\n> create all the views that the spec defines.\n\nI'm sure you will get more help if it is available as add-on for 7.3.\n\n-- \nHannu Krosing <hannu@tm.ee>\n", "msg_date": "15 Dec 2002 00:10:16 +0500", "msg_from": "Hannu Krosing <hannu@tm.ee>", "msg_from_op": false, "msg_subject": "Re: Information schema now available" }, { "msg_contents": "Hannu Krosing writes:\n\n> Could you also post it somewhere as a plain SQL script for 7.3 ?\n\nIt depends on SQL features not present in 7.3. But you can download it\nfrom the CVS web interface and make the adjustments yourself if you want\nto use it.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Sun, 15 Dec 2002 17:22:31 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": true, "msg_subject": "Re: Information schema now available" } ]
[ { "msg_contents": "Dear all,\n\nWhy not use libgda, Gnome-DB database provider, to be able to attach foreign \ntables inside PostgreSQL. Would it be hard to achieve?\n\nMany users are looking for such a solution to be able to query/update tables \noutside PostgreSQL in Oracle, MS SQL Server, IBM DB2 or even MySQL databases.\n\nCheers,\nJean-Michel POURE\n", "msg_date": "Sat, 14 Dec 2002 10:24:38 +0100", "msg_from": "Jean-Michel POURE <jm.poure@freesurf.fr>", "msg_from_op": true, "msg_subject": "Big 7.4 items : table attachment" } ]
[ { "msg_contents": "\n\n> -----Original Message-----\n> From: Hannu Krosing [mailto:hannu@tm.ee] \n> Sent: 14 December 2002 19:10\n> To: Peter Eisentraut\n> Cc: PostgreSQL Development\n> Subject: Re: [HACKERS] Information schema now available\n> \n> \n> Peter Eisentraut kirjutas L, 14.12.2002 kell 05:32:\n> > A basic version of the SQL information schema is now available in \n> > newly initdb'ed database installations.\n> \n> Could you also post it somewhere as a plain SQL script for 7.3 ?\n> \n> IMHO this should become the default way for \\d, ODBC, JDBC, \n> and other similar interfaces for getting at this information \n> and making it available for 7.3 would give the implementors \n> of those a head start.\n\nYes. I agree. pgAdmin III is at a stage in development now where I could\nquite easily convert. In a couple of months it'll be a great deal more\nwork.\n\nRegards, Dave.\n", "msg_date": "Sat, 14 Dec 2002 19:58:33 -0000", "msg_from": "\"Dave Page\" <dpage@vale-housing.co.uk>", "msg_from_op": true, "msg_subject": "Re: Information schema now available" } ]
[ { "msg_contents": "The description of AsciiRow says that for each field it contains \"the \nvalue of the field itself in ASCII characters\". The description is here:\n \nhttp://postgresql.org/users-lounge/docs/7.3/postgres/protocol-message-formats.html\n\nLooking at the bytes comming over the Frontend/Backend protocol, \nhowever, it seems that the field is really always a String encoded \naccording to the settings in the database - is that right?\n\nThanks.\n\n", "msg_date": "Mon, 16 Dec 2002 11:24:21 +1100", "msg_from": "Bruce Badger <bruce_badger@BadgerSE.com>", "msg_from_op": true, "msg_subject": "AsciiRow" }, { "msg_contents": "Bruce Badger <bruce_badger@BadgerSE.com> writes:\n> The description of AsciiRow says that for each field it contains \"the \n> value of the field itself in ASCII characters\". The description is here:\n \n> http://postgresql.org/users-lounge/docs/7.3/postgres/protocol-message-formats.html\n\n> Looking at the bytes comming over the Frontend/Backend protocol, \n> however, it seems that the field is really always a String encoded \n> according to the settings in the database - is that right?\n\nI believe this documentation's reference to ASCII shouldn't be taken\nvery literally; it really just means \"text\" (as opposed to some internal\nbinary representation) and is not meant to imply anything about one\ncharacter set encoding versus another. You're correct that the data\nwould actually appear in the currently-selected client encoding,\nassuming that the backend is configured to support encoding conversions.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 16 Dec 2002 13:13:26 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: AsciiRow " }, { "msg_contents": "Hello everybody,\n\nSome days ago was released pgaccess 0.98.8 This release marks the end \nfirst phase of the work of the Redux team that came together in April \nthis year - Chris, Bartus, Brett, and some others.\n\nAfter the release was renewed the practice weekly releases to be made. \nThey have the name pgaccess 0.98.8.20021216 - the last is the date of \nthe release, they mark the progress towards pgaccess 0.98.9\n\nThe anonymus CVS access still exists.\n\nThere are many new features, fixed bugs and improvements in pgaccess \n0.98.8 in comparison to 0.98.7 - I will not use the space here, but all \ninterested can find them on the web site, and test it themselves.\n\nWhat is may be most interesting is that several people start real small \napplications based on pgaccess. Bartus does something for the company he \nworks for - that is pgaccess based. Chris has several small applications \nin the pipeline (and also uses pgaccess for the company he works for). \nThere are one or two more names of people who are known to work actively \non using pgaccess as application building environment.\n\nWe hope that with this news we have made many of you happy. Those who \ndid not believe that anything good will come out of it - now have at \nleast a little bit proof.\n\nWhere are we heading from now? Improving pgaccess as an administrative \ntool. And preparing the small foundations for making it an easier and \nbetter applications building environment.\n\nWhat we would like to see is a strong PostgreSQL. On UNIX, on Windows, \non Mac OS. On anything where Tcl/Tk works.\n\npgaccess has still open areas where people can work. Once on the very \ncore and the surrounding - documentation, tutorial. Second - broad and \nendless - on small applications built using pgaccess. If anybody is \ninteresting - more information can be found on the web site.\n\nWith best Christmas wishes for everybody,\n\nThe pgaccess development team\n\nwww.pgaccess.org\n\n\n", "msg_date": "Mon, 16 Dec 2002 23:28:01 +0100", "msg_from": "Iavor Raytchev <pobox@verysmall.org>", "msg_from_op": false, "msg_subject": "pgaccess 0.98.8 - released" } ]
[ { "msg_contents": "Hello hackers,\n\nI wonder what the PageGetMaxOffsetNumber macro is supposed to do (in a\nbtree index page)?\n\nThe scenario is the following: I need to get the pointer to a btree\npage (leaf or internal) in its parent page, to call _bt_itemdel on it.\nThe only thing I can think of is to use its parent link and then iterate\nin every OffsetNumber to see which item points down to the page I'm\nlooking for. I think the gotcha here is that the parent page can split\nand I may have to walk right in order to find the new parent (Is this\nright? Can the parent go someplace else?)\n\nI iterate over the elements of the parent page in a for loop, and the\nupper bound is rarely reached because the item is found. However\nsometimes the item isn't found, and PageGetItem fails its assertion\nbecause the item isn't used (LP_USED). I have found that\nPageGetMaxOffsetNumber (the upper bound) returns a consistent value\nthat's far too high (4294967291, 0xFFFFFFFB) and therefore the for loop\neventually falls out of bounds.\n\n\nThe btree freelist patch is \"almost ready\" (which means it works in the\ntrivial cases I've tested and there are some corner cases I haven't\neven covered), however I am stuck on this. Can anyone give me a light?\nMaybe there's another way to get at the pointer to a page in its parent?\n\n-- \nAlvaro Herrera (<alvherre[a]dcc.uchile.cl>)\n\"Crear es tan dificil como ser libre\" (Elsa Triolet)\n", "msg_date": "Sun, 15 Dec 2002 23:49:57 -0300", "msg_from": "Alvaro Herrera <alvherre@dcc.uchile.cl>", "msg_from_op": true, "msg_subject": "PageGetMaxOffsetNumber" }, { "msg_contents": "On Sun, Dec 15, 2002 at 11:49:57PM -0300, Alvaro Herrera wrote:\n\n> I iterate over the elements of the parent page in a for loop, and the\n> upper bound is rarely reached because the item is found. However\n> sometimes the item isn't found, and PageGetItem fails its assertion\n> because the item isn't used (LP_USED). I have found that\n> PageGetMaxOffsetNumber (the upper bound) returns a consistent value\n> that's far too high (4294967291, 0xFFFFFFFB) and therefore the for loop\n> eventually falls out of bounds.\n\nFWIW, the code that is supposed to do this (and appears to work fine on\nmost cases) is the following. buf is the page I'm going to free and\npblkno is the BlockNumber of its parent as seen in btpo_parent.\n\n/*\n * Delete the pointer to a child page. If the parent page is empty after\n * the deletion, delete the pointer from its parent too.\n */\nstatic void\n_bt_deletefromparent(Relation rel, BlockNumber pblkno, Buffer buf)\n{\n\tBuffer\t\t\tpbuf;\n\tOffsetNumber\toffnum;\n\tPage\t\t\tppage;\n\tBTPageOpaque\tpop;\n\tBlockNumber\t\tblkno = BufferGetBlockNumber(buf),\n\t\t\t\t\tmax;\n\n\tpbuf = _bt_getbuf(rel, pblkno, BT_WRITE);\n\tAssert(!BufferIsInvalid(pbuf));\n\n\tppage = BufferGetPage(pbuf);\n\tpop = (BTPageOpaque) PageGetSpecialPointer(ppage);\n\n\t/*\n\t * Repeat until the correct parent page is found.\tSplits may\n\t * cause the parent page to move right.\n\t */\n\tfor (;;)\n\t{\n\t\tBlockNumber\tnext;\n\n\t\t/* Make sure no one else tries to look at the page. */\n\t\tLockBuffer(pbuf, BUFFER_LOCK_UNLOCK);\n\t\tLockBufferForCleanup(pbuf);\n\t\tmax = PageGetMaxOffsetNumber(pop);\n\n\t\t/*\n\t\t * Look every offset of the page for the item pointing to the\n\t\t * dead page.\n\t\t */\n\t\tfor (offnum = FirstOffsetNumber; offnum <= max; offnum++)\n\t\t{\n\t\t\tBTItem\t\t\titem;\n\t\t\tItemPointer\t\tiptr;\n\n\t\t\titem = (BTItem) PageGetItem(ppage, PageGetItemId(ppage, offnum));\n\t\t\tiptr = &(item->bti_itup.t_tid);\n\t\t\tif (ItemPointerGetBlockNumber(iptr) == blkno)\n\t\t\t{\n\t\t\t\t/* Ok, we found the page. Now delete the pointer. */\n\t\t\t\tItemPointer\t\tiptrd = palloc(SizeOfIptrData);\n\n\t\t\t\tItemPointerSet(iptrd, pblkno, offnum);\n\t\t\t\t_bt_itemdel(rel, pbuf, iptrd);\n\t\t\t\t_bt_wrtnorelbuf(rel, pbuf);\n\t\t\t\tLockBuffer(pbuf, BUFFER_LOCK_UNLOCK);\n\n\t\t\t\t/*\n\t\t\t\t * If the parent page is empty after this deletion,\n\t\t\t\t * mark it dead and free it too.\n\t\t\t\t */\n\t\t\t\tif (_bt_pageisempty(ppage))\n\t\t\t\t{\n\t\t\t\t\tpop->btpo_flags |= BTP_DEAD;\n\t\t\t\t\t_bt_processdead(rel, pbuf);\n\t\t\t\t}\n\t\t\t\tReleaseBuffer(pbuf);\n\t\t\t\treturn;\n\t\t\t}\n\t\t}\n\n\t\t/*\n\t\t * If we just finished scanning the rightmost page of the upper level,\n\t\t * there's some wrong.\n\t\t */\n\t\tif (P_RIGHTMOST(pop))\n\t\t\telog(ERROR, \"Unable to find parent page!\");\n\n\t\t/* Oops, the parent was split.\tCheck its right sibling. */\n\t\tnext = pop->btpo_next;\n\t\t_bt_relbuf(rel, pbuf);\n\t\tpbuf = _bt_getbuf(rel, next, BT_WRITE);\n\t\tppage = BufferGetPage(pbuf);\n\t\tpop = (BTPageOpaque) PageGetSpecialPointer(ppage);\n\t}\n}\n\n-- \nAlvaro Herrera (<alvherre[a]dcc.uchile.cl>)\n\"Aprende a avergonzarte mas ante ti que ante los demas\" (Democrito)\n", "msg_date": "Mon, 16 Dec 2002 00:05:20 -0300", "msg_from": "Alvaro Herrera <alvherre@dcc.uchile.cl>", "msg_from_op": true, "msg_subject": "Re: PageGetMaxOffsetNumber" }, { "msg_contents": "On Sun, 15 Dec 2002 23:49:57 -0300, Alvaro Herrera\n<alvherre@dcc.uchile.cl> wrote:\n>PageGetMaxOffsetNumber (the upper bound) returns a consistent value\n>that's far too high (4294967291, 0xFFFFFFFB)\n\nAlvaro, maybe this comment from bufpage.h can shed some light on it?\n\n/*\n * PageGetMaxOffsetNumber\n * Returns the maximum offset number used by the given page.\n * Since offset numbers are 1-based, this is also the number\n * of items on the page.\n *\n * NOTE: to ensure sane behavior if the page is not initialized\n * (pd_lower == 0), cast the unsigned values to int before dividing.\n * That way we get -1 or so, not a huge positive number...\n */\n\n#define PageGetMaxOffsetNumber(page) \\\n (((int) (((PageHeader) (page))->pd_lower - SizeOfPageHeaderData)) \\\n / ((int) sizeof(ItemIdData)))\n\n0xFFFFFFFB is -5.\nWith SizeOfPageHeaderData == 20 and sizeof(ItemIdData) == 4 you get\nthis result for pd_lower == 0.\n\nServus\n Manfred\n", "msg_date": "Mon, 16 Dec 2002 09:58:55 +0100", "msg_from": "Manfred Koizar <mkoi-pg@aon.at>", "msg_from_op": false, "msg_subject": "Re: PageGetMaxOffsetNumber" }, { "msg_contents": "Alvaro Herrera <alvherre@dcc.uchile.cl> writes:\n> \tppage = BufferGetPage(pbuf);\n> \tpop = (BTPageOpaque) PageGetSpecialPointer(ppage);\n\n> \t\tmax = PageGetMaxOffsetNumber(pop);\n\nI believe you want PageGetMaxOffsetNumber(ppage) ...\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 16 Dec 2002 13:21:40 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: PageGetMaxOffsetNumber " } ]
[ { "msg_contents": "hello, Joe Conway. \n\nit seems the crosstab function you write in contrib/tablefunc can't work.\nit saids that A column definition list is required for functions returning RECORD.\nHow can i make it work?\nthank you very much\n\nbest regard,\n\nJinqiang Han\n\n\n\n\n", "msg_date": "Mon, 16 Dec 2002 12:8:48 +0800", "msg_from": "\"Jinqiang Han\" <postgresql@db.pku.edu.cn>", "msg_from_op": true, "msg_subject": "about table function" }, { "msg_contents": "Jinqiang Han wrote:\n> it seems the crosstab function you write in contrib/tablefunc can't work.\n> it saids that A column definition list is required for functions returning RECORD.\n> How can i make it work?\n\n>From contrib/tablefunc/README.tablefunc:\n\nOutputs\n\n Returns setof record, which must defined with a column definition\n in the FROM clause of the SELECT statement, e.g.:\n\n SELECT *\n FROM crosstab(sql, 2)\n AS ct(row_name text, category_1 text, category_2 text);\n\n the example crosstab function produces a set something like:\n <== values columns ==>\n row_name category_1 category_2\n ---------+------------+------------\n row1 val1 val2\n row2 val5 val6\n\n\nYou are apparently forgetting the\n \"AS ct(row_name text, category_1 text, category_2 text)\"\npart.\n\nhope this helps,\n\nJoe\n\n", "msg_date": "Sun, 15 Dec 2002 20:25:44 -0800", "msg_from": "Joe Conway <mail@joeconway.com>", "msg_from_op": false, "msg_subject": "Re: about table function" } ]
[ { "msg_contents": "I've now got a copy of pg_filedump and compiled it, can you tell me the\ncommand line parameters to pass it (and the file that I must process) so I\ncan give you exactly what you require.\n\nThanks\nSteve\n\n-----Original Message-----\nFrom: Tom Lane [mailto:tgl@sss.pgh.pa.us]\nSent: 13 December 2002 14:43\nTo: Steve King\nCc: 'Hannu Krosing'; 'pgsql-hackers@postgresql.org'\nSubject: Re: [HACKERS] FW: Duplicate oids!\n\n\nSteve King <steve.king@ecmsys.co.uk> writes:\n> The ctids are different however vaccum is run on this table and the record\n> is updated.\n\nIt would be useful to look at xmin,xmax,cmin,cmax of these tuples too.\n\nActually, if you don't mind grabbing a copy of pg_filedump --- see\nhttp://sources.redhat.com/rhdb/tools.html --- then a dump of page 7\nof that relation would be good to look at. I am wondering about the\nstates of the infomask bits for these tuples...\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 16 Dec 2002 10:56:39 -0000", "msg_from": "Steve King <steve.king@ecmsys.co.uk>", "msg_from_op": true, "msg_subject": "Re: FW: Duplicate oids!" }, { "msg_contents": "Steve King <steve.king@ecmsys.co.uk> writes:\n> I've now got a copy of pg_filedump and compiled it, can you tell me the\n> command line parameters to pass it (and the file that I must process) so I\n> can give you exactly what you require.\n\nI'd recommend\n\n\tpg_filedump -f -i -R <blocknum> <filename>\n\nwhere <blocknum> is whatever page you need to look at (the high part of\nthe TIDs of the bad tuples), and <filename> is going to be of the form\n\n\t$PGDATA/base/<dbnum>/<filenum>\n\nYou get the DB number from\n\n\tselect oid from pg_database where datname = 'yourdbname';\n\nand the file number from\n\n\tselect relfilenode from pg_class where relname = 'yourtablename';\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 16 Dec 2002 09:59:21 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: FW: Duplicate oids! " } ]
[ { "msg_contents": "Hi,\n\nI have a trigger in my database that checks to see if there is another \nrecord in the table, and when there is if the type is correct. (if the \nfirst one is of type \"parent\", the other has to be of type \"child\").\n\nWhen updating these records in a transaction, the trigger only works when \nI make it deferrable. I hacked this by updating pg_trigger and setting the\ncolumn tgdeferrable to true.\n\nIS there a way to do this in SQL? I can only find documentation on setting \na constraint to deferrable, but not on a trigger.\n\nTIA,\nReinoud\n\n-- \n__________________________________________________\n\"Nothing is as subjective as reality\"\nReinoud van Leeuwen reinoud.v@n.leeuwen.net\nhttp://www.xs4all.nl/~reinoud\n__________________________________________________\n", "msg_date": "Mon, 16 Dec 2002 15:18:11 +0100", "msg_from": "Reinoud van Leeuwen <reinoud@xs4all.nl>", "msg_from_op": true, "msg_subject": "how to make a trigger deferrable" } ]
[ { "msg_contents": "CVSROOT:\t/cvsroot\nModule name:\tpgsql-server\nChanges by:\ttgl@postgresql.org\t02/12/16 11:22:46\n\nModified files:\n\tsrc/backend/nodes: nodes.c \n\tsrc/backend/utils/mmgr: mcxt.c \n\tsrc/include : c.h \n\tsrc/include/nodes: nodes.h \n\tsrc/include/utils: memutils.h palloc.h \n\nLog message:\n\tCode review for palloc0 patch --- avoid dangerous and unnecessary\n\tpractice of evaluating MemSet's arguments multiple times, except for\n\tthe special case of newNode(), where we can assume the argument is\n\ta constant sizeof() operator.\n\tAlso, add GetMemoryChunkContext() to mcxt.c's API, in preparation for\n\tfixing recent GEQO breakage.\n\n", "msg_date": "Mon, 16 Dec 2002 11:22:46 -0500 (EST)", "msg_from": "tgl@postgresql.org (Tom Lane)", "msg_from_op": true, "msg_subject": "pgsql-server/src backend/nodes/nodes.c backend ..." }, { "msg_contents": "Hi guys,\n\nJust a thought - do we explicitly wipe password strings from RAM after using\nthem?\n\nI just read an article (by MS in fact) that illustrates a cute problem.\nImagine you memset the password to zeros after using it. There is a good\nchance that the compiler will simply remove the memset from the object code\nas it will seem like it can be optimised away...\n\nJust wondering...\n\nChris\n\n", "msg_date": "Tue, 17 Dec 2002 10:07:55 +0800", "msg_from": "\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au>", "msg_from_op": false, "msg_subject": "Password security question" }, { "msg_contents": "On Tue, 17 Dec 2002, Christopher Kings-Lynne wrote:\n\n> Hi guys,\n> \n> Just a thought - do we explicitly wipe password strings from RAM after using\n> them?\n> \n> I just read an article (by MS in fact) that illustrates a cute problem.\n> Imagine you memset the password to zeros after using it. There is a good\n> chance that the compiler will simply remove the memset from the object code\n> as it will seem like it can be optimised away...\n\nBugtraq discussion claims that GCC >=3 are not affected by this. Variables\nwhich are affected by code that cannot be optimised away should be marked\nvolitile anyway.\n\nGavin\n\n", "msg_date": "Tue, 17 Dec 2002 13:17:49 +1100 (EST)", "msg_from": "Gavin Sherry <swm@linuxworld.com.au>", "msg_from_op": false, "msg_subject": "Re: Password security question" }, { "msg_contents": "\n\nChristopher Kings-Lynne wrote:\n\n>Hi guys,\n>\n>Just a thought - do we explicitly wipe password strings from RAM after using\n>them?\n>\n>I just read an article (by MS in fact) that illustrates a cute problem.\n>Imagine you memset the password to zeros after using it. There is a good\n>chance that the compiler will simply remove the memset from the object code\n>as it will seem like it can be optimised away...\n>\n>Just wondering...\n>\n>Chris\n> \n>\nCould you post that link? That seems wrong, an explicit memset certainly \nchanges the operation of the code, and thus should not be optimized away.\n\n> \n>\n\n\n", "msg_date": "Tue, 17 Dec 2002 11:49:47 -0500", "msg_from": "mlw <pgsql@mohawksoft.com>", "msg_from_op": false, "msg_subject": "Re: Password security question" }, { "msg_contents": "On Tue, 2002-12-17 at 10:49, mlw wrote:\n> Christopher Kings-Lynne wrote:\n> \n> >Hi guys,\n> >\n> >Just a thought - do we explicitly wipe password strings from RAM after using\n> >them?\n> >\n> >I just read an article (by MS in fact) that illustrates a cute problem.\n> >Imagine you memset the password to zeros after using it. There is a good\n> >chance that the compiler will simply remove the memset from the object code\n> >as it will seem like it can be optimised away...\n> >\n> >Just wondering...\n> >\n> >Chris\n> > \n> >\n> Could you post that link? That seems wrong, an explicit memset certainly \n> changes the operation of the code, and thus should not be optimized away.\n> \n> > \n> >\n> \n\nI'd like to see the link too.\n\nI can imagine that it would be possible for it to optimize it away if\nthere wasn't an additional read/write access which followed. In other\nwords, why do what is more or less a no-op if it's never accessed again.\n\n\n-- \nGreg Copeland <greg@copelandconsulting.net>\nCopeland Computer Consulting\n\n", "msg_date": "17 Dec 2002 11:00:19 -0600", "msg_from": "Greg Copeland <greg@CopelandConsulting.Net>", "msg_from_op": false, "msg_subject": "Re: Password security question" }, { "msg_contents": "http://msdn.microsoft.com/library/en-us/dncode/html/secure10102002.asp\n\n", "msg_date": "Tue, 17 Dec 2002 12:11:21 -0500", "msg_from": "\"Ken Hirsch\" <kenhirsch@myself.com>", "msg_from_op": false, "msg_subject": "Re: Password security question" }, { "msg_contents": "On Tue, 2002-12-17 at 11:11, Ken Hirsch wrote:\n> http://msdn.microsoft.com/library/en-us/dncode/html/secure10102002.asp\n> \n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org\n\n\nThanks. Seems I hit the nail on the head. ;)\n\n\n-- \nGreg Copeland <greg@copelandconsulting.net>\nCopeland Computer Consulting\n\n", "msg_date": "17 Dec 2002 11:28:11 -0600", "msg_from": "Greg Copeland <greg@CopelandConsulting.Net>", "msg_from_op": false, "msg_subject": "Re: Password security question" }, { "msg_contents": "\n\nKen Hirsch wrote:\n\n>http://msdn.microsoft.com/library/en-us/dncode/html/secure10102002.asp\n>\n> \n>\nWell, OK, that isn't as bizarre as one could have expected.\n\n> \n>\n\n\n", "msg_date": "Tue, 17 Dec 2002 13:26:12 -0500", "msg_from": "mlw <pgsql@mohawksoft.com>", "msg_from_op": false, "msg_subject": "Re: Password security question" }, { "msg_contents": "\n\nGreg Copeland wrote:\n\n>On Tue, 2002-12-17 at 10:49, mlw wrote:\n> \n>\n>>Christopher Kings-Lynne wrote:\n>>\n>> \n>>\n>>>Hi guys,\n>>>\n>>>Just a thought - do we explicitly wipe password strings from RAM after using\n>>>them?\n>>>\n>>>I just read an article (by MS in fact) that illustrates a cute problem.\n>>>Imagine you memset the password to zeros after using it. There is a good\n>>>chance that the compiler will simply remove the memset from the object code\n>>>as it will seem like it can be optimised away...\n>>>\n>>>Just wondering...\n>>>\n>>>Chris\n>>> \n>>>\n>>> \n>>>\n>>Could you post that link? That seems wrong, an explicit memset certainly \n>>changes the operation of the code, and thus should not be optimized away.\n>>\n>> \n>>\n>>> \n>>>\n>>> \n>>>\n>\n>I'd like to see the link too.\n>\n>I can imagine that it would be possible for it to optimize it away if\n>there wasn't an additional read/write access which followed. In other\n>words, why do what is more or less a no-op if it's never accessed again.\n> \n>\nIt has been my experience that the MSC optimizer uses a patented \nHeisenberg optimizer. :)\n\n>\n> \n>\n\n\n", "msg_date": "Tue, 17 Dec 2002 13:30:21 -0500", "msg_from": "mlw <pgsql@mohawksoft.com>", "msg_from_op": false, "msg_subject": "Re: Password security question" } ]
[ { "msg_contents": "Tom, Folks:\n\nJoe and I were discussing your recent discussion about the costs of VACUUM and \ntuple maintainence, and I had an interesting idea.\n\nHow hard would it be to add a \"WITH (VACUUM)\" option to UPDATE and DELETE \nqueries? This option would cause the regular vacuum activity -- purging the \ndead tuple and its index references -- to be done immediately, as part of the \nstatement, instead of being deferred.\n\nI agree that we don't want immediate tuple maintainence, most of the time. \nHowever, a couple of my clients would really like to have a few queries do \ntheir vacuum immediately, and are currently getting around that by vacuuming \nevery 5 mintutes.\n\nEasy? Hard? Insane? What do you think?\n\n-- \n-Josh Berkus\n Aglio Database Solutions\n San Francisco\n\n", "msg_date": "Mon, 16 Dec 2002 14:00:53 -0800", "msg_from": "Josh Berkus <josh@agliodbs.com>", "msg_from_op": true, "msg_subject": "Suggestion; \"WITH VACUUM\" option" }, { "msg_contents": "On Mon, 16 Dec 2002, Josh Berkus wrote:\n\n> Tom, Folks:\n>\n> Joe and I were discussing your recent discussion about the costs of VACUUM and\n> tuple maintainence, and I had an interesting idea.\n>\n> How hard would it be to add a \"WITH (VACUUM)\" option to UPDATE and DELETE\n> queries? This option would cause the regular vacuum activity -- purging the\n> dead tuple and its index references -- to be done immediately, as part of the\n> statement, instead of being deferred.\n>\n> I agree that we don't want immediate tuple maintainence, most of the time.\n> However, a couple of my clients would really like to have a few queries do\n> their vacuum immediately, and are currently getting around that by vacuuming\n> every 5 mintutes.\n>\n> Easy? Hard? Insane? What do you think?\n\nJust curious, but wouldn't it be just as simple to issue a VACUUM call\nright after the UPDATE/DELETE?\n", "msg_date": "Mon, 16 Dec 2002 18:15:32 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: Suggestion; \"WITH VACUUM\" option" }, { "msg_contents": "\nMarc,\n\n> > Easy? Hard? Insane? What do you think?\n> \n> Just curious, but wouldn't it be just as simple to issue a VACUUM call\n> right after the UPDATE/DELETE?\n\nWell, you can't do that as part of a transaction or procedure, whereas ....\n\nHmmm. Couldn't do \"with vacuum\" as part of a transaction, either. \n\nFunny, it sounded like a really good idea on the phone. \n\nIn the words of Rosanne Rosanna Danna, \n\"Neeeveeeeeer miiiiiiiiind.\"\n\n-- \n-Josh Berkus\n Aglio Database Solutions\n San Francisco\n\n", "msg_date": "Mon, 16 Dec 2002 14:25:59 -0800", "msg_from": "Josh Berkus <josh@agliodbs.com>", "msg_from_op": true, "msg_subject": "Re: Suggestion; \"WITH VACUUM\" option" }, { "msg_contents": "Josh Berkus <josh@agliodbs.com> writes:\n> How hard would it be to add a \"WITH (VACUUM)\" option to UPDATE and DELETE \n> queries? This option would cause the regular vacuum activity -- purging the\n> dead tuple and its index references -- to be done immediately, as part of the\n> statement, instead of being deferred.\n\n> Easy? Hard? Insane? What do you think?\n\nImpossible. You can't vacuum a tuple until the last open transaction\nthat can see it is gone. It is therefore *impossible* for a transaction\nto vacuum away its own detritus; until the transaction commits, you\ncan't even start to wonder whether other open transactions see it or\nnot.\n\nVacuuming has to be done later, and that being the case, I don't see any\nreal advantage to altering the \"background vacuum\" design we have.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 16 Dec 2002 17:33:14 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Suggestion; \"WITH VACUUM\" option " }, { "msg_contents": "On Mon, 16 Dec 2002, Tom Lane wrote:\n\n> Josh Berkus <josh@agliodbs.com> writes:\n> > How hard would it be to add a \"WITH (VACUUM)\" option to UPDATE and DELETE \n> > queries? This option would cause the regular vacuum activity -- purging the\n> > dead tuple and its index references -- to be done immediately, as part of the\n> > statement, instead of being deferred.\n> \n> > Easy? Hard? Insane? What do you think?\n> \n> Impossible. You can't vacuum a tuple until the last open transaction\n> that can see it is gone. It is therefore *impossible* for a transaction\n> to vacuum away its own detritus; until the transaction commits, you\n> can't even start to wonder whether other open transactions see it or\n> not.\n> \n> Vacuuming has to be done later, and that being the case, I don't see any\n> real advantage to altering the \"background vacuum\" design we have.\n\nThen, would a \"commit with vacuum\" work? OR a \"begin transaction with \nvacuum\" Just tossing them out there...\n\n", "msg_date": "Mon, 16 Dec 2002 15:52:30 -0700 (MST)", "msg_from": "\"scott.marlowe\" <scott.marlowe@ihs.com>", "msg_from_op": false, "msg_subject": "Re: Suggestion; \"WITH VACUUM\" option " }, { "msg_contents": "On Mon, 16 Dec 2002, scott.marlowe wrote:\n\n> On Mon, 16 Dec 2002, Tom Lane wrote:\n>\n> > Josh Berkus <josh@agliodbs.com> writes:\n> > > How hard would it be to add a \"WITH (VACUUM)\" option to UPDATE and DELETE\n> > > queries? This option would cause the regular vacuum activity -- purging the\n> > > dead tuple and its index references -- to be done immediately, as part of the\n> > > statement, instead of being deferred.\n> >\n> > > Easy? Hard? Insane? What do you think?\n> >\n> > Impossible. You can't vacuum a tuple until the last open transaction\n> > that can see it is gone. It is therefore *impossible* for a transaction\n> > to vacuum away its own detritus; until the transaction commits, you\n> > can't even start to wonder whether other open transactions see it or\n> > not.\n> >\n> > Vacuuming has to be done later, and that being the case, I don't see any\n> > real advantage to altering the \"background vacuum\" design we have.\n>\n> Then, would a \"commit with vacuum\" work? OR a \"begin transaction with\n> vacuum\" Just tossing them out there...\n\nTom will correct me here, but I believe what he was trying to get across\nisn't that the 'current transaction' is the problem ... the problem is the\nother connections who have open transactions ... my simplistic\nunderstanding (and I hope it isn't too flawed) of MVCC is that as long as\n*one* transaction is outstanding on a tuple, that tuple can't be\nphysically removed ... as far as any new transactions are concerned, it\nhas disappeared ... so if I open a transaction, then you open one 'with\nvacuum', your 'with vacuum' will fail unless I happen to be out of my\ntransaction before you ...\n\n\n", "msg_date": "Mon, 16 Dec 2002 19:43:35 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: Suggestion; \"WITH VACUUM\" option " }, { "msg_contents": "On Mon, 16 Dec 2002, Marc G. Fournier wrote:\n\n> On Mon, 16 Dec 2002, scott.marlowe wrote:\n> \n> > On Mon, 16 Dec 2002, Tom Lane wrote:\n> >\n> > > Josh Berkus <josh@agliodbs.com> writes:\n> > > > How hard would it be to add a \"WITH (VACUUM)\" option to UPDATE and DELETE\n> > > > queries? This option would cause the regular vacuum activity -- purging the\n> > > > dead tuple and its index references -- to be done immediately, as part of the\n> > > > statement, instead of being deferred.\n> > >\n> > > > Easy? Hard? Insane? What do you think?\n> > >\n> > > Impossible. You can't vacuum a tuple until the last open transaction\n> > > that can see it is gone. It is therefore *impossible* for a transaction\n> > > to vacuum away its own detritus; until the transaction commits, you\n> > > can't even start to wonder whether other open transactions see it or\n> > > not.\n> > >\n> > > Vacuuming has to be done later, and that being the case, I don't see any\n> > > real advantage to altering the \"background vacuum\" design we have.\n> >\n> > Then, would a \"commit with vacuum\" work? OR a \"begin transaction with\n> > vacuum\" Just tossing them out there...\n> \n> Tom will correct me here, but I believe what he was trying to get across\n> isn't that the 'current transaction' is the problem ... the problem is the\n> other connections who have open transactions ... my simplistic\n> understanding (and I hope it isn't too flawed) of MVCC is that as long as\n> *one* transaction is outstanding on a tuple, that tuple can't be\n> physically removed ... as far as any new transactions are concerned, it\n> has disappeared ... so if I open a transaction, then you open one 'with\n> vacuum', your 'with vacuum' will fail unless I happen to be out of my\n> transaction before you ...\n\nOh, yeah I have no doubt of that. I was thinking more along the lines of \nwhen a transaction ends it throws a background \"vacuum table1;vacuum \ntable2;vacuum tablen\" command into some kind of vacuuming hopper. I.e. it \ndoesn't block waiting, it runs it as though it were run AFTER the \ntransaction. If there are a few tuples from other transactions we can't \nreclaim, no big deal. \n\nThe other option would be some kind of GUC that set a max number of rows \ndeleted/updated in a table in a transaction that would trigger this kind \nof thing automagically.\n\nBut I could see such a setting causing just as much harm (chaos theory \nanyone? :-) as good.\n\n", "msg_date": "Mon, 16 Dec 2002 17:17:34 -0700 (MST)", "msg_from": "\"scott.marlowe\" <scott.marlowe@ihs.com>", "msg_from_op": false, "msg_subject": "Re: Suggestion; \"WITH VACUUM\" option " }, { "msg_contents": "\"scott.marlowe\" <scott.marlowe@ihs.com> writes:\n> Oh, yeah I have no doubt of that. I was thinking more along the lines of \n> when a transaction ends it throws a background \"vacuum table1;vacuum \n> table2;vacuum tablen\" command into some kind of vacuuming hopper.\n\nActually, the plans I liked best for driving auto-vacuum were\nessentially an indirect version of that: the FSM module would keep track\nof committed deletes + aborted inserts for each active table, and then\nthe autovacuum scheduler could use that info to decide which tables are\nhighest-priority to vacuum.\n\n(Or possibly the runtime stats module would be a better place to track\nit than FSM.)\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 16 Dec 2002 20:16:27 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Suggestion; \"WITH VACUUM\" option " }, { "msg_contents": "\n\nTom Lane wrote:\n\n>Josh Berkus <josh@agliodbs.com> writes:\n> \n>\n>>How hard would it be to add a \"WITH (VACUUM)\" option to UPDATE and DELETE \n>>queries? This option would cause the regular vacuum activity -- purging the\n>>dead tuple and its index references -- to be done immediately, as part of the\n>>statement, instead of being deferred.\n>> \n>>\n>\n> \n>\n>>Easy? Hard? Insane? What do you think?\n>> \n>>\n>\n>Impossible. You can't vacuum a tuple until the last open transaction\n>that can see it is gone. It is therefore *impossible* for a transaction\n>to vacuum away its own detritus; until the transaction commits, you\n>can't even start to wonder whether other open transactions see it or\n>not.\n>\n>Vacuuming has to be done later, and that being the case, I don't see any\n>real advantage to altering the \"background vacuum\" design we have.\n> \n>\nThis does raise an interresting question, and I understand that it is \n*impossible* to do with PostgreSQL as it currently exists, however, let \nme just toss this out there:\n\nSuppose you do this:\n\nupdate largetable set foo=bar;\n\nLets also assume that \"largetable\" has tens of millions of rows. I have \ndatabases like this, and I sometimes do operations like this. I have \nfound it more efficient to break up the update into a series of:\n\nupdate largetable set foo=bar where somefield < a;\nvacuum\nupdate largetable set foo=bar where somefield < b;\nvacuum\nupdate largetable set foo=bar where somefield < c;\nvacuum\nupdate largetable set foo=bar where not foo = bar;\nvacuum\n\nOn some of my databases a statement which updates all the rows is \nunworkable in PostgreSQL, on Oracle, however, there is no poblem.\n\nFor my use, it is a pain in the neck to deal with, but not unworkable. \nFor some other users, it may be a bigger problem.\n\n\n\n> \n>\n\n> \n>\n\n", "msg_date": "Tue, 17 Dec 2002 08:02:34 -0500", "msg_from": "mlw <pgsql@mohawksoft.com>", "msg_from_op": false, "msg_subject": "Re: Suggestion; \"WITH VACUUM\" option" }, { "msg_contents": "On Tue, 17 Dec 2002, mlw wrote:\n\n> update largetable set foo=bar;\n>\n> Lets also assume that \"largetable\" has tens of millions of rows.\n[..]\n> On some of my databases a statement which updates all the rows is\n> unworkable in PostgreSQL, on Oracle, however, there is no poblem.\n\n.. provided you have a lot of rollback space, which is\nessentially what the datafile growth here is providing.\n\nMatthew.\n\n", "msg_date": "Tue, 17 Dec 2002 18:38:11 +0000 (GMT)", "msg_from": "Matthew Kirkwood <matthew@hairy.beasts.org>", "msg_from_op": false, "msg_subject": "Re: Suggestion; \"WITH VACUUM\" option" } ]
[ { "msg_contents": "Steve King <steve.king@ecmsys.co.uk> writes:\n> Files output from pg_filedump are below,\n> I have two tables with duplicate oids and these are the pg_filedumps for\n> them.\n\nHmm. You seem to have a rather unusual usage pattern for these tables\n--- it looks like there are *lots* of failed (rolled back) updates of\nthe same tuple. Can you give us some details about the update commands\nbeing issued against these tables?\n\nAs an example, the page from the \"machine\" table has many entries like\nthis:\n\n Item 9 -- Length: 261 Offset: 5288 (0x14a8) Flags: USED\n OID: 9646238 CID: min(2) max(0) XID: min(119186229) max(0)\n Block Id: 7 linp Index: 9 Attributes: 41 Size: 32\n infomask: 0x2a02 (HASVARLENA|XMIN_INVALID|XMAX_INVALID|UPDATED) \n\nwhich is evidently a failed (aborted) update of the row with OID\n9646238. There are 27 other items on the same page with different\nXIDs --- all failed commands, according to the XMIN_INVALID bits ---\nand these two:\n\n Item 10 -- Length: 261 Offset: 7928 (0x1ef8) Flags: USED\n OID: 9646238 CID: min(5) max(2) XID: min(119162725) max(119208868)\n Block Id: 759 linp Index: 30 Attributes: 41 Size: 32\n infomask: 0x2902 (HASVARLENA|XMIN_COMMITTED|XMAX_INVALID|UPDATED) \n\n Item 18 -- Length: 261 Offset: 7664 (0x1df0) Flags: USED\n OID: 9646238 CID: min(2) max(2) XID: min(119162754) max(119208939)\n Block Id: 760 linp Index: 3 Attributes: 41 Size: 32\n infomask: 0x2902 (HASVARLENA|XMIN_COMMITTED|XMAX_INVALID|UPDATED) \n\nThese are both still valid (since their attempted deleting transactions\nboth rolled back, according to the XMAX_INVALID bits). But how did\ntransactions 119162725 and 119162754 both manage to commit updates of\nthe same row, without one invalidating the other?\n\nI wonder whether this is evidence of a race condition in the\nconcurrent-update logic. Can't tell without much more detail, though.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 16 Dec 2002 18:07:05 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": true, "msg_subject": "Re: FW: Duplicate oids! " } ]
[ { "msg_contents": "No :)\n\nMagnus\n", "msg_date": "Tue, 17 Dec 2002 01:33:02 +0100", "msg_from": "\"Magnus Naeslund(f)\" <mag@fbab.net>", "msg_from_op": true, "msg_subject": "Re: Is anybody out there !!!" }, { "msg_contents": "", "msg_date": "Tue, 17 Dec 2002 00:31:32 -0800", "msg_from": "\"zahid rahman\" <zahidr@btconnect.com>", "msg_from_op": false, "msg_subject": "Is anybody out there !!!" } ]
[ { "msg_contents": "(I forwarded the mail to the general list so that we keep onlist, next\ntime use 'reply to all' or similar...)\n\nzahid rahman <zahidr@btconnect.com> wrote:\n> Hi,\n>\n> Excellent!\n>\n> I have just downloaded postgresql.\n> I am in the process of downloading cygwin as per install file.\n> I follow instructions. I am a good boy!\n> I am a windows 2000 user. Ok ! I am not so good.\n>\n> My next instruction is that I will require a GCC compiler.\n> Where can I get so that I go to the sql prompt and run sql create\n> command\n> to create that elusive table without any hiccups.\n>\n\nI'm sorry to say that i don't know anything about using postgresql in\ncygwin.\nBut i think the safest bet is to use the cygwin install program to\ninstall postgresql.\nAfter that you just start postgresql as you normally would do i think.\n\nFor more info, see: http://www.se.postgresql.org/docs/faq-mswin.html\n\nMagnus\n\n", "msg_date": "Tue, 17 Dec 2002 02:00:47 +0100", "msg_from": "\"Magnus Naeslund(f)\" <mag@fbab.net>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] following instructions GCC" }, { "msg_contents": "\nIs this it ?\n\nOutput of cygwin....\n\nAdministrator@ABC-NQ0MBUYZ4EH ~\n$ pwd\n/home/Administrator\n\nAdministrator@ABC-NQ0MBUYZ4EH ~\n$ cd ..\n\nAdministrator@ABC-NQ0MBUYZ4EH /home\n$ cd ..\n\nAdministrator@ABC-NQ0MBUYZ4EH /\n$ ls -l\ntotal 890\ndrwxrwxrwx+ 3 Administ None 200704 Dec 17 01:44 bin\n-rwx------+ 1 Administ None 79657 Dec 17 01:21\ncygipc-1.13-2-src.tar.bz\n2\n-rwx------+ 1 Administ None 386405 Dec 17 01:21\ncygipc-1.13-2.tar.bz2\n-rwxrwxrwx 1 Administ None 88 Dec 17 02:04 cygwin.bat\n-rwxrwxrwx 1 Administ None 766 Dec 17 01:43 cygwin.ico\ndrwxrwxrwx+ 17 Administ None 8192 Dec 17 01:44 etc\ndrwxrwxrwx+ 3 Administ None 0 Dec 17 00:22\nftp%3a%2f%2fftp.mirror.a\nc.uk%2fsites%2fsources.redhat.com%2fpub%2fcygwin\ndrwxrwxrwx+ 3 Administ None 0 Dec 17 01:45 home\ndrwxrwxrwx+ 22 Administ None 53248 Dec 17 01:44 lib\ndrwxrwxrwx+ 2 Administ None 0 Dec 17 01:37 sbin\n-rwx------+ 1 Administ None 176128 Dec 17 00:20 setup.exe\ndrwxrwxrwx+ 2 Administ None 0 Dec 17 01:44 tmp\ndrwxrwxrwx+ 21 Administ None 4096 Dec 17 01:43 usr\ndrwxrwxrwx+ 8 Administ None 0 Dec 17 01:43 var\n\nAdministrator@ABC-NQ0MBUYZ4EH /\n$ ipc-daemon &\n[1] 768\n\nAdministrator@ABC-NQ0MBUYZ4EH /\n$ initdb -D /usr/share/postgresql/data\nThe files belonging to this database system will be owned by user\n\"Administrator\n\".\nThis user must also own the server process.\n\nThe database cluster will be initialized with locale C.\n\ncreating directory /usr/share/postgresql/data... ok\ncreating directory /usr/share/postgresql/data/base... ok\ncreating directory /usr/share/postgresql/data/global... ok\ncreating directory /usr/share/postgresql/data/pg_xlog... ok\ncreating directory /usr/share/postgresql/data/pg_clog... ok\ncreating template1 database in /usr/share/postgresql/data/base/1... ok\ncreating configuration files... ok\ninitializing pg_shadow... ok\nenabling unlimited row size for system tables... ok\ninitializing pg_depend... ok\ncreating system views... ok\nloading pg_description... ok\ncreating conversions... ok\nsetting privileges on built-in objects... ok\nvacuuming database template1... ok\ncopying template1 to template0... ok\n\nSuccess. You can now start the database server using:\n\n /usr/bin/postmaster -D /usr/share/postgresql/data\nor\n /usr/bin/pg_ctl -D /usr/share/postgresql/data -l logfile start\n\nAdministrator@ABC-NQ0MBUYZ4EH /\n$ pg_ctl -D /usr/share/postgresql/data -l logfile start\npostmaster successfully started\n\nAdministrator@ABC-NQ0MBUYZ4EH /\n$ psql template1\nWelcome to psql 7.3, the PostgreSQL interactive terminal.\n\nType: \\copyright for distribution terms\n \\h for help with SQL commands\n \\? for help on internal slash commands\n \\g or terminate with semicolon to execute query\n \\q to quit\n\ntemplate1=# \\h\nAvailable help:\n ABORT CREATE TABLE EXECUTE\n ALTER DATABASE CREATE TABLE AS EXPLAIN\n ALTER GROUP CREATE TRIGGER FETCH\n ALTER TABLE CREATE TYPE GRANT\n ALTER TRIGGER CREATE USER INSERT\n ALTER USER CREATE VIEW LISTEN\n ANALYZE DEALLOCATE LOAD\n BEGIN DECLARE LOCK\n CHECKPOINT DELETE MOVE\n CLOSE DROP AGGREGATE NOTIFY\n CLUSTER DROP CAST PREPARE\n COMMENT DROP CONVERSION REINDEX\n COMMIT DROP DATABASE RESET\n COPY DROP DOMAIN REVOKE\n CREATE AGGREGATE DROP FUNCTION ROLLBACK\n CREATE CAST DROP GROUP SELECT\n CREATE CONSTRAINT TRIGGER DROP INDEX SELECT INTO\n CREATE CONVERSION DROP LANGUAGE SET\n CREATE DATABASE DROP OPERATOR CLASS SET CONSTRAINTS\n CREATE DOMAIN DROP OPERATOR SET SESSION\nAUTHORIZATION\n CREATE FUNCTION DROP RULE SET TRANSACTION\n CREATE GROUP DROP SCHEMA SHOW\n CREATE INDEX DROP SEQUENCE START TRANSACTION\n CREATE LANGUAGE DROP TABLE TRUNCATE\n:\n\n\n-----Original Message-----\nFrom: Magnus Naeslund(f) [mailto:mag@fbab.net] \nSent: 16 December 2002 17:01\nTo: zahid rahman\nCc: pgsql-general@postgresql.org\nSubject: Re: [HACKERS] following instructions GCC\n\n(I forwarded the mail to the general list so that we keep onlist, next\ntime use 'reply to all' or similar...)\n\nzahid rahman <zahidr@btconnect.com> wrote:\n> Hi,\n>\n> Excellent!\n>\n> I have just downloaded postgresql.\n> I am in the process of downloading cygwin as per install file.\n> I follow instructions. I am a good boy!\n> I am a windows 2000 user. Ok ! I am not so good.\n>\n> My next instruction is that I will require a GCC compiler.\n> Where can I get so that I go to the sql prompt and run sql create\n> command\n> to create that elusive table without any hiccups.\n>\n\nI'm sorry to say that i don't know anything about using postgresql in\ncygwin.\nBut i think the safest bet is to use the cygwin install program to\ninstall postgresql.\nAfter that you just start postgresql as you normally would do i think.\n\nFor more info, see: http://www.se.postgresql.org/docs/faq-mswin.html\n\nMagnus\n\n", "msg_date": "Tue, 17 Dec 2002 02:31:48 -0800", "msg_from": "\"zahid rahman\" <zahidr@btconnect.com>", "msg_from_op": false, "msg_subject": "Re: Is that it ?" } ]
[ { "msg_contents": "\n\nI have the following problem\n\nI want to use using a function on result of subselect:\n\nI create the following function:\n\nhannu=# create or replace function pg_fields(pg_user) returns text as '\nhannu'# tup = args[0]\nhannu'# return tup[\"usename\"] + \":\" + str(tup[\"usesysid\"])\nhannu'# ' LANGUAGE 'plpython';\nCREATE FUNCTION\n\nAnd it runs fine straight on table/view:\n\nhannu=# select pg_fields(pg_user) from pg_user;\n pg_fields \n------------\n postgres:1\n hannu:100\n(2 rows)\n\n\nBut I am unable to run it on a subselect, whatever I do:\n\nhannu=# \nhannu=# select pg_fields(pg_user) from (select * from pg_user) as\npg_user;\nERROR: Cannot pass result of sub-select or join pg_user to a function\nhannu=# \nhannu=# select pg_fields(pg_user) from (select pg_user from pg_user) as\npg_user;\nERROR: You can't use relation names alone in the target list, try\nrelation.*.\nhannu=# select pg_fields(pg_user) from (select pg_user.* from pg_user)\nas pg_user;\nERROR: Cannot pass result of sub-select or join pg_user to a function\n\n\nI there a way to:\n\na) tell PostgreSQL that the funtion can take any row type as an argument\n\nor \n\nb) to cast the result of subquery to a known row type\n\n-- \nHannu Krosing <hannu@tm.ee>\n", "msg_date": "17 Dec 2002 13:07:29 +0500", "msg_from": "Hannu Krosing <hannu@tm.ee>", "msg_from_op": true, "msg_subject": "using a function on result of subselect" } ]
[ { "msg_contents": "hi\n postgresql supports dynamic sql with parameters in SQL function bodies,\n but not in interactive queries. why?\n\n when i wrote a dynamic sql with parameters, ODBC just filled the values of parameters into query string and sent it to server as a static query string.\ni think it's not right solution to dynamic sql with parameters.\n\nthanks\n\n\n \n \n \n\n\n", "msg_date": "Tue, 17 Dec 2002 16:10:40 +0800", "msg_from": "\"anthony sun\" <google_2000@163.com>", "msg_from_op": true, "msg_subject": "dynamic sql with parameters in interactive queries" } ]
[ { "msg_contents": "Update on replication:\n\nWe have several things happening with Postgres-R replication:\n\n\to Someone is porting the 7.2-based Postgres-R code to 7.3\n\to Darren and I are in discussion with the Spread folks,\n\t attempting to get a more BSD-friendly license from them\n\to People are evaluating the Postgres-R approach and comparing\n\t it to more traditional 2-phase commit replication.\n\nWith these things moving forward, we will be in a much better position\nto get synchronous replication integrated in PostgreSQL.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 12:56:45 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Update on replication" }, { "msg_contents": "> \to People are evaluating the Postgres-R approach and comparing\n> \t it to more traditional 2-phase commit replication.\n\nNot that the Postgres-R approach can replace 2-phase commit methods.\n\n2PC is still needed for support with external transaction managers (XA\ndrivers for JDBC).\n\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "17 Dec 2002 13:14:26 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "Rod Taylor wrote:\n-- Start of PGP signed section.\n> > \to People are evaluating the Postgres-R approach and comparing\n> > \t it to more traditional 2-phase commit replication.\n> \n> Not that the Postgres-R approach can replace 2-phase commit methods.\n> \n> 2PC is still needed for support with external transaction managers (XA\n> drivers for JDBC).\n\nYes, good point. Let me add that eventually, I think we will have:\n\n\tsync replication (Postgres-R?)\n\tasync replication, using PITR\n\tPITR\n\ttwo-phase commit for distributed transactions\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 13:17:23 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Update on replication" }, { "msg_contents": "On Tue, Dec 17, 2002 at 12:56:45PM -0500, Bruce Momjian wrote:\n> Update on replication:\n> \n> We have several things happening with Postgres-R replication:\n> \n> \to Someone is porting the 7.2-based Postgres-R code to 7.3\n\nYou mean 7.4devel?\n\n> With these things moving forward, we will be in a much better position\n> to get synchronous replication integrated in PostgreSQL.\n\nWhat about asynchronous (triggered?) replication? Is something like\nrserv or dbmirror going to be moved to main?\n\n-- \nAlvaro Herrera (<alvherre[a]dcc.uchile.cl>)\n\"Porque francamente, si para saber manejarse a uno mismo hubiera que\nrendir examen... �Qui�n es el machito que tendr�a carnet?\" (Mafalda)\n", "msg_date": "Tue, 17 Dec 2002 15:18:42 -0300", "msg_from": "Alvaro Herrera <alvherre@dcc.uchile.cl>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "Alvaro Herrera wrote:\n> On Tue, Dec 17, 2002 at 12:56:45PM -0500, Bruce Momjian wrote:\n> > Update on replication:\n> > \n> > We have several things happening with Postgres-R replication:\n> > \n> > \to Someone is porting the 7.2-based Postgres-R code to 7.3\n> \n> You mean 7.4devel?\n\nSorry, right.\n\n> > With these things moving forward, we will be in a much better position\n> > to get synchronous replication integrated in PostgreSQL.\n> \n> What about asynchronous (triggered?) replication? Is something like\n> rserv or dbmirror going to be moved to main?\n\nI think eventually we will have some async replication in the main\nserver, probably using PITR logs in some way.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 13:21:03 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Update on replication" }, { "msg_contents": "I just got my copy of SysAdmin Magazine and was surprised to see an \narticle about Usogres -- The PostgreSQL Replication Tool.\n\nI don't remember seeing it mentioned on this or the General list. Though\nI just started reading the article and don't have a firm grasp on it yet, \nI do remember a discussion of replication using this technique - described \nin the first two paragraphs.\n\n\nFyi,\nRod\n-- \n \"Open Source Software - Sometimes you get more than you paid for...\"\n\n", "msg_date": "Tue, 17 Dec 2002 11:10:19 -0800 (PST)", "msg_from": "\"Roderick A. Anderson\" <raanders@acm.org>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "Roderick A. Anderson wrote:\n> I just got my copy of SysAdmin Magazine and was surprised to see an \n> article about Usogres -- The PostgreSQL Replication Tool.\n> \n> I don't remember seeing it mentioned on this or the General list. Though\n> I just started reading the article and don't have a firm grasp on it yet, \n> I do remember a discussion of replication using this technique - described \n> in the first two paragraphs.\n\nI saw Usogres when I was first in Japan. Interesting in that is\nintercepts queries and passes them two different servers.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 14:16:19 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Update on replication" }, { "msg_contents": "On Tue, 17 Dec 2002, Alvaro Herrera wrote:\n\n> What about asynchronous (triggered?) replication? Is something like\n> rserv or dbmirror going to be moved to main?\n\n From what I've been able to tell *so far*, Postgres-R is going to preclude\nthe ability for either to work ... Vadim is currently reviewing the code,\nand based on his assessment of whether or not that is the case, I'm going\nto be pushing for postgres-r to be its own project/fork of PostgreSQL,\nlike RedHat Database ...\n\n\n\n", "msg_date": "Tue, 17 Dec 2002 16:06:38 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "Marc G. Fournier wrote:\n> On Tue, 17 Dec 2002, Alvaro Herrera wrote:\n> \n> > What about asynchronous (triggered?) replication? Is something like\n> > rserv or dbmirror going to be moved to main?\n> \n> From what I've been able to tell *so far*, Postgres-R is going to preclude\n> the ability for either to work ... Vadim is currently reviewing the code,\n> and based on his assessment of whether or not that is the case, I'm going\n> to be pushing for postgres-r to be its own project/fork of PostgreSQL,\n> like RedHat Database ...\n\nHow is Postgres-R going to prevent async replication from also being\nadopted in CVS?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 15:31:48 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Update on replication" }, { "msg_contents": "\"Marc G. Fournier\" <scrappy@hub.org> writes:\n> On Tue, 17 Dec 2002, Alvaro Herrera wrote:\n>> What about asynchronous (triggered?) replication? Is something like\n>> rserv or dbmirror going to be moved to main?\n\n>> From what I've been able to tell *so far*, Postgres-R is going to preclude\n> the ability for either to work ...\n\nWhy do you say that? If it can't coexist with other solutions, then it\nsurely will not be accepted, but I can't think of any reason why it\nwould preclude other approaches.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 17 Dec 2002 18:24:36 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Update on replication " }, { "msg_contents": "On Tue, 17 Dec 2002, Tom Lane wrote:\n\n> \"Marc G. Fournier\" <scrappy@hub.org> writes:\n> > On Tue, 17 Dec 2002, Alvaro Herrera wrote:\n> >> What about asynchronous (triggered?) replication? Is something like\n> >> rserv or dbmirror going to be moved to main?\n>\n> >> From what I've been able to tell *so far*, Postgres-R is going to preclude\n> > the ability for either to work ...\n>\n> Why do you say that? If it can't coexist with other solutions, then it\n> surely will not be accepted, but I can't think of any reason why it\n> would preclude other approaches.\n\nOkay, if this is the case, that does change things somewhat, but Bruce\nseems to indiate that co-existance will be a problem:\n\n\n========\n> On Tue, 17 Dec 2002, Bruce Momjian wrote:\n>\n> > > The other concern is how does integrating Postgres-R affect the ability to\n> > > investigate other solutions?\n> > >\n> > > As I said, I don't doubt taht there are aspects of Postgres-R that would\n> > > benefit the server as a whole, and those bits-n-pieces should be looked at\n> > > on an individual basis, but to just slap it in completely and hope that it\n> > > doesn't cause problems for alternative solutions is kinda irresponsible\n> > > ...\n> >\n> > It certainly will cause problems with other replication solutions.\n", "msg_date": "Tue, 17 Dec 2002 20:18:31 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: Update on replication " }, { "msg_contents": "\nWhen I said:\n\n> > > It certainly will cause problems with other replication solutions.\n\nI meant it would cause other solutions to be less desirable, meaning. as\nyou said, \"affect the ability to investigate other solutions?\" With a\nworking solution, others may be less likely to investigate because\nPostgres-R will be our official solution. (I believe that was your\nmajor negative point.) However, with the hooks already there, people\nmay be _more_ likely to investigate solutions, so there is really no way\nto know.\n\n---------------------------------------------------------------------------\n\nMarc G. Fournier wrote:\n> On Tue, 17 Dec 2002, Tom Lane wrote:\n> \n> > \"Marc G. Fournier\" <scrappy@hub.org> writes:\n> > > On Tue, 17 Dec 2002, Alvaro Herrera wrote:\n> > >> What about asynchronous (triggered?) replication? Is something like\n> > >> rserv or dbmirror going to be moved to main?\n> >\n> > >> From what I've been able to tell *so far*, Postgres-R is going to preclude\n> > > the ability for either to work ...\n> >\n> > Why do you say that? If it can't coexist with other solutions, then it\n> > surely will not be accepted, but I can't think of any reason why it\n> > would preclude other approaches.\n> \n> Okay, if this is the case, that does change things somewhat, but Bruce\n> seems to indiate that co-existance will be a problem:\n> \n> \n> ========\n> > On Tue, 17 Dec 2002, Bruce Momjian wrote:\n> >\n> > > > The other concern is how does integrating Postgres-R affect the ability to\n> > > > investigate other solutions?\n> > > >\n> > > > As I said, I don't doubt taht there are aspects of Postgres-R that would\n> > > > benefit the server as a whole, and those bits-n-pieces should be looked at\n> > > > on an individual basis, but to just slap it in completely and hope that it\n> > > > doesn't cause problems for alternative solutions is kinda irresponsible\n> > > > ...\n> > >\n> > > It certainly will cause problems with other replication solutions.\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 5: Have you checked our extensive FAQ?\n> \n> http://www.postgresql.org/users-lounge/docs/faq.html\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 19:27:37 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Update on replication" }, { "msg_contents": "On Tue, 17 Dec 2002, Bruce Momjian wrote:\n\n> I meant it would cause other solutions to be less desirable, meaning. as\n> you said, \"affect the ability to investigate other solutions?\" With a\n> working solution, others may be less likely to investigate because\n> Postgres-R will be our official solution. (I believe that was your\n> major negative point.) However, with the hooks already there, people\n> may be _more_ likely to investigate solutions, so there is really no way\n> to know.\n\nOkay, but if we are just adding hooks to allow Postgres-R to tie in, can't\nthose hooks be done in such a way as to be loadable, not compiled in?\n", "msg_date": "Tue, 17 Dec 2002 20:30:40 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "> Marc G. Fournier wrote:\n> > On Tue, 17 Dec 2002, Alvaro Herrera wrote:\n> > \n> > > What about asynchronous (triggered?) replication? Is something like\n> > > rserv or dbmirror going to be moved to main?\n> > \n> > From what I've been able to tell *so far*, Postgres-R is going to preclude\n> > the ability for either to work ... Vadim is currently reviewing the code,\n> > and based on his assessment of whether or not that is the case, I'm going\n> > to be pushing for postgres-r to be its own project/fork of PostgreSQL,\n> > like RedHat Database ...\n> \n> How is Postgres-R going to prevent async replication from also being\n> adopted in CVS?\n\nAs far as I know, all trigger based async replication solutions have a\nlimitation. They do not handle high load (and probably cannot handle\nlarge objects. am I correct?). I think we should move to other async\nreplication soltions, such as PostgreSQL-R or (not yet existing)\ntransaction log based replications.\n--\nTatsuo Ishii\n", "msg_date": "Wed, 18 Dec 2002 10:12:38 +0900 (JST)", "msg_from": "Tatsuo Ishii <t-ishii@sra.co.jp>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "> I just got my copy of SysAdmin Magazine and was surprised to see an \n> article about Usogres -- The PostgreSQL Replication Tool.\n> \n> I don't remember seeing it mentioned on this or the General list. Though\n> I just started reading the article and don't have a firm grasp on it yet, \n> I do remember a discussion of replication using this technique - described \n> in the first two paragraphs.\n\nGlad to hear that. Usogres was developed in Japan and pretty popular\nammong Japanese PostgreSQL community.\n\nBTW, there is a commercial product called QueryMaster, which takes\nsimilar approach to Usogres. It copies the input query and distribute\nto multiple PostgreSQL servers. As long as one of a server is working,\nusers even do not notice some of them are failing.\n--\nTatsuo Ishii\n", "msg_date": "Wed, 18 Dec 2002 10:13:06 +0900 (JST)", "msg_from": "Tatsuo Ishii <t-ishii@sra.co.jp>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "On Wed, 18 Dec 2002, Tatsuo Ishii wrote:\n\n> > I just got my copy of SysAdmin Magazine and was surprised to see an\n> > article about Usogres -- The PostgreSQL Replication Tool.\n> >\n> > I don't remember seeing it mentioned on this or the General list. Though\n> > I just started reading the article and don't have a firm grasp on it yet,\n> > I do remember a discussion of replication using this technique - described\n> > in the first two paragraphs.\n>\n> Glad to hear that. Usogres was developed in Japan and pretty popular\n> ammong Japanese PostgreSQL community.\n>\n> BTW, there is a commercial product called QueryMaster, which takes\n> similar approach to Usogres. It copies the input query and distribute\n> to multiple PostgreSQL servers. As long as one of a server is working,\n> users even do not notice some of them are failing.\n\nHow come these solutions are such well kept secrets? I've heard of\nneither in relation to past discussions about replication, or have I just\nmissed them? :(\n\n", "msg_date": "Tue, 17 Dec 2002 21:38:26 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "On Tue, 2002-12-17 at 19:38, Marc G. Fournier wrote:\n> On Wed, 18 Dec 2002, Tatsuo Ishii wrote:\n> \n> > > I just got my copy of SysAdmin Magazine and was surprised to see an\n> > > article about Usogres -- The PostgreSQL Replication Tool.\n> > >\n> > > I don't remember seeing it mentioned on this or the General list. Though\n> > > I just started reading the article and don't have a firm grasp on it yet,\n> > > I do remember a discussion of replication using this technique - described\n> > > in the first two paragraphs.\n> >\n> > Glad to hear that. Usogres was developed in Japan and pretty popular\n> > ammong Japanese PostgreSQL community.\n> >\n> > BTW, there is a commercial product called QueryMaster, which takes\n> > similar approach to Usogres. It copies the input query and distribute\n> > to multiple PostgreSQL servers. As long as one of a server is working,\n> > users even do not notice some of them are failing.\n> \n> How come these solutions are such well kept secrets? I've heard of\n> neither in relation to past discussions about replication, or have I just\n> missed them? :(\n> \n\nGood questions. I've heard of Usogres, more or less in passing, but\nheard it didn't work very well. Also heard that it wasn't reliable or a\nserious solution. I seem to remember hearing that it only worked on\nmuch older versions of PostgreSQL. Of course, I'm not attempting to\nassert any true to what I've heard but since it's being talking about,\nperhaps someone can clarify how well it REALLY works. Perhaps even\nprovide some more details on it?\n\nNever heard of QueryMaster. Perhaps someone would like to talk a little\nmore about that as well?\n\n\n\n\n-- \nGreg Copeland <greg@copelandconsulting.net>\nCopeland Computer Consulting\n\n", "msg_date": "17 Dec 2002 19:43:05 -0600", "msg_from": "Greg Copeland <greg@CopelandConsulting.Net>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "On Tue, 17 Dec 2002, Greg Copeland wrote:\n\n> On Tue, 2002-12-17 at 19:38, Marc G. Fournier wrote:\n> > On Wed, 18 Dec 2002, Tatsuo Ishii wrote:\n> >\n> > > > I just got my copy of SysAdmin Magazine and was surprised to see an\n> > > > article about Usogres -- The PostgreSQL Replication Tool.\n> > > >\n> > > > I don't remember seeing it mentioned on this or the General list. Though\n> > > > I just started reading the article and don't have a firm grasp on it yet,\n> > > > I do remember a discussion of replication using this technique - described\n> > > > in the first two paragraphs.\n> > >\n> > > Glad to hear that. Usogres was developed in Japan and pretty popular\n> > > ammong Japanese PostgreSQL community.\n> > >\n> > > BTW, there is a commercial product called QueryMaster, which takes\n> > > similar approach to Usogres. It copies the input query and distribute\n> > > to multiple PostgreSQL servers. As long as one of a server is working,\n> > > users even do not notice some of them are failing.\n> >\n> > How come these solutions are such well kept secrets? I've heard of\n> > neither in relation to past discussions about replication, or have I just\n> > missed them? :(\n> >\n>\n> Good questions. I've heard of Usogres, more or less in passing, but\n> heard it didn't work very well. Also heard that it wasn't reliable or a\n> serious solution. I seem to remember hearing that it only worked on\n> much older versions of PostgreSQL. Of course, I'm not attempting to\n> assert any true to what I've heard but since it's being talking about,\n> perhaps someone can clarify how well it REALLY works. Perhaps even\n> provide some more details on it?\n>\n> Never heard of QueryMaster. Perhaps someone would like to talk a little\n> more about that as well?\n\nJust checked on Usogres, and it appears to be relatively up to date, in\nthat it is known to work up to 7.2.1:\n\n\t\thttp://usogres.good-day.net/working.php3\n\nSearching Google for QueryMaster finds a few Japanese sites, but I can't\nread Japanese :(\n\nTatsuo, can you help?\n", "msg_date": "Tue, 17 Dec 2002 21:47:19 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "Marc G. Fournier wrote:\n> Just checked on Usogres, and it appears to be relatively up to date, in\n> that it is known to work up to 7.2.1:\n> \n> \t\thttp://usogres.good-day.net/working.php3\n> \n> Searching Google for QueryMaster finds a few Japanese sites, but I can't\n> read Japanese :(\n> \n> Tatsuo, can you help?\n\nUse Altavista:\n\n\thttp://world.altavista.com/\n\nThey have a Japense/English translation.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 20:49:05 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Update on replication" }, { "msg_contents": "> > Glad to hear that. Usogres was developed in Japan and pretty popular\n> > ammong Japanese PostgreSQL community.\n> >\n> > BTW, there is a commercial product called QueryMaster, which takes\n> > similar approach to Usogres. It copies the input query and distribute\n> > to multiple PostgreSQL servers. As long as one of a server is working,\n> > users even do not notice some of them are failing.\n> \n> How come these solutions are such well kept secrets? I've heard of\n> neither in relation to past discussions about replication, or have I just\n> missed them? :(\n\nAs far as I can remember, Usogres has appeared on this list for\nseveral times.\n\nAbout QueryMaster: it is born in US and is a fairly new\nproduct. Unfortunately I couldn't find web pages for it. I'll talk to\na person who is selling it in Japan if he could provide info in\nEnglish...\n--\nTatsuo Ishii\n", "msg_date": "Wed, 18 Dec 2002 10:58:42 +0900 (JST)", "msg_from": "Tatsuo Ishii <t-ishii@sra.co.jp>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "On Tue, Dec 17, 2002 at 07:43:05PM -0600, Greg Copeland wrote:\n> On Tue, 2002-12-17 at 19:38, Marc G. Fournier wrote:\n\n> > How come these solutions are such well kept secrets? I've heard of\n> > neither in relation to past discussions about replication, or have I just\n> > missed them? :(\n> \n> Good questions. I've heard of Usogres, more or less in passing, but\n> heard it didn't work very well.\n\nI'm sure this is because there little or no documentation of the system\nin english. I tried Usogres sometime ago, and felt really lost. I\nmanaged to install it and maybe even got it to work, but the\ndocumentation was so lacking that I didn't feel confident enough with\nit.\n\nI think the same can be said of QueryMaster, plus it isn't mentioned\nanywhere. How can I possibly search for documentation on a project if I\ndon't know it even exists?\n\nOne would be tempted to say that GBorg is a good place to list\nPostgreSQL related projects. Sadly, GBorg doesn't even have a\nconnection to the outside world. What value does it have if people\ndon't know it exists, or can't remember how to get to it? Let alone use\nit to find a project...\n\n-- \nAlvaro Herrera (<alvherre[a]dcc.uchile.cl>)\n\"Now I have my system running, not a byte was off the shelf;\nIt rarely breaks and when it does I fix the code myself.\nIt's stable, clean and elegant, and lightning fast as well,\nAnd it doesn't cost a nickel, so Bill Gates can go to hell.\"\n", "msg_date": "Tue, 17 Dec 2002 23:07:22 -0300", "msg_from": "Alvaro Herrera <alvherre@dcc.uchile.cl>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "On Wed, 18 Dec 2002, Tatsuo Ishii wrote:\n\n> > > Glad to hear that. Usogres was developed in Japan and pretty popular\n> > > ammong Japanese PostgreSQL community.\n> > >\n> > > BTW, there is a commercial product called QueryMaster, which takes\n> > > similar approach to Usogres. It copies the input query and distribute\n> > > to multiple PostgreSQL servers. As long as one of a server is working,\n> > > users even do not notice some of them are failing.\n> >\n> > How come these solutions are such well kept secrets? I've heard of\n> > neither in relation to past discussions about replication, or have I just\n> > missed them? :(\n>\n> As far as I can remember, Usogres has appeared on this list for\n> several times.\n>\n> About QueryMaster: it is born in US and is a fairly new\n> product. Unfortunately I couldn't find web pages for it. I'll talk to\n> a person who is selling it in Japan if he could provide info in\n> English...\n\n*rofl* I like that ... a US product, but all the pdf's and docs I can\nfind on it are in Japanese ... I think its almost time to learn a third\nlanguage :)\n\n\n", "msg_date": "Tue, 17 Dec 2002 22:09:37 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "On Tue, 2002-12-17 at 20:07, Alvaro Herrera wrote:\n> On Tue, Dec 17, 2002 at 07:43:05PM -0600, Greg Copeland wrote:\n> > On Tue, 2002-12-17 at 19:38, Marc G. Fournier wrote:\n> \n> > > How come these solutions are such well kept secrets? I've heard of\n> > > neither in relation to past discussions about replication, or have I just\n> > > missed them? :(\n> > \n> > Good questions. I've heard of Usogres, more or less in passing, but\n> > heard it didn't work very well.\n> \n> I'm sure this is because there little or no documentation of the system\n> in english. I tried Usogres sometime ago, and felt really lost. I\n> managed to install it and maybe even got it to work, but the\n> documentation was so lacking that I didn't feel confident enough with\n> it.\n> \n> I think the same can be said of QueryMaster, plus it isn't mentioned\n> anywhere. How can I possibly search for documentation on a project if I\n> don't know it even exists?\n> \n> One would be tempted to say that GBorg is a good place to list\n> PostgreSQL related projects. Sadly, GBorg doesn't even have a\n> connection to the outside world. What value does it have if people\n> don't know it exists, or can't remember how to get to it? Let alone use\n> it to find a project...\n\nThere are a couple of links to it from PostgreSQL's site, but you sorta\nhave to look and hunt around. I've requested higher visibility but for\nwhatever reason it seemed to be snubbed rather quickly.\n\nI do agree, GBorg needs MUCH higher visibility!\n\n\n-- \nGreg Copeland <greg@copelandconsulting.net>\nCopeland Computer Consulting\n\n", "msg_date": "17 Dec 2002 20:33:41 -0600", "msg_from": "Greg Copeland <greg@CopelandConsulting.Net>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "On Tue, 17 Dec 2002, Greg Copeland wrote:\n\n> There are a couple of links to it from PostgreSQL's site, but you sorta\n> have to look and hunt around. I've requested higher visibility but for\n> whatever reason it seemed to be snubbed rather quickly.\n>\n> I do agree, GBorg needs MUCH higher visibility!\n\nDave is pretty much ready to put teh portal in place\n(http://wwwdevel.postgresql.org) which has several links to GBorg right\noff the main page ...\n\n", "msg_date": "Tue, 17 Dec 2002 22:41:14 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "> There are a couple of links to it from PostgreSQL's site, but you sorta\n> have to look and hunt around. I've requested higher visibility but for\n> whatever reason it seemed to be snubbed rather quickly.\n>\n> I do agree, GBorg needs MUCH higher visibility!\n\nYes - I would love to move phpPgAdmin back to GBorg to attract some actual\ninterested developers, but I'm not sure it would...\n\nChris\n\n", "msg_date": "Wed, 18 Dec 2002 10:42:01 +0800", "msg_from": "\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "On Tue, 2002-12-17 at 21:33, Greg Copeland wrote:\n> I do agree, GBorg needs MUCH higher visibility!\n\nI'm just curious: why do we need GBorg at all? Does it offer anything\nthat SourceForge, or a similar service does not offer?\n\nEspecially given that (a) most other OSS projects don't have a site for\n\"related projects\" (unless you count something like CPAN, which is\ntotally different) (b) GBorg is completely unknown to anyone outside the\nPostgreSQL community and even to many people within it...\n\nCheers,\n\nNeil\n-- \nNeil Conway <neilc@samurai.com> || PGP Key ID: DB3C29FC\n\n\n\n", "msg_date": "17 Dec 2002 21:55:43 -0500", "msg_from": "Neil Conway <neilc@samurai.com>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "Neil Conway wrote:\n> On Tue, 2002-12-17 at 21:33, Greg Copeland wrote:\n> > I do agree, GBorg needs MUCH higher visibility!\n> \n> I'm just curious: why do we need GBorg at all? Does it offer anything\n> that SourceForge, or a similar service does not offer?\n> \n> Especially given that (a) most other OSS projects don't have a site for\n> \"related projects\" (unless you count something like CPAN, which is\n> totally different) (b) GBorg is completely unknown to anyone outside the\n> PostgreSQL community and even to many people within it...\n\nI think gborg allows us to collect all relivant projects in one place.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 22:00:36 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Update on replication" }, { "msg_contents": "On Tue, 2002-12-17 at 20:55, Neil Conway wrote:\n> On Tue, 2002-12-17 at 21:33, Greg Copeland wrote:\n> > I do agree, GBorg needs MUCH higher visibility!\n> \n> I'm just curious: why do we need GBorg at all? Does it offer anything\n> that SourceForge, or a similar service does not offer?\n> \n> Especially given that (a) most other OSS projects don't have a site for\n> \"related projects\" (unless you count something like CPAN, which is\n> totally different) (b) GBorg is completely unknown to anyone outside the\n> PostgreSQL community and even to many people within it...\n> \n\n\nPart I can answer, part I can not. Since I'm not the one that pushed\nthe projects to that site, I can't answer that part of the equation. \nAddressing the part of your question that I think I can, I do like the\nconcept of one-stop-shopping for all PostgreSQL needs. All Things\nProgreSQL is a pretty neat concept. Of course, it rather defeats the\nwhole purpose if no one, including potential developers, have no idea it\nexists.\n\n\n\n-- \nGreg Copeland <greg@copelandconsulting.net>\nCopeland Computer Consulting\n\n", "msg_date": "17 Dec 2002 21:08:48 -0600", "msg_from": "Greg Copeland <greg@CopelandConsulting.Net>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "On Tue, 2002-12-17 at 22:00, Bruce Momjian wrote:\n> I think gborg allows us to collect all relivant projects in one place.\n\nYes, but so would a webpage with a list of URLs, or a\nfreshmeat/google/dmoz directory, or an SF foundry, or [ any number of\nother mechanisms for collecting groups of related websites ].\n\n(Not to mention that you're assuming that GBorg includes \"all relevant\nprojects\" -- it doesn't do that now, nor is it likely to in the future.)\n\nCheers,\n\nNeil\n-- \nNeil Conway <neilc@samurai.com> || PGP Key ID: DB3C29FC\n\n\n\n", "msg_date": "17 Dec 2002 22:09:46 -0500", "msg_from": "Neil Conway <neilc@samurai.com>", "msg_from_op": false, "msg_subject": "Re: Update on replication" }, { "msg_contents": "Neil Conway wrote:\n> On Tue, 2002-12-17 at 22:00, Bruce Momjian wrote:\n> > I think gborg allows us to collect all relivant projects in one place.\n> \n> Yes, but so would a webpage with a list of URLs, or a\n> freshmeat/google/dmoz directory, or an SF foundry, or [ any number of\n> other mechanisms for collecting groups of related websites ].\n> \n> (Not to mention that you're assuming that GBorg includes \"all relevant\n> projects\" -- it doesn't do that now, nor is it likely to in the future.)\n\nGood point. I put my stuff on there because it was PostgreSQL-specific,\nand it had nice features.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 22:12:11 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: Update on replication" }, { "msg_contents": "The reason I favour a GBorg is that VA Linux (who own sourceforge) have yet\nto turn in a profit and so maylook to trim some of it's assets in order to\nimprove profitability at some point in the future.\n\nI think it would be a bad move to shift everything to sourceforge, only to\nfind that a year or more down the line the site dissappears/degrades to a\nlevel where it causes problems for the project, and loose the time we could\nhave spent building up the reputation of GBorg.\n\nAl.\n\n\n----- Original Message -----\nFrom: \"Neil Conway\" <neilc@samurai.com>\nTo: \"Greg Copeland\" <greg@CopelandConsulting.Net>\nCc: \"Alvaro Herrera\" <alvherre@dcc.uchile.cl>; \"Marc G. Fournier\"\n<scrappy@hub.org>; \"Tatsuo Ishii\" <t-ishii@sra.co.jp>; <raanders@acm.org>;\n\"PostgresSQL Hackers Mailing List\" <pgsql-hackers@postgresql.org>\nSent: Wednesday, December 18, 2002 2:55 AM\nSubject: [mail] Re: [HACKERS] Update on replication\n\n\n> On Tue, 2002-12-17 at 21:33, Greg Copeland wrote:\n> > I do agree, GBorg needs MUCH higher visibility!\n>\n> I'm just curious: why do we need GBorg at all? Does it offer anything\n> that SourceForge, or a similar service does not offer?\n>\n> Especially given that (a) most other OSS projects don't have a site for\n> \"related projects\" (unless you count something like CPAN, which is\n> totally different) (b) GBorg is completely unknown to anyone outside the\n> PostgreSQL community and even to many people within it...\n>\n> Cheers,\n>\n> Neil\n> --\n> Neil Conway <neilc@samurai.com> || PGP Key ID: DB3C29FC\n>\n>\n>\n>\n> ---------------------------(end of broadcast)---------------------------\n> TIP 4: Don't 'kill -9' the postmaster\n>\n\n\n", "msg_date": "Wed, 18 Dec 2002 08:03:35 -0000", "msg_from": "\"Al Sutton\" <al@alsutton.com>", "msg_from_op": false, "msg_subject": "Re: [mail] Re: Update on replication" }, { "msg_contents": "> I'm just curious: why do we need GBorg at all? Does it offer anything\n> that SourceForge, or a similar service does not offer?\n\nSAY NO TO SOURCEFORGE !\n\nPlease find enclosed some extracts from licensing terms \n(http://sourceforge.net/tos/tos.php) :\n\na) Acceptance of Terms\n\n\"We reserve the right, at our discretion, to change, modify, add or remove \nportions of these terms periodically. Such modifications shall be effective \nimmediately upon posting of the modified agreement to the website. Your \ncontinued use of the SourceForge.net website following the posting of changes \nto these terms and conditions will mean that you accept those changes.\"\n\n-> My point of view : SourceForge has an unlimited right to change the content \nof the TOS. A simpe post of the modified TOS is sufficient. For example, they \nmay at any time charge access to their web site.\n\nb) Licensing of code\n\n\"In each such case, the submitting user grants SourceForge.net the \nroyalty-free, perpetual, irrevocable, non-exclusive and fully sublicensable \nright and license to use, reproduce, modify, adapt, publish, translate, \ncreate derivative works from, distribute, perform and display such Content \n(in whole or part) worldwide and/or to incorporate it in other works in any \nform, media, or technology now known or later developed, all subject to the \nterms of any applicable approved license.\"\n\n-> My point of view : More surprisingly, SourceForge owns all content posted \non SourceForge. For legal reasons, SF licensing agreement is subject to the \nterms of any applicable approved license. But, the words *** or later \ndevelopment *** are thrilling.\n\nc) Termination\n\n\"We may terminate a user's account in our absolute discretion and for any \nreason. We are especially likely to terminate for reasons that include, but \nare not limited to, the following: 1.) violation of these Terms; 2.) abuse of \nsite resources or attempt to gain unauthorized entry to the site or site \nresources; 3.) use of Service in a manner inconsistent with the Purpose; 4.) \na user's request for such termination; and 5.) requirement of applicable law, \nregulation, court or governing agency order.\n\n-> My point of view : As a consequence, SourceForge does not allow a user to \nunregister from SourceForge. I tried to unregister a project from SF, without \nsuccess. Why? Because SF owns the (your) project rights.\n\nd) Incompatibility with local european laws\n\nThis contract does not comply with the European laws :\n\n- SF may change the licensing terms at any time, without limitation of the \nfuture modified clauses. In the European Union, you cannot grant an \n***unlimited right*** on your ***future and undefined*** work without a \n***defined counterpart*** (example=a salary).\n\n- In French law, an author right is devided in two separate rights : the owner \nright and the moral right. Every author (or community of authors) owns a \nmoral right on his/her work, even after the selling of rights. It has several \nconsequences which I wron't bother you with (even in RedHat attitude = the \nrenaming of a software is unmoral, bacause it makes you believe RedHat is the \noriginal author of PostgreSQL).\n\ne) SourceForge is a closed-source service\n\nSourceForge migrated (=rewrote) their database server-side code from \nPostgreSQL to Oracle, mainly for legal reasond. This new work gives them the \nability to change licences.\n\nAs a consequence, SF did not release their code for a year or so. SF can now \nbe considered as a closed-source service.\n\nf) I would discourage anyone from registering on SourceForge. Any organization \nis meant to be created, to live ... and die. Microsoft, Oracle and \nSourceForge will probably die sooner or later. \n\nOn the other hand, open-source project provider, like Savannah or better \nGBOrg, which are owned by non-profit organization or individuals releasing \ntheir source code under the GNU, will never die. This make a huge difference.\n\nNow, what if Microsoft purchased OSDN? What would be the consequences?\n\nCheers,\nJean-Michel\n\n", "msg_date": "Wed, 18 Dec 2002 11:17:56 +0100", "msg_from": "Jean-Michel POURE <jm.poure@freesurf.fr>", "msg_from_op": false, "msg_subject": "SourceForge policy on http://sourceforge.net/tos/tos.php" }, { "msg_contents": "Jean-Michel POURE <jm.poure@freesurf.fr> writes:\n\n> a) Acceptance of Terms\n> \n> \"We reserve the right, at our discretion, to change, modify, add or remove \n> portions of these terms periodically. Such modifications shall be effective \n> immediately upon posting of the modified agreement to the website. Your \n> continued use of the SourceForge.net website following the posting of changes \n> to these terms and conditions will mean that you accept those changes.\"\n> \n> -> My point of view : SourceForge has an unlimited right to change the content \n> of the TOS. A simpe post of the modified TOS is sufficient. For example, they \n> may at any time charge access to their web site.\n\nThis is standard legalities. They are a corporate entity; if there is\na gap that leaves them liable for someone else's actions, they have to\neither be able to update their TOS -- or they must stop providing the\nservice.\n\n> b) Licensing of code\n> \n> \"In each such case, the submitting user grants SourceForge.net the \n> royalty-free, perpetual, irrevocable, non-exclusive and fully sublicensable \n> right and license to use, reproduce, modify, adapt, publish, translate, \n> create derivative works from, distribute, perform and display such Content \n> (in whole or part) worldwide and/or to incorporate it in other works in any \n> form, media, or technology now known or later developed, all subject to the \n> terms of any applicable approved license.\"\n> \n> -> My point of view : More surprisingly, SourceForge owns all content posted \n> on SourceForge. For legal reasons, SF licensing agreement is subject to the \n> terms of any applicable approved license. But, the words *** or later \n> development *** are thrilling.\n\nThat is simply their legal way of saying \"You may only publish 'open\nsource' software and content on SourceForge\" -- which is made clear\nwhen you sign up for a new project anyway. If you want them to\ndistribute your software over the Internet, why do you care if they\nlater distribute it over Planetnet? (The \"in any ... technology\n... later developed\" does not scare me.) You must choose which\nlicense to use; if you do not want to use any of the approved\nlicenses, *then* you might not want to use SourceForge.\n\n> c) Termination\n> \n> \"We may terminate a user's account in our absolute discretion and for any \n> reason. We are especially likely to terminate for reasons that include, but \n> are not limited to, the following: 1.) violation of these Terms; 2.) abuse of \n> site resources or attempt to gain unauthorized entry to the site or site \n> resources; 3.) use of Service in a manner inconsistent with the Purpose; 4.) \n> a user's request for such termination; and 5.) requirement of applicable law, \n> regulation, court or governing agency order.\n> \n> -> My point of view : As a consequence, SourceForge does not allow a user to \n> unregister from SourceForge. I tried to unregister a project from SF, without \n> success. Why? Because SF owns the (your) project rights.\n\nYour project has nothing to do with their ability to terminate\naccounts. When I tried to unregister a project from SF, it took\nmonths, but they eventually did it. Maybe it helped that we were just\nrenaming the project to something else on SF and had already moved the\nCVS tree.\n\n> d) Incompatibility with local european laws\n> \n> This contract does not comply with the European laws :\n> \n> - SF may change the licensing terms at any time, without limitation of the \n> future modified clauses. In the European Union, you cannot grant an \n> ***unlimited right*** on your ***future and undefined*** work without a \n> ***defined counterpart*** (example=a salary).\n\nWhere does it say SF may change the licensing terms? If you do not\nwant them to change the licensing terms much, you can do something\nlike using GPL instead of BSD. Neither do they claim any right at all\non your future work, unless you send those works to them -- \"Licensing\nof code\" merely says that once you send them something, they have an\nunlimited right to use and redistribute it.\n\n> - In French law, an author right is devided in two separate rights : the owner \n> right and the moral right. Every author (or community of authors) owns a \n> moral right on his/her work, even after the selling of rights. It has several \n> consequences which I wron't bother you with (even in RedHat attitude = the \n> renaming of a software is unmoral, bacause it makes you believe RedHat is the \n> original author of PostgreSQL).\n\nThis is partially a license issue, and partially an international law\nissue, but it has little bearing on whether one uses SourceForge or\nnot; *any* user of the code can exercise the same rights. You say\nbelow that you prefer Savannah over SourceForge, but Savannah does\nlittle enough to acknowledge that their code is based on SourceForge's\ncode. Why criticize RedHat for doing the same thing?\n\n> e) SourceForge is a closed-source service\n> \n> SourceForge migrated (=rewrote) their database server-side code from \n> PostgreSQL to Oracle, mainly for legal reasond. This new work gives them the \n> ability to change licences.\n> \n> As a consequence, SF did not release their code for a year or so. SF can now \n> be considered as a closed-source service.\n\nLast time I checked, SourceForge 2.5 and before were still GPLed, and\nsome people do work to maintain it. However, use of a service is\ndifferent than licensing software. Google is closed source -- do you\nrefuse to use the services they provide?\n\n> f) I would discourage anyone from registering on SourceForge. Any organization \n> is meant to be created, to live ... and die. Microsoft, Oracle and \n> SourceForge will probably die sooner or later. \n\nMicrosoft, Oracle, SourceForge, and the FSF can all die, yes.\n\n> On the other hand, open-source project provider, like Savannah or better \n> GBOrg, which are owned by non-profit organization or individuals releasing \n> their source code under the GNU, will never die. This make a huge difference.\n\nWhy not? People get hit by busses. Organizations lose favor, or run\nlow on funding, and eventually go away too. Licensing and wide\ndistribution of the code is what guarantees it will never die, not the\nmass distributor(s).\n\n> Now, what if Microsoft purchased OSDN? What would be the consequences?\n\nWhat if Microsoft hired all of the Postgres committers? (Empty\nquestions can be easily turned around, so it is better to be specific\nin your concerns.)\n\nMichael Poole\n", "msg_date": "18 Dec 2002 08:36:47 -0500", "msg_from": "Michael Poole <poole@troilus.org>", "msg_from_op": false, "msg_subject": "Re: SourceForge policy on http://sourceforge.net/tos/tos.php" } ]
[ { "msg_contents": "I see a few malloc's in backend/access/transam/xlog.c that don't check\nto see if malloc returns NULL/failure. I think there should be at least\nan Assert() in there.\n\nAlso, seems we use malloc() a few other places where palloc should be\nused, like variable.c. Is that correct?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 13:30:21 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "malloc in xlog.c" }, { "msg_contents": "Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> I see a few malloc's in backend/access/transam/xlog.c that don't check\n> to see if malloc returns NULL/failure. I think there should be at least\n> an Assert() in there.\n\nIt'll dump core just fine without the help of an Assert ;-). I don't\nsee that an Assert adds much of anything.\n\nIn practice those mallocs are done during startup, so it seems very\nunlikely that they could fail, anyway.\n\n> Also, seems we use malloc() a few other places where palloc should be\n> used, like variable.c. Is that correct?\n\nThose are correct because of guc.c's conventions. Note the comments.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 17 Dec 2002 13:39:11 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: malloc in xlog.c " } ]
[ { "msg_contents": "I think the python interface in /interfaces/python should be moved to\ngborg. It already has its own web site:\n\n\thttp://www.druid.net/pygresql/\n\nand there is also another one, pyPgSQL, at:\n\n\thttp://pypgsql.sourceforge.net/\n\nIt would be good to get both of them listed in the gborg interfaces\nsection. They don't need to move their web sites there. They can just\nprovide a link to their main site from gborg.\n\nAlso, other interface authors should consider adding a page for\nthemselves on gborg too so there is one place people can look for\nPostgreSQL interfaces.\n\nI have CC'ed both python interface authors.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 14:10:38 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "python interface" }, { "msg_contents": "Bruce Momjian kirjutas K, 18.12.2002 kell 00:10:\n> I think the python interface in /interfaces/python should be moved to\n> gborg. It already has its own web site:\n> \n> \thttp://www.druid.net/pygresql/\n> \n> and there is also another one, pyPgSQL, at:\n> \n> \thttp://pypgsql.sourceforge.net/\n\nAnd the active third one, psycopg at:\n\n http://initd.org/software/initd/psycopg\n\nAnd an old, seemingly abandoned one at:\n\n http://www.advogato.org/proj/python-postgresql/\n\nBut it would be nice if there were still some rpm's built by default for\nthose, even from gborg.\n\n> It would be good to get both of them listed in the gborg interfaces\n> section. They don't need to move their web sites there. They can just\n> provide a link to their main site from gborg.\n> \n> Also, other interface authors should consider adding a page for\n> themselves on gborg too so there is one place people can look for\n> PostgreSQL interfaces.\n> \n> I have CC'ed both python interface authors.\n-- \nHannu Krosing <hannu@tm.ee>\n", "msg_date": "18 Dec 2002 01:40:36 +0500", "msg_from": "Hannu Krosing <hannu@tm.ee>", "msg_from_op": false, "msg_subject": "Re: python interface" }, { "msg_contents": "On Tuesday 17 December 2002 14:10, Bruce Momjian wrote:\n> I think the python interface in /interfaces/python should be moved to\n> gborg. It already has its own web site:\n>\n> \thttp://www.druid.net/pygresql/\n\nI would love to add PyGreSQL to gborg but no matter how often I log in it \ntells me that I need to be logged in to create a new project.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.\n", "msg_date": "Fri, 3 Jan 2003 08:05:55 -0500", "msg_from": "\"D'Arcy J.M. Cain\" <darcy@druid.net>", "msg_from_op": false, "msg_subject": "Re: python interface" }, { "msg_contents": "\"D'Arcy J.M. Cain\" <darcy@druid.net> writes:\n> I would love to add PyGreSQL to gborg but no matter how often I log in it \n> tells me that I need to be logged in to create a new project.\n\nWeird. Maybe you're blocking cookies, or something like that?\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 03 Jan 2003 15:24:32 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: python interface " }, { "msg_contents": "On Friday 03 January 2003 15:24, Tom Lane wrote:\n> \"D'Arcy J.M. Cain\" <darcy@druid.net> writes:\n> > I would love to add PyGreSQL to gborg but no matter how often I log in it\n> > tells me that I need to be logged in to create a new project.\n>\n> Weird. Maybe you're blocking cookies, or something like that?\n\nHmmm. I am using Opera and I have noticed cookie related strangeness before \nbut cookies do work. I use them with mailman all the time.\n\nI'll try another browser tonight.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.\n", "msg_date": "Fri, 3 Jan 2003 15:53:13 -0500", "msg_from": "\"D'Arcy J.M. Cain\" <darcy@druid.net>", "msg_from_op": false, "msg_subject": "Re: python interface" }, { "msg_contents": "\nLet me know how it goes, and what the project is ... that way we can move\nthe current CVS over so that we don't lose the extensive logging history\n...\n\n\n\nOn Fri, 3 Jan 2003, D'Arcy J.M. Cain wrote:\n\n> On Friday 03 January 2003 15:24, Tom Lane wrote:\n> > \"D'Arcy J.M. Cain\" <darcy@druid.net> writes:\n> > > I would love to add PyGreSQL to gborg but no matter how often I log in it\n> > > tells me that I need to be logged in to create a new project.\n> >\n> > Weird. Maybe you're blocking cookies, or something like that?\n>\n> Hmmm. I am using Opera and I have noticed cookie related strangeness before\n> but cookies do work. I use them with mailman all the time.\n>\n> I'll try another browser tonight.\n>\n> --\n> D'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\n> http://www.druid.net/darcy/ | and a sheep voting on\n> +1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.\n>\n> ---------------------------(end of broadcast)---------------------------\n> TIP 5: Have you checked our extensive FAQ?\n>\n> http://www.postgresql.org/users-lounge/docs/faq.html\n>\n", "msg_date": "Sat, 4 Jan 2003 20:58:45 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: python interface" }, { "msg_contents": "On Saturday 04 January 2003 19:58, Marc G. Fournier wrote:\n> Let me know how it goes, and what the project is ... that way we can move\n> the current CVS over so that we don't lose the extensive logging history\n\nHmm. I didn't think that it was suggested to move the repository again. I \nthought I was just going to add a pointer to the PyGreSQL page but leave the \ncode where it was. If we move it won't it make it harder to add it with just \na config option like we do now?\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.\n", "msg_date": "Wed, 8 Jan 2003 08:57:42 -0500", "msg_from": "\"D'Arcy J.M. Cain\" <darcy@druid.net>", "msg_from_op": false, "msg_subject": "Re: python interface" }, { "msg_contents": "D'Arcy J.M. Cain wrote:\n> On Saturday 04 January 2003 19:58, Marc G. Fournier wrote:\n> > Let me know how it goes, and what the project is ... that way we can move\n> > the current CVS over so that we don't lose the extensive logging history\n> \n> Hmm. I didn't think that it was suggested to move the repository again. I \n> thought I was just going to add a pointer to the PyGreSQL page but leave the \n> code where it was. If we move it won't it make it harder to add it with just \n> a config option like we do now?\n\nYes, it will make it harder, but we can release independently, and\nothers may get more involved in it. We have moved the CVS for most of\nthe interfaces, and they are doing fine on gborg.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 8 Jan 2003 10:16:55 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: python interface" } ]
[ { "msg_contents": "As suggested, I intend to create a 'CoerceToDomain' node in the\nexpression tree rather than attempting to apply the create the\nconstraints tests immediately.\n\n1. Create a new function in the executor to handle domain coercions:\nExecEvalCoerceToDomain()\n\n2. Move coerce_type_constraints to the executor (builds expression tree\nof constraints) under the name ExecCoerceTypeConstraints().\n\n3. On initial pass, CoerceToDomain will have a 'raw' expression tree\n(simple arg of data to coerce). After passing through\nExecCoerceTypeConstraints a 'cooked' expression tree will contain the\nconstraint tests. Subsequent tuples will simply use the pre-cooked\ntree.\n\n\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "17 Dec 2002 14:45:11 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": true, "msg_subject": "Coerce to Domain" }, { "msg_contents": "Rod Taylor <rbt@rbt.ca> writes:\n> 3. On initial pass, CoerceToDomain will have a 'raw' expression tree\n> (simple arg of data to coerce). After passing through\n> ExecCoerceTypeConstraints a 'cooked' expression tree will contain the\n> constraint tests.\n\nUh ... why? The cooked tree should be stored in pg_constraint, no?\nWhat's the point of redoing the parse analysis phase?\n\nBTW, this should be a lot easier to do cleanly (ie, without memory\nleaks) now that there's a notion of expression state trees. You can\nattach the information loaded from pg_constraint to the expression state\nnode for the CoerceToDomain node (being careful to copy it into\nestate->es_query_cxt). Maybe that was already apparent to you ...\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 17 Dec 2002 18:15:54 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Coerce to Domain " }, { "msg_contents": "On Tue, 2002-12-17 at 18:15, Tom Lane wrote:\n> Rod Taylor <rbt@rbt.ca> writes:\n> > 3. On initial pass, CoerceToDomain will have a 'raw' expression tree\n> > (simple arg of data to coerce). After passing through\n> > ExecCoerceTypeConstraints a 'cooked' expression tree will contain the\n> > constraint tests.\n> \n> Uh ... why? The cooked tree should be stored in pg_constraint, no?\n> What's the point of redoing the parse analysis phase?\n\nOk.. Cooked was the wrong word. The individual constraints will not be\nretested, but the group as a whole will be prepared (discovered?, looked\nup?).\n\n> node for the CoerceToDomain node (being careful to copy it into\n> estate->es_query_cxt). Maybe that was already apparent to you ...\n\nSomewhat apparent given your recent commits. I've not looked to find\nout exactly what it is yet, but I'm guessing recent changes to\nEvalPlanQual() will tell me.\n\n-- \nRod Taylor <rbt@rbt.ca>\n\nPGP Key: http://www.rbt.ca/rbtpub.asc", "msg_date": "17 Dec 2002 19:18:14 -0500", "msg_from": "Rod Taylor <rbt@rbt.ca>", "msg_from_op": true, "msg_subject": "Re: Coerce to Domain" }, { "msg_contents": "Rod Taylor <rbt@rbt.ca> writes:\n> Somewhat apparent given your recent commits. I've not looked to find\n> out exactly what it is yet, but I'm guessing recent changes to\n> EvalPlanQual() will tell me.\n\nDon't look at EvalPlanQual() ... you'll just get confused ;-).\nIt's a mess, and not related to constraints.\n\nThe code around init_fcache and ExecEvalFunc might be more enlightening\nfor your immediate purposes.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 17 Dec 2002 19:23:59 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: Coerce to Domain " } ]
[ { "msg_contents": "\nooops, sorry, that was what I meant (its been one of those days *grin*)\n\n\nOn Tue, 17 Dec 2002, Mikheev, Vadim wrote:\n\n> > From what I've been able to tell *so far*, Postgres-R is\n> > going to preclude the ability for either to work ...\n> > Vadim is currently reviewing the code,\n>\n> Not the code, just Darren' pdf (\"slide show\" -:()\n> and discussion in hackers' list.\n>\n>\n> _____________________________________________________\n> Sector Data, LLC, is not affiliated with Sector, Inc., or SIAC\n>\n", "msg_date": "Tue, 17 Dec 2002 16:45:23 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": true, "msg_subject": "Re: Update on replication" } ]
[ { "msg_contents": "\n\n> Not the code, just Darren' pdf (\"slide show\" -:()\n> and discussion in hackers' list.\n> \n\n\nYou might want to read this paper. Its not very long, and\nwill give you much more insite on the postgres-r work.\n\nhttp://www.cs.mcgill.ca/~kemme/papers/vldb00.html\n\nDarren\n\n", "msg_date": "Tue, 17 Dec 2002 16:02:31 -0500", "msg_from": "<darren@up.hrcoxmail.com>", "msg_from_op": true, "msg_subject": "Re: Update on replication" } ]
[ { "msg_contents": "Folks,\n\nAs you may know, we don't check the tgenabled status of deferred\ntriggers (e.g. AFTER triggers) -- they are added to the deferred queue\nand executed regardless.\n\nThis is pretty clearly a bug, but when it's been mentioned before, there\nwere three possible fixes suggested:\n\n(1) check tgenabled when the trigger is fired (and we're considering\nadding it to the deferred event queue)\n\n(2) check tgenabled when the deferred event is activated\n\n(3) check tgenabled at both times -- if #1 AND #2 is true, actually\ninvoke the deferred trigger\n\nI think we should implement (1), for the following reasons:\n\n- it's simpler: KISS, if nothing else. Adding it to the deferred queue\nand then delaying the \"is it enabled?\" decision to a later point in time\nadds complexity for no gain in useful functionality.\n\n- less functional: in theory, someone could twiddle the tgenabled bit\nfor individual firings of the trigger, to only add some of the firings\nof the trigger to the event queue. At the least with the current\nimplementation of deferred triggers, this wouldn't be possible AFAICS.\n\nAny comments?\n\nIf no has a problem with (1), I'll send in a patch implementing it to\n-patches.\n\nCheers,\n\nNeil\n-- \nNeil Conway <neilc@samurai.com> || PGP Key ID: DB3C29FC\n\n\n\n", "msg_date": "17 Dec 2002 16:19:25 -0500", "msg_from": "Neil Conway <neilc@samurai.com>", "msg_from_op": true, "msg_subject": "disabled, deferred triggers" } ]
[ { "msg_contents": "I have prepared the 7.3 CVS branch in preparation of a 7.3.1 release\nsoon. Please check it.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 21:17:24 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "7.3.1 stamped" }, { "msg_contents": "Bruce Momjian writes:\n\n> I have prepared the 7.3 CVS branch in preparation of a 7.3.1 release\n> soon. Please check it.\n\nIt would be of advantage if it were announced to the development group\nahead of time when a minor release is planned, so work can be planned\nbetter. It is certainly extremely clumsy if the so-called release already\nsits on the FTP server while some of us are stilling trying to fix things\nup.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Thu, 19 Dec 2002 00:39:05 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 stamped" } ]
[ { "msg_contents": "I have not been aggressive about backpatching documentation improvements\ninto 7.3.1. Is that something I should check?\n\nAs I remember, we didn't update the official docs for minor releases. \nIs that still true?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 21:25:11 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "7.3.1 documentation updates" }, { "msg_contents": "On Tue, 17 Dec 2002, Bruce Momjian wrote:\n\n> I have not been aggressive about backpatching documentation improvements\n> into 7.3.1. Is that something I should check?\n>\n> As I remember, we didn't update the official docs for minor releases.\n> Is that still true?\n\nThey say hydergine helps the memory. The idocs were never updated\nto current version until around the .1 release since there were usually\ndiscrepencies and it was a pain to update them.\n\nVince.\n-- \n Fast, inexpensive internet service 56k and beyond! http://www.pop4.net/\n http://www.meanstreamradio.com http://www.unknown-artists.com\n Internet radio: It's not file sharing, it's just radio.\n\n", "msg_date": "Tue, 17 Dec 2002 21:34:24 -0500 (EST)", "msg_from": "Vince Vielhaber <vev@michvhf.com>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 documentation updates" }, { "msg_contents": "Peter Eisentraut wrote:\n> Bruce Momjian writes:\n> \n> > I have not been aggressive about backpatching documentation improvements\n> > into 7.3.1. Is that something I should check?\n> \n> Probably in the future it would be good to put a minimal amount of work\n> in polishing the documentation in the stable branch.\n> \n> > As I remember, we didn't update the official docs for minor releases.\n> > Is that still true?\n> \n> It never was.\n\nI meant we don't create new doc tarballs for minor releases. Is that\nstill true?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 18 Dec 2002 19:00:22 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": true, "msg_subject": "Re: 7.3.1 documentation updates" }, { "msg_contents": "Bruce Momjian writes:\n\n> I have not been aggressive about backpatching documentation improvements\n> into 7.3.1. Is that something I should check?\n\nProbably in the future it would be good to put a minimal amount of work\nin polishing the documentation in the stable branch.\n\n> As I remember, we didn't update the official docs for minor releases.\n> Is that still true?\n\nIt never was.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Thu, 19 Dec 2002 01:06:15 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 documentation updates" }, { "msg_contents": "Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> I meant we don't create new doc tarballs for minor releases. Is that\n> still true?\n\nIf someone's willing to do the work of regenerating the doc tarballs,\nI think it would be a real good idea to update them for minor releases.\nPeter, are you volunteering for that?\n\n\nI will take the blame for the lack of advance notice of this release.\ncore had discussed it over the weekend, and agreed at my suggestion to\nplan a Wednesday release. We should have let -hackers know of the plan,\nand in fact I intended to do that but forgot. Mea culpa.\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 18 Dec 2002 19:20:12 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 documentation updates " }, { "msg_contents": "Tom Lane writes:\n\n> If someone's willing to do the work of regenerating the doc tarballs,\n> I think it would be a real good idea to update them for minor releases.\n\nThey are being updated, they have been for quite a while, and there's no\none volunteering because it's done rather automatically. (The 7.3.1\n\"preview\" from the other day contained freshly built documentation, and I\ndidn't have anything to do with it.)\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Fri, 20 Dec 2002 00:45:24 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 documentation updates " } ]
[ { "msg_contents": "Could you put a note in HISTORY about the incompatability with pre-7.3\nSSL clients?\n\n --Nate\n", "msg_date": "Tue, 17 Dec 2002 20:42:31 -0600 (CST)", "msg_from": "Nathan Mueller <nmueller@cs.wisc.edu>", "msg_from_op": true, "msg_subject": "Re: 7.3.1 stamped" }, { "msg_contents": "Nathan Mueller wrote:\n> Could you put a note in HISTORY about the incompatability with pre-7.3\n> SSL clients?\n\nI believe that pre7-3 SSL clients will work in 7.3.1, or am I wrong?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 21:50:28 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 stamped" } ]
[ { "msg_contents": "> I believe that pre7-3 SSL clients will work in 7.3.1, or am I wrong?\n\nIn 7.3 the SSL protocol switched from SSLv2 to TLSv1. If the server\nmethod is switched to SSLv23_method it will be backwords compatable with\npre-7.3 clients without sacrificing the added security of TLSv1 for\nnewer stuff. There's been a lot of other changes to the SSL code between\n7.2 and 7.3, but I've tested this out and haven't found any problems.\nI've included a patch to src/backend/libpq/be-secure.c if you're\ninterested.\n\n --Nate\n\n--- be-secure.c 13 Dec 2002 18:06:44 -0000 1.3\n+++ be-secure.c 18 Dec 2002 04:16:19 -0000\n@@ -587,7 +587,7 @@\n {\n SSL_library_init();\n SSL_load_error_strings();\n- SSL_context = SSL_CTX_new(TLSv1_method());\n+ SSL_context = SSL_CTX_new(SSLv23_method());\n if (!SSL_context)\n {\n postmaster_error(\"failed to create SSL\n context: %s\",\n", "msg_date": "Tue, 17 Dec 2002 22:18:38 -0600 (CST)", "msg_from": "Nathan Mueller <nmueller@cs.wisc.edu>", "msg_from_op": true, "msg_subject": "Re: 7.3.1 stamped" }, { "msg_contents": "\nI am confused. How can we switch back to SSLv23_method and still be\ncompatible with TLSv1_method. Does SSLv23_method support both?\n\nThe SSL author didn't like SSLv23_method (especially SSLv2) and I am not\nconfident to question his decision. We will just have to break backward\ncompatibility with pre-7.3 clients. No one else has mentioned it as a\nproblem, and in fact most have probably already upgraded to 7.3, so we\nshould be OK.\n\n---------------------------------------------------------------------------\n\nNathan Mueller wrote:\n> > I believe that pre7-3 SSL clients will work in 7.3.1, or am I wrong?\n> \n> In 7.3 the SSL protocol switched from SSLv2 to TLSv1. If the server\n> method is switched to SSLv23_method it will be backwords compatable with\n> pre-7.3 clients without sacrificing the added security of TLSv1 for\n> newer stuff. There's been a lot of other changes to the SSL code between\n> 7.2 and 7.3, but I've tested this out and haven't found any problems.\n> I've included a patch to src/backend/libpq/be-secure.c if you're\n> interested.\n> \n> --Nate\n> \n> --- be-secure.c 13 Dec 2002 18:06:44 -0000 1.3\n> +++ be-secure.c 18 Dec 2002 04:16:19 -0000\n> @@ -587,7 +587,7 @@\n> {\n> SSL_library_init();\n> SSL_load_error_strings();\n> - SSL_context = SSL_CTX_new(TLSv1_method());\n> + SSL_context = SSL_CTX_new(SSLv23_method());\n> if (!SSL_context)\n> {\n> postmaster_error(\"failed to create SSL\n> context: %s\",\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 23:22:04 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 stamped" } ]
[ { "msg_contents": "> I am confused. How can we switch back to SSLv23_method and still be\n> compatible with TLSv1_method. Does SSLv23_method support both?\n\nSSLv23 understands SSLv2, SSLv3 and TLSv1. When used in a client it uses\nSSLv2 but tells the server it can understand the other ones too. Check\nout the SSL_CTX_new manpage for a lot more details.\n\n> The SSL author didn't like SSLv23_method (especially SSLv2) and\n> I am not\n> confident to question his decision. We will just have to break\n> backward\n> compatibility with pre-7.3 clients. No one else has mentioned it as a\n> problem, and in fact most have probably already upgraded to 7.3, so we\n> should be OK.\n\nI agree, TLSv1 is a lot better but there's no point in breaking\nbackwords compatibility when you don't have to. Also, given my problems\nwith 7.3's SSL I'd be surprised if a lot of people who use it have made\nthe switch.\n\n --Nate\n", "msg_date": "Tue, 17 Dec 2002 22:27:13 -0600 (CST)", "msg_from": "Nathan Mueller <nmueller@cs.wisc.edu>", "msg_from_op": true, "msg_subject": "Re: 7.3.1 stamped" }, { "msg_contents": "Nathan Mueller wrote:\n> > I am confused. How can we switch back to SSLv23_method and still be\n> > compatible with TLSv1_method. Does SSLv23_method support both?\n> \n> SSLv23 understands SSLv2, SSLv3 and TLSv1. When used in a client it uses\n> SSLv2 but tells the server it can understand the other ones too. Check\n> out the SSL_CTX_new manpage for a lot more details.\n> \n> > The SSL author didn't like SSLv23_method (especially SSLv2) and\n> > I am not\n> > confident to question his decision. We will just have to break\n> > backward\n> > compatibility with pre-7.3 clients. No one else has mentioned it as a\n> > problem, and in fact most have probably already upgraded to 7.3, so we\n> > should be OK.\n> \n> I agree, TLSv1 is a lot better but there's no point in breaking\n> backwords compatibility when you don't have to. Also, given my problems\n> with 7.3's SSL I'd be surprised if a lot of people who use it have made\n> the switch.\n\nWell, we break backward compatibility so people can't use SSL2 to\nconnect to the server. Backward compatibility to a broken protocol\nisn't what I would call secure. Is that accurate?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 23:29:10 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 stamped" } ]
[ { "msg_contents": "> Well, we break backward compatibility so people can't use SSL2 to\n> connect to the server. Backward compatibility to a broken protocol\n> isn't what I would call secure. Is that accurate?\n\nI suppose. As long as the incompatibilty is mentioned in HISTORY I'm\nfine.\n\n --Nate\n", "msg_date": "Tue, 17 Dec 2002 22:48:40 -0600 (CST)", "msg_from": "Nathan Mueller <nmueller@cs.wisc.edu>", "msg_from_op": true, "msg_subject": "Re: 7.3.1 stamped" }, { "msg_contents": "Nathan Mueller wrote:\n> > Well, we break backward compatibility so people can't use SSL2 to\n> > connect to the server. Backward compatibility to a broken protocol\n> > isn't what I would call secure. Is that accurate?\n> \n> I suppose. As long as the incompatibilty is mentioned in HISTORY I'm\n> fine.\n\nUh, I didn't get it in there because we were still discussing it while I\nwas packaging 7.3.1. Sorry. I can mention it in 7.3.2. \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Tue, 17 Dec 2002 23:50:09 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 stamped" }, { "msg_contents": "On Tue, 17 Dec 2002, Nathan Mueller wrote:\n\n> > Well, we break backward compatibility so people can't use SSL2 to\n> > connect to the server. Backward compatibility to a broken protocol\n> > isn't what I would call secure. Is that accurate?\n>\n> I suppose. As long as the incompatibilty is mentioned in HISTORY I'm\n> fine.\n\nI read the SSL_CTX_new man page, and they recommend using SSLv23_method to\nprovide backwards compatibility ... if someone doesn't wan tto use SSL2,\nthey have the option to use TLS, but we shouldn't be forcigin them to use\none or the othe r...\n\nI have made the change and am just building v7.3.1 right now ... should be\navailable in a few minutes, and I'll announce it this evening as being\navailable ... can you grab a copy and make sure that it works as expected?\n", "msg_date": "Wed, 18 Dec 2002 09:37:15 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 stamped" }, { "msg_contents": "Marc G. Fournier wrote:\n> On Tue, 17 Dec 2002, Nathan Mueller wrote:\n> \n> > > Well, we break backward compatibility so people can't use SSL2 to\n> > > connect to the server. Backward compatibility to a broken protocol\n> > > isn't what I would call secure. Is that accurate?\n> >\n> > I suppose. As long as the incompatibilty is mentioned in HISTORY I'm\n> > fine.\n> \n> I read the SSL_CTX_new man page, and they recommend using SSLv23_method to\n> provide backwards compatibility ... if someone doesn't wan tto use SSL2,\n> they have the option to use TLS, but we shouldn't be forcigin them to use\n> one or the othe r...\n> \n> I have made the change and am just building v7.3.1 right now ... should be\n> available in a few minutes, and I'll announce it this evening as being\n> available ... can you grab a copy and make sure that it works as expected?\n\nOK, I see from your commit message:\n\n From the SSL_CTX_new man page:\n\n \"SSLv23_method(void), SSLv23_server_method(void), SSLv23_client_method(void)\n\n A TLS/SSL connection established with these methods will understand the SSLv2,\n SSLv3, and TLSv1 protocol. A client will send out SSLv2 client hello messages\n and will indicate that it also understands SSLv3 and TLSv1. A server will\n understand SSLv2, SSLv3, and TLSv1 client hello messages. This is the best\n choice when compatibility is a concern.\"\n\n This will maintain backwards compatibility for those us that don't use\n TLS connections ...\n\nMy question is whether it is safe to be making this change in a minor\nrelease? Does it work with 7.3 to 7.3.1 combinations? My other\nquestion is, if SSLv2 isn't secure, couldn't a client say they only\nsupport SSLv2, and hence break into the server? That was my original\nhesitancy, that and the fact Bear the SSL guy didn't want it.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 18 Dec 2002 14:43:09 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 stamped" }, { "msg_contents": "On Wed, 18 Dec 2002, Bruce Momjian wrote:\n\n> Marc G. Fournier wrote:\n> > On Tue, 17 Dec 2002, Nathan Mueller wrote:\n> >\n> > > > Well, we break backward compatibility so people can't use SSL2 to\n> > > > connect to the server. Backward compatibility to a broken protocol\n> > > > isn't what I would call secure. Is that accurate?\n> > >\n> > > I suppose. As long as the incompatibilty is mentioned in HISTORY I'm\n> > > fine.\n> >\n> > I read the SSL_CTX_new man page, and they recommend using SSLv23_method to\n> > provide backwards compatibility ... if someone doesn't wan tto use SSL2,\n> > they have the option to use TLS, but we shouldn't be forcigin them to use\n> > one or the othe r...\n> >\n> > I have made the change and am just building v7.3.1 right now ... should be\n> > available in a few minutes, and I'll announce it this evening as being\n> > available ... can you grab a copy and make sure that it works as expected?\n>\n> OK, I see from your commit message:\n>\n> From the SSL_CTX_new man page:\n>\n> \"SSLv23_method(void), SSLv23_server_method(void), SSLv23_client_method(void)\n>\n> A TLS/SSL connection established with these methods will understand the SSLv2,\n> SSLv3, and TLSv1 protocol. A client will send out SSLv2 client hello messages\n> and will indicate that it also understands SSLv3 and TLSv1. A server will\n> understand SSLv2, SSLv3, and TLSv1 client hello messages. This is the best\n> choice when compatibility is a concern.\"\n>\n> This will maintain backwards compatibility for those us that don't use\n> TLS connections ...\n>\n> My question is whether it is safe to be making this change in a minor\n> release? Does it work with 7.3 to 7.3.1 combinations? My other\n> question is, if SSLv2 isn't secure, couldn't a client say they only\n> support SSLv2, and hence break into the server? That was my original\n> hesitancy, that and the fact Bear the SSL guy didn't want it.\n\nWow, which part of \"A TLS/SSL connection established with these methods\nwill understand the SSLv2, SSLv3, and TLSv1 protocol\" are you finiding\nparticularly confusing? As nate explained to you, and the man page\nsection I commited states, TLSv1_method *only* supports TLS connections\n... SSLv23_method supports SSLv2, v3 and TLSv1 ...\n\nAs for 'break into the server\" ... ummm ... isn't that what pg_hba.conf is\nfor? I don't know about servers you run, but I don't let just anyone\nconnect to my server, and, in fact, close down the databases themsleves to\nspecific users ... if you don't trust the client, why are you giving him\naccss to your data, regardless of the protocol being used to encrypt the\nsessino??\n\n", "msg_date": "Wed, 18 Dec 2002 15:55:21 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 stamped" }, { "msg_contents": "Marc G. Fournier wrote:\n> > My question is whether it is safe to be making this change in a minor\n> > release? Does it work with 7.3 to 7.3.1 combinations? My other\n> > question is, if SSLv2 isn't secure, couldn't a client say they only\n> > support SSLv2, and hence break into the server? That was my original\n> > hesitancy, that and the fact Bear the SSL guy didn't want it.\n> \n> Wow, which part of \"A TLS/SSL connection established with these methods\n> will understand the SSLv2, SSLv3, and TLSv1 protocol\" are you finiding\n> particularly confusing? As nate explained to you, and the man page\n> section I commited states, TLSv1_method *only* supports TLS connections\n> ... SSLv23_method supports SSLv2, v3 and TLSv1 ...\n> \n> As for 'break into the server\" ... ummm ... isn't that what pg_hba.conf is\n> for? I don't know about servers you run, but I don't let just anyone\n> connect to my server, and, in fact, close down the databases themsleves to\n> specific users ... if you don't trust the client, why are you giving him\n> accss to your data, regardless of the protocol being used to encrypt the\n> sessino??\n\nI wasn't sure how insecure SSL2 was, and whether it allowed you to\nauthenticate without a password or something.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 18 Dec 2002 14:58:20 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 stamped" }, { "msg_contents": "On Wed, 18 Dec 2002, Marc G. Fournier wrote:\n\n> On Wed, 18 Dec 2002, Bruce Momjian wrote:\n> \n> > OK, I see from your commit message:\n> >\n> > From the SSL_CTX_new man page:\n> >\n> > \"SSLv23_method(void), SSLv23_server_method(void), SSLv23_client_method(void)\n> >\n> > A TLS/SSL connection established with these methods will understand the SSLv2,\n> > SSLv3, and TLSv1 protocol. A client will send out SSLv2 client hello messages\n> > and will indicate that it also understands SSLv3 and TLSv1. A server will\n> > understand SSLv2, SSLv3, and TLSv1 client hello messages. This is the best\n> > choice when compatibility is a concern.\"\n> >\n> > This will maintain backwards compatibility for those us that don't use\n> > TLS connections ...\n> >\n> > My question is whether it is safe to be making this change in a minor\n> > release? Does it work with 7.3 to 7.3.1 combinations? My other\n> > question is, if SSLv2 isn't secure, couldn't a client say they only\n> > support SSLv2, and hence break into the server? That was my original\n> > hesitancy, that and the fact Bear the SSL guy didn't want it.\n> \n> Wow, which part of \"A TLS/SSL connection established with these methods\n> will understand the SSLv2, SSLv3, and TLSv1 protocol\" are you finiding\n> particularly confusing? As nate explained to you, and the man page\n> section I commited states, TLSv1_method *only* supports TLS connections\n> ... SSLv23_method supports SSLv2, v3 and TLSv1 ...\n> \n> As for 'break into the server\" ... ummm ... isn't that what pg_hba.conf is\n> for? I don't know about servers you run, but I don't let just anyone\n> connect to my server, and, in fact, close down the databases themsleves to\n> specific users ... if you don't trust the client, why are you giving him\n> accss to your data, regardless of the protocol being used to encrypt the\n> sessino??\n\nBut, insecure SSL allows for \"man in the middle\" type of attacks. I.e. \nsomeone can sniff your secure (?) connection and get the password out of \nit, then spoof your IP and get in. The REASON for including TLS/SSL was \nto give people the ability to connect in a secure method so that IF \nsomeone is trying to listen in, they can't grab your name/password or \nyour data. \n\nAllowing SSL connects means that that could happen. Disallowing them \ninconveniences the user. My suggestion would be to implement another GUC \nthat by default turns off the insecure connections, and has to be \nuncommented and changed by the dba to allow the server to serve the \ninsecure SSL method. Best of both worlds.\n\n", "msg_date": "Wed, 18 Dec 2002 13:23:05 -0700 (MST)", "msg_from": "\"scott.marlowe\" <scott.marlowe@ihs.com>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 stamped" }, { "msg_contents": "> > Wow, which part of \"A TLS/SSL connection established with these methods\n> > will understand the SSLv2, SSLv3, and TLSv1 protocol\" are you finiding\n> > particularly confusing? As nate explained to you, and the man page\n> > section I commited states, TLSv1_method *only* supports TLS connections\n> > ... SSLv23_method supports SSLv2, v3 and TLSv1 ...\n> > \n> > As for 'break into the server\" ... ummm ... isn't that what pg_hba.conf is\n> > for? I don't know about servers you run, but I don't let just anyone\n> > connect to my server, and, in fact, close down the databases themsleves to\n> > specific users ... if you don't trust the client, why are you giving him\n> > accss to your data, regardless of the protocol being used to encrypt the\n> > sessino??\n> \n> But, insecure SSL allows for \"man in the middle\" type of attacks. I.e. \n> someone can sniff your secure (?) connection and get the password out of \n> it, then spoof your IP and get in. The REASON for including TLS/SSL was \n> to give people the ability to connect in a secure method so that IF \n> someone is trying to listen in, they can't grab your name/password or \n> your data. \n> \n> Allowing SSL connects means that that could happen. Disallowing them \n> inconveniences the user. My suggestion would be to implement another GUC \n> that by default turns off the insecure connections, and has to be \n> uncommented and changed by the dba to allow the server to serve the \n> insecure SSL method. Best of both worlds.\n\nAt this point, all the SSL2 problems are conjecture on my part, which I\ndon't understand. I hesitate to do anything until someone really\nknowledgeable can comment. Re-enabling SSL2 as part of 7.3.1 makes\nsense until we can get a definative answer on the risks involved.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 18 Dec 2002 15:30:36 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 stamped" }, { "msg_contents": "On Wed, 18 Dec 2002, Bruce Momjian wrote:\n\n> Marc G. Fournier wrote:\n> > > My question is whether it is safe to be making this change in a minor\n> > > release? Does it work with 7.3 to 7.3.1 combinations? My other\n> > > question is, if SSLv2 isn't secure, couldn't a client say they only\n> > > support SSLv2, and hence break into the server? That was my original\n> > > hesitancy, that and the fact Bear the SSL guy didn't want it.\n> > \n> > Wow, which part of \"A TLS/SSL connection established with these methods\n> > will understand the SSLv2, SSLv3, and TLSv1 protocol\" are you finiding\n> > particularly confusing? As nate explained to you, and the man page\n> > section I commited states, TLSv1_method *only* supports TLS connections\n> > ... SSLv23_method supports SSLv2, v3 and TLSv1 ...\n> > \n> > As for 'break into the server\" ... ummm ... isn't that what pg_hba.conf is\n> > for? I don't know about servers you run, but I don't let just anyone\n> > connect to my server, and, in fact, close down the databases themsleves to\n> > specific users ... if you don't trust the client, why are you giving him\n> > accss to your data, regardless of the protocol being used to encrypt the\n> > sessino??\n> \n> I wasn't sure how insecure SSL2 was, and whether it allowed you to\n> authenticate without a password or something.\n\nSSL2 seems to get a lot of votes for being broken in ways that cannot be \nfixed because they aren't simple buffer overflows. see:\n\nhttp://www.lne.com/ericm/papers/ssl_servers.html#1.2\n\nMy suggestion would be to eventually phase out ssl2 in favor of ssl3 or \ntls. And, as we are phasing it out, make it an opt-in thing, where the \ndba has to turn on ssl2 KNOWING he is turning on a flawed protocol.\n\n", "msg_date": "Wed, 18 Dec 2002 13:46:14 -0700 (MST)", "msg_from": "\"scott.marlowe\" <scott.marlowe@ihs.com>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 stamped" }, { "msg_contents": "scott.marlowe wrote:\n> > I wasn't sure how insecure SSL2 was, and whether it allowed you to\n> > authenticate without a password or something.\n> \n> SSL2 seems to get a lot of votes for being broken in ways that cannot be \n> fixed because they aren't simple buffer overflows. see:\n> \n> http://www.lne.com/ericm/papers/ssl_servers.html#1.2\n> \n> My suggestion would be to eventually phase out ssl2 in favor of ssl3 or \n> tls. And, as we are phasing it out, make it an opt-in thing, where the \n> dba has to turn on ssl2 KNOWING he is turning on a flawed protocol.\n\nThat was sort of my point --- if we allow SSLv2 in the server, are we\nopen to any security problems? Maybe not. I just don't know.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 18 Dec 2002 16:28:13 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: 7.3.1 stamped" }, { "msg_contents": "On Wed, 18 Dec 2002, Bruce Momjian wrote:\n\n> scott.marlowe wrote:\n> > > I wasn't sure how insecure SSL2 was, and whether it allowed you to\n> > > authenticate without a password or something.\n> >\n> > SSL2 seems to get a lot of votes for being broken in ways that cannot be\n> > fixed because they aren't simple buffer overflows. see:\n> >\n> > http://www.lne.com/ericm/papers/ssl_servers.html#1.2\n> >\n> > My suggestion would be to eventually phase out ssl2 in favor of ssl3 or\n> > tls. And, as we are phasing it out, make it an opt-in thing, where the\n> > dba has to turn on ssl2 KNOWING he is turning on a flawed protocol.\n>\n> That was sort of my point --- if we allow SSLv2 in the server, are we\n> open to any security problems? Maybe not. I just don't know.\n\nMy understanding of SSL/TLS is that the DBA himself has to enable it ...\nthere has to be a server/client key setup, similar to how it gets done\nwith Apache for https connections ... can someone confirm whether this is\nthe case?\n", "msg_date": "Wed, 18 Dec 2002 22:30:18 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "SSL/TLS support (Was: Re: 7.3.1 stamped)" }, { "msg_contents": "Marc G. Fournier wrote:\n> > > My suggestion would be to eventually phase out ssl2 in favor of ssl3 or\n> > > tls. And, as we are phasing it out, make it an opt-in thing, where the\n> > > dba has to turn on ssl2 KNOWING he is turning on a flawed protocol.\n> >\n> > That was sort of my point --- if we allow SSLv2 in the server, are we\n> > open to any security problems? Maybe not. I just don't know.\n> \n> My understanding of SSL/TLS is that the DBA himself has to enable it ...\n> there has to be a server/client key setup, similar to how it gets done\n> with Apache for https connections ... can someone confirm whether this is\n> the case?\n\nUh, I just followed the steps in the PostgresSQL install instructions:\n\t\n\topenssl req -new -text -out server.req\n\topenssl rsa -in privkey.pem -out server.key\n\trm privkey.pem\n\topenssl req -x509 -in server.req -text -key server.key -out server.crt\n\tchmod og-rwx server.key\n\nIsn't that similar to what was required in 7.2.X?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 18 Dec 2002 21:34:34 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: SSL/TLS support (Was: Re: 7.3.1 stamped)" }, { "msg_contents": "Bruce Momjian wrote:\n> Marc G. Fournier wrote:\n> \n>>>>My suggestion would be to eventually phase out ssl2 in favor of ssl3 or\n>>>>tls. And, as we are phasing it out, make it an opt-in thing, where the\n>>>>dba has to turn on ssl2 KNOWING he is turning on a flawed protocol.\n>>>\n>>>That was sort of my point --- if we allow SSLv2 in the server, are we\n>>>open to any security problems? Maybe not. I just don't know.\n\nThere are some weaknesses in SSLv2 that were fixed in SSLv3, but \nit takes a knowledgeable attacker to exploit them. Anyone who is \nseriously concerned can easily change the startup code in both \nclient and server and migrate to TLSv1. We kept the current \napproach solely for backward compatibilty.\n\nBear\n\n", "msg_date": "Sat, 28 Dec 2002 14:15:48 -0700", "msg_from": "Bear Giles <bgiles@coyotesong.com>", "msg_from_op": false, "msg_subject": "Re: SSL/TLS support (Was: Re: 7.3.1 stamped)" } ]
[ { "msg_contents": "Looks like they've caught up on a lot of our features. I have to say I\nappreciate them adding SERIAL as an alias for AUTO_INCREMENT. Perhaps we\nshould return the favour? :)\n\n(BTW, sorry to ppl who aren't interested, but I think it's important to see\nwhat other db's are doing. Also, someone asked me what kind of subselects\nthey support.)\n\nThe fact that they have GIS, subselects and prepared statements now is\ninteresting.\n\nD.1.1 Changes in release 4.1.0\n\nFaster binary protocol used with prepared statements.\nIn CREATE TABLE foo (a int not null primary key) the PRIMARY word is now\noptional.\nIn CREATE TABLE the attribute SERIAL is now an alias for BIGINT NOT NULL\nAUTO_INCREMENT UNIQUE.\nSELECT ... FROM DUAL is an alias for SELECT .... (To be compatible with some\nother databases).\nIf once creates a too long CHAR/VARCHAR it's now automatically changed to\nTEXT or BLOB; One will get a warning in this case.\nOne can specify the different BLOB/TEXT types with the syntax BLOB(length)\nand TEXT(length). MySQL will automatically change it to one of the internal\nBLOB/TEXT types.\nCHAR BYTE is an alias for CHAR BINARY.\nVARCHARACTER is an alias for VARCHAR.\nNew operators integer MOD integer and integer DIV integer.\nSERIAL DEFAULT VALUE is an alias for AUTO_INCREMENT.\nTRUE and FALSE are added as alias for 0 and 1.\nAliases are now forced in derived tables, as per SQL-99\nFixed SELECT .. LIMIT 0 to return proper row count for SQL_CALC_FOUND_ROWS.\nOne can specify many temporary directories to be used in a round-robin\nfashion with: --tmpdir=dirname1:dirname2:dirname3.\nSubqueries: SELECT * from t1 where t1.a=(SELECT t2.b FROM t2).\nDerived tables: SELECT a from t1, (select * from t2) WHERE t1.a=t2.a\nCharacter sets to be defined per column, table and database.\nUnicode (UTF8) support.\nBTREE index on HEAP tables.\nFaster embedded server.\nOne can add a comment per column in CREATE TABLE.\nSHOW FULL COLUMNS FROM table_name shows column comments.\nALTER DATABASE.\nSupport for GIS (Geometrical data).\nSHOW WARNINGS; Shows warnings from the last command.\nOne can specify a column type for a colum in CREATE TABLE ... SELECT by\ndefining the column in the CREATE part.\nCREATE TABLE foo (a tinyint not null) SELECT b+1 AS 'a' FROM bar;\nexpr SOUNDS LIKE expr same as SOUNDEX(expr)=SOUNDEX(expr).\nVARIANCE(expr) returns the standard variance of expr\n\nChris\n\n", "msg_date": "Wed, 18 Dec 2002 14:49:44 +0800", "msg_from": "\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au>", "msg_from_op": true, "msg_subject": "MySQL 4.1 Features" }, { "msg_contents": "At 02:49 PM 18/12/2002 +0800, Christopher Kings-Lynne wrote:\n>Looks like they've caught up on a lot of our features.\n\nAm I missing something here, or is 4.1 vapourware? Based on mysql.com, \n3.23.45a is current production, and 4.0.5a is current beta.\n\nI am happy to believe I missed something, but if not, I think we should \neither compare our plans for 7.4 with 4.1, or 3.2/4.0 with 7.3.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Wed, 18 Dec 2002 18:05:03 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": false, "msg_subject": "Re: MySQL 4.1 Features" }, { "msg_contents": "> Am I missing something here, or is 4.1 vapourware? Based on mysql.com, \n> 3.23.45a is current production, and 4.0.5a is current beta.\n> \n> I am happy to believe I missed something, but if not, I think we should \n> either compare our plans for 7.4 with 4.1, or 3.2/4.0 with 7.3.\n\nOK, fair enough.\n\nChris\n\n", "msg_date": "Wed, 18 Dec 2002 15:19:21 +0800", "msg_from": "\"Christopher Kings-Lynne\" <chriskl@familyhealth.com.au>", "msg_from_op": true, "msg_subject": "Re: MySQL 4.1 Features" }, { "msg_contents": "On Wed, 2002-12-18 at 01:49, Christopher Kings-Lynne wrote:\n> Looks like they've caught up on a lot of our features. I have to say I\n> appreciate them adding SERIAL as an alias for AUTO_INCREMENT. Perhaps we\n> should return the favour? :)\n\nWell, it's not the same as PostgreSQL's serial (which is a 32-bit NOT\nNULL incrementing integer -- MySQL's serial is a 64-bit unique, NOT NULL\nincrementing integer).\n\nPersonally, I think the implementations are different enough that making\nAUTO_INCREMENT an alias for SERIAL on the PostgreSQL side of things\nwould just confuse the situation further...\n\nCheers,\n\nNeil\n-- \nNeil Conway <neilc@samurai.com> || PGP Key ID: DB3C29FC\n\n\n\n", "msg_date": "18 Dec 2002 18:24:20 -0500", "msg_from": "Neil Conway <neilc@samurai.com>", "msg_from_op": false, "msg_subject": "Re: MySQL 4.1 Features" } ]
[ { "msg_contents": "Hi, all\r\nI have installed the Postgresql 7.3 . But I think something is wrong with authority.\r\nI have made the following operations:\r\n1. I enter the psql and run 'alter user postgres with password 'postgres''\r\n2. I change the pg_hba.conf and set the auth_type from 'trust' to 'password'\r\n3. Then I can not connect to server.\r\nI have test connect from local or connect from host, but it just return 'authority fail for user postgres'.\r\nI can connect to postgresql 7.23 successfully after above steps;\r\nGreat thanks for any message.\r\nJosh\r\n\n\n\n\n\nHi, all\nI have installed the Postgresql 7.3 . But I think something is wrong with \r\nauthority.\nI have made the following operations:\n1. I enter the psql and run 'alter user postgres with password \r\n'postgres''\n2. I change the pg_hba.conf and set the auth_type from 'trust' to \r\n'password'\n3. Then I can not connect to server.\nI have test connect from local or connect from host, but it just return \r\n'authority fail for user postgres'.\nI can connect to postgresql 7.23 successfully after above steps;\nGreat thanks for any message.\nJosh", "msg_date": "Wed, 18 Dec 2002 17:20:40 +0800", "msg_from": "postgresql <postgresql@db.pku.edu.cn>", "msg_from_op": true, "msg_subject": "a problem in authority" }, { "msg_contents": "On Wed, 2002-12-18 at 09:20, postgresql wrote:\n> 2. I change the pg_hba.conf and set the auth_type from 'trust' to\n> 'password'\n> \n> 3. Then I can not connect to server.\n\nTry using 'md5' instead of 'password' in pg_hba.conf.\n\n-- \nOliver Elphick <olly@lfix.co.uk>\nLFIX Limited\n\n", "msg_date": "18 Dec 2002 10:17:39 +0000", "msg_from": "Oliver Elphick <olly@lfix.co.uk>", "msg_from_op": false, "msg_subject": "Re: a problem in authority" } ]
[ { "msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\nHash: SHA1\n\nHi all!\nI get the following error when trying to connect to a pg-7.3 database on\nSolaris8 with the 7.3 JDBC-driver.\nHere is my output on the Solaris box:\nwww=> select version();\n version\n- -----------------------------------------------------------------------\n PostgreSQL 7.3 on sparc-sun-solaris2.8, compiled by GCC gcc (GCC) 3.2\n(1 row)\n\nwww=> set client_encoding = 'UNICODE';\nERROR: Conversion between UNICODE and LATIN1 is not supported\nwww=>\n\n\nHere is the same commands on my Linux box:\nandreak=# select version();\n version\n- --------------------------------------------------------------------------------------------------\n PostgreSQL 7.3 on i686-pc-linux-gnu, compiled by GCC gcc (GCC) 3.2 (Mandrake\nLinux 9.0 3.2-1mdk)\n(1 row)\n\nandreak=# set client_encoding = 'UNICODE';\nSET\nandreak=#\n\n=================\n\nDoes anybody know why this doesn't work on Solaris8? They are both compiled\nwith \"./configure --with-pam\" and multibyte support is enabled as default in\n7.3. I can allways recompile the JDBC-drivers with \"set client_encoding =\n'LATIN1'; show autocommit\" in AbstractJdbc1Connection.java, but that seems\nhacky.\n\n- --\nAndreas Joseph Krogh <andreak@officenet.no>\nThe difference between insanity and genius is measured by success\n-----BEGIN PGP SIGNATURE-----\nVersion: GnuPG v1.0.7 (GNU/Linux)\n\niD8DBQE+AGxvUopImDh2gfQRAoLbAJ9y9ZOZNS513zahIkPMDP5tJ1zSGACgj6Xo\nlxVLkYwE/I/sxRRRNvRSZFo=\n=FsMP\n-----END PGP SIGNATURE-----\n\n", "msg_date": "Wed, 18 Dec 2002 13:39:11 +0100", "msg_from": "Andreas Joseph Krogh <andreak@officenet.no>", "msg_from_op": true, "msg_subject": "Conversion between UNICODE and LATIN1 is not supported" }, { "msg_contents": "Andreas,\n\nThis would be better addressed to the hackers list, and I am forwarding\nit on.\n\nDave\nOn Wed, 2002-12-18 at 07:39, Andreas Joseph Krogh wrote:\n> -----BEGIN PGP SIGNED MESSAGE-----\n> Hash: SHA1\n> \n> Hi all!\n> I get the following error when trying to connect to a pg-7.3 database on\n> Solaris8 with the 7.3 JDBC-driver.\n> Here is my output on the Solaris box:\n> www=> select version();\n> version\n> - -----------------------------------------------------------------------\n> PostgreSQL 7.3 on sparc-sun-solaris2.8, compiled by GCC gcc (GCC) 3.2\n> (1 row)\n> \n> www=> set client_encoding = 'UNICODE';\n> ERROR: Conversion between UNICODE and LATIN1 is not supported\n> www=>\n> \n> \n> Here is the same commands on my Linux box:\n> andreak=# select version();\n> version\n> - --------------------------------------------------------------------------------------------------\n> PostgreSQL 7.3 on i686-pc-linux-gnu, compiled by GCC gcc (GCC) 3.2 (Mandrake\n> Linux 9.0 3.2-1mdk)\n> (1 row)\n> \n> andreak=# set client_encoding = 'UNICODE';\n> SET\n> andreak=#\n> \n> =================\n> \n> Does anybody know why this doesn't work on Solaris8? They are both compiled\n> with \"./configure --with-pam\" and multibyte support is enabled as default in\n> 7.3. I can allways recompile the JDBC-drivers with \"set client_encoding =\n> 'LATIN1'; show autocommit\" in AbstractJdbc1Connection.java, but that seems\n> hacky.\n> \n> - --\n> Andreas Joseph Krogh <andreak@officenet.no>\n> The difference between insanity and genius is measured by success\n> -----BEGIN PGP SIGNATURE-----\n> Version: GnuPG v1.0.7 (GNU/Linux)\n> \n> iD8DBQE+AGxvUopImDh2gfQRAoLbAJ9y9ZOZNS513zahIkPMDP5tJ1zSGACgj6Xo\n> lxVLkYwE/I/sxRRRNvRSZFo=\n> =FsMP\n> -----END PGP SIGNATURE-----\n> \n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org\n-- \nDave Cramer <dave@fastcrypt.com>\nCramer Consulting\n\n", "msg_date": "18 Dec 2002 08:00:57 -0500", "msg_from": "Dave Cramer <dave@fastcrypt.com>", "msg_from_op": false, "msg_subject": "Re: [JDBC] Conversion between UNICODE and LATIN1 is not supported" }, { "msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\nHash: SHA1\n\nOn Thursday 19 December 2002 08:37, you wrote:\n> Andreas Joseph Krogh wrote:\n> > -----BEGIN PGP SIGNED MESSAGE-----\n> > Hash: SHA1\n> >\n> > Hi all!\n> > I get the following error when trying to connect to a pg-7.3 database on\n> > Solaris8 with the 7.3 JDBC-driver.\n> > Here is my output on the Solaris box:\n> > www=> select version();\n> > version\n> > - -----------------------------------------------------------------------\n> > PostgreSQL 7.3 on sparc-sun-solaris2.8, compiled by GCC gcc (GCC) 3.2\n> > (1 row)\n> >\n> > www=> set client_encoding = 'UNICODE';\n> > ERROR: Conversion between UNICODE and LATIN1 is not supported\n>\n> This is not JDBC interface, it is PSQL command line utility!\n>\n> The message most likely is due to your database being LATIN1 encoded and I\n> must agree with PostgreSQL - Unicode cannot be converted to Latin1, not all\n> of it.\n>\n> So, check the database encoding: \"psql -l\".\n\nI'm sorry if I was unclear, I initially got the error using the JDBC-driver \nwhich issues the commands:\n\t\t\tjava.sql.ResultSet acRset =\n\t\t\t\tExecSQL(\"set client_encoding = 'UNICODE'; show autocommit\");\n\n\t\t\t//set encoding to be unicode\n\t\t\tencoding = Encoding.getEncoding(\"UNICODE\", null);\n\n\t\t\tif (!acRset.next())\n\t\t\t{\n\t\t\t\tthrow new PSQLException(\"postgresql.con.failed\", \"failed getting \nautocommit status\");\n\t\t\t}\n\nin AbstractJdbc1Connection.openConnection()\n\nI had to recompile the JDBC-drivers with the method changed to \n\"set client_encoding = 'LATIN1'\", as my database is as you guessed created \nwith \"createdb -E LATIN1 dbname\"\n\nThen I realised I could reproduse it from the psql cmd-line, and posted the \nerror I got from there. Sorry for the confusion.\n\nBut my question still stands, why does this work on Linux and not Solaris8? \nThe docs say PostgreSQL can convert to and from LATIN1<->UNICODE as long as \nmultibyte support is enabled, which it is default in 7.3.\n\n- -- \nAndreas Joseph Krogh <andreak@officenet.no>\nThe difference between insanity and genius is measured by success\n-----BEGIN PGP SIGNATURE-----\nVersion: GnuPG v1.0.7 (GNU/Linux)\n\niD8DBQE+AXljUopImDh2gfQRAlnYAJkBD+9kBPH8ME0doqOPctHApgHLowCeKkxK\n5m5GngiT8tW7gUAPPTRV9JI=\n=Hosd\n-----END PGP SIGNATURE-----\n\n", "msg_date": "Thu, 19 Dec 2002 08:46:35 +0100", "msg_from": "Andreas Joseph Krogh <andreak@officenet.no>", "msg_from_op": true, "msg_subject": "Re: [JDBC] Conversion between UNICODE and LATIN1 is not supported" }, { "msg_contents": "Hello,\n\nDo you have other databases that work with UNICODE? You can check the\nencoding of all your databases with 'psql -l'. You can check if you can\ncreate any unicode databases with 'createdb -E UNICODE mytestdatabase'. \n\nI didn't use postgresql on solaris yet, so the following is just a\nguess: maybe you need to install additional solaris packages. For\nexample to get tomcat/java working with utf-8/unicode/.. instead of\nplain ascii, we need at least the following on our systems:\n\nSUNWloc System Localization\nSUNWlocx System Localization (64-bit)\nSUNWeuluf UTF-8 L10N For Language Environment User Files\nSUNWeulux UTF-8 L10N For Language Environment User Files (64-bit)\n\nKind regards,\nDries Verachtert\n\n\nOn Wed, 2002-12-18 at 13:39, Andreas Joseph Krogh wrote:\n> -----BEGIN PGP SIGNED MESSAGE-----\n> Hash: SHA1\n> \n> Hi all!\n> I get the following error when trying to connect to a pg-7.3 database on\n> Solaris8 with the 7.3 JDBC-driver.\n> Here is my output on the Solaris box:\n> www=> select version();\n> version\n> - -----------------------------------------------------------------------\n> PostgreSQL 7.3 on sparc-sun-solaris2.8, compiled by GCC gcc (GCC) 3.2\n> (1 row)\n> \n> www=> set client_encoding = 'UNICODE';\n> ERROR: Conversion between UNICODE and LATIN1 is not supported\n> www=>\n> \n> \n> Here is the same commands on my Linux box:\n> andreak=# select version();\n> version\n> - --------------------------------------------------------------------------------------------------\n> PostgreSQL 7.3 on i686-pc-linux-gnu, compiled by GCC gcc (GCC) 3.2 (Mandrake\n> Linux 9.0 3.2-1mdk)\n> (1 row)\n> \n> andreak=# set client_encoding = 'UNICODE';\n> SET\n> andreak=#\n> \n> =================\n> \n> Does anybody know why this doesn't work on Solaris8? They are both compiled\n> with \"./configure --with-pam\" and multibyte support is enabled as default in\n> 7.3. I can allways recompile the JDBC-drivers with \"set client_encoding =\n> 'LATIN1'; show autocommit\" in AbstractJdbc1Connection.java, but that seems\n> hacky.\n> \n> - --\n> Andreas Joseph Krogh <andreak@officenet.no>\n> The difference between insanity and genius is measured by success\n> -----BEGIN PGP SIGNATURE-----\n> Version: GnuPG v1.0.7 (GNU/Linux)\n> \n> iD8DBQE+AGxvUopImDh2gfQRAoLbAJ9y9ZOZNS513zahIkPMDP5tJ1zSGACgj6Xo\n> lxVLkYwE/I/sxRRRNvRSZFo=\n> =FsMP\n> -----END PGP SIGNATURE-----\n> \n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org\n\n\n", "msg_date": "19 Dec 2002 16:09:13 +0100", "msg_from": "Dries Verachtert <dries@ulyssis.org>", "msg_from_op": false, "msg_subject": "Re: Conversion between UNICODE and LATIN1 is not supported" }, { "msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\nHash: SHA1\n\nOn Thursday 19 December 2002 16:09, you wrote:\n> Hello,\n>\n> Do you have other databases that work with UNICODE? You can check the\n> encoding of all your databases with 'psql -l'. You can check if you can\n> create any unicode databases with 'createdb -E UNICODE mytestdatabase'.\n>\n> I didn't use postgresql on solaris yet, so the following is just a\n> guess: maybe you need to install additional solaris packages. For\n> example to get tomcat/java working with utf-8/unicode/.. instead of\n> plain ascii, we need at least the following on our systems:\n>\n> SUNWloc System Localization\n> SUNWlocx System Localization (64-bit)\n> SUNWeuluf UTF-8 L10N For Language Environment User Files\n> SUNWeulux UTF-8 L10N For Language Environment User Files (64-bit)\n\nNo, all our databases are LATIN1, it's the JDBC driver which want to \"set \nclient_encoding = 'UNICODE'\". I havn't tried to connect to a UNICODE (-E \nUNICODE) database with the JDBC driver.\n\nI now created a test database <andreak> with -E UNICODE\nHere is some output:\n\n[andreak@www2] ~$ uname -a\nSunOS www2.java.no 5.8 Generic_108528-16 sun4u sparc SUNW,Ultra-250\n[andreak@www2] ~$ psql -l\n List of databases\n Name | Owner | Encoding\n- -----------+----------+-----------\n andreak | andreak | UNICODE\n template0 | postgres | SQL_ASCII\n template1 | postgres | SQL_ASCII\n www | onp | LATIN1\n(4 rows)\n\n[andreak@www2] ~$ psql andreak\nWelcome to psql 7.3, the PostgreSQL interactive terminal.\n\nType: \\copyright for distribution terms\n \\h for help with SQL commands\n \\? for help on internal slash commands\n \\g or terminate with semicolon to execute query\n \\q to quit\n\nandreak=# select version();\n version\n- -----------------------------------------------------------------------\n PostgreSQL 7.3 on sparc-sun-solaris2.8, compiled by GCC gcc (GCC) 3.2\n(1 row)\n\nandreak=# show client_encoding ;\n client_encoding\n- -----------------\n UNICODE\n(1 row)\n\nandreak=# set client_encoding = 'LATIN1';\nERROR: Conversion between LATIN1 and UNICODE is not supported\nandreak=# set client_encoding = 'UNICODE';\nSET\nandreak=# \\q\n[andreak@www2] ~$ psql www\nWelcome to psql 7.3, the PostgreSQL interactive terminal.\n\nType: \\copyright for distribution terms\n \\h for help with SQL commands\n \\? for help on internal slash commands\n \\g or terminate with semicolon to execute query\n \\q to quit\n\nwww=# set client_encoding = 'LATIN1';\nSET\nwww=# set client_encoding = 'UNICODE';\nERROR: Conversion between UNICODE and LATIN1 is not supported\nwww=#\n\nThis works fine on Linux, I havn't tried to patch Solaris with l10n-stuff yet, \nbut it would be nice to get some feedback and understand why this is \nhappening and why I don't get a more descriptive error. For some reason none \nof the conversions work at all, except, of course, conversion between similar \nencodings like UNICODE<->UNICODE.\n\n- -- \nAndreas Joseph Krogh <andreak@officenet.no>\n\tThere will always be someone who agrees with you\n\tbut is, inexplicably, a moron.\n-----BEGIN PGP SIGNATURE-----\nVersion: GnuPG v1.0.7 (GNU/Linux)\n\niD8DBQE+AeQ4UopImDh2gfQRAuRvAKCbk4pyp7CbXiA4NGq8ntKpZ2IriwCeKcaC\n/pPGUgPzEuzDmIdX3shR8Ps=\n=EYP3\n-----END PGP SIGNATURE-----\n\n", "msg_date": "Thu, 19 Dec 2002 16:22:32 +0100", "msg_from": "Andreas Joseph Krogh <andreak@officenet.no>", "msg_from_op": true, "msg_subject": "Re: [JDBC] Conversion between UNICODE and LATIN1 is not supported" } ]
[ { "msg_contents": "\nGoing to announce later this evening to give the mirrors a chance to catch\nup ... let me know if there are any problems ..\n\n\n", "msg_date": "Wed, 18 Dec 2002 10:18:00 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": true, "msg_subject": "v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "Is this going to be announced to a wider press audience? Has anyone gone\nover the \"list of things to do when we release\" to make sure things like\nthe websites getting updated or perhaps getting rpm builds coordinated\nhas been done?\n\nRobert Treat\n\n\nOn Wed, 2002-12-18 at 09:18, Marc G. Fournier wrote:\n> \n> Going to announce later this evening to give the mirrors a chance to catch\n> up ... let me know if there are any problems ..\n> \n\n\n\n", "msg_date": "18 Dec 2002 09:30:20 -0500", "msg_from": "Robert Treat <xzilla@users.sourceforge.net>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On Wed, 18 Dec 2002, Robert Treat wrote:\n\n> Is this going to be announced to a wider press audience? Has anyone gone\n> over the \"list of things to do when we release\" to make sure things like\n> the websites getting updated or perhaps getting rpm builds coordinated\n> has been done?\n\nNo, we don't do that with minor releases ... nothing has changed that\nneeds to be announced, other then a few bugs fixed ...\n\n\n>\n> Robert Treat\n>\n>\n> On Wed, 2002-12-18 at 09:18, Marc G. Fournier wrote:\n> >\n> > Going to announce later this evening to give the mirrors a chance to catch\n> > up ... let me know if there are any problems ..\n> >\n>\n>\n>\n>\n", "msg_date": "Wed, 18 Dec 2002 10:51:11 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": true, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On Wed, 2002-12-18 at 09:51, Marc G. Fournier wrote:\n> On Wed, 18 Dec 2002, Robert Treat wrote:\n> \n> > Is this going to be announced to a wider press audience? Has anyone gone\n> > over the \"list of things to do when we release\" to make sure things like\n> > the websites getting updated or perhaps getting rpm builds coordinated\n> > has been done?\n> \n> No, we don't do that with minor releases ... nothing has changed that\n> needs to be announced, other then a few bugs fixed ...\n> \n\nSurly the websites will still need to be updated, rpm's need to be\nbuilt, things like that...\n\nRobert Treat\n\n", "msg_date": "18 Dec 2002 10:43:27 -0500", "msg_from": "Robert Treat <xzilla@users.sourceforge.net>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On Wed, 18 Dec 2002, Robert Treat wrote:\n\n> On Wed, 2002-12-18 at 09:51, Marc G. Fournier wrote:\n> > On Wed, 18 Dec 2002, Robert Treat wrote:\n> >\n> > > Is this going to be announced to a wider press audience? Has anyone gone\n> > > over the \"list of things to do when we release\" to make sure things like\n> > > the websites getting updated or perhaps getting rpm builds coordinated\n> > > has been done?\n> >\n> > No, we don't do that with minor releases ... nothing has changed that\n> > needs to be announced, other then a few bugs fixed ...\n> >\n>\n> Surly the websites will still need to be updated, rpm's need to be\n> built, things like that...\n\nOf course, but that has nothing to do with announcing to a wider press\naudience ... :)\n\n", "msg_date": "Wed, 18 Dec 2002 11:59:04 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": true, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "\"Marc G. Fournier\" <scrappy@hub.org> writes:\n> Going to announce later this evening to give the mirrors a chance to catch\n> up ... let me know if there are any problems ..\n\nTarball looks good from here.\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 18 Dec 2002 11:57:48 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ... " }, { "msg_contents": "On Wed, 18 Dec 2002, Tom Lane wrote:\n\n> \"Marc G. Fournier\" <scrappy@hub.org> writes:\n> > Going to announce later this evening to give the mirrors a chance to catch\n> > up ... let me know if there are any problems ..\n>\n> Tarball looks good from here.\n\nGreat, put out a short techy announcement this evening when I geth home\n...\n\n", "msg_date": "Wed, 18 Dec 2002 13:16:42 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": true, "msg_subject": "Re: v7.3.1 tar ready ... please check it ... " }, { "msg_contents": "Marc G. Fournier writes:\n\n> Going to announce later this evening to give the mirrors a chance to catch\n> up ... let me know if there are any problems ..\n\nPlenty...\n\nThe release notes are missing at least one item and contain at least one\nfactual mistake that needs to be fixed. The HISTORY file needs to be\nreformatted because the line breaks look pretty ugly. The list of\nsupported platforms needs to be revised and all platforms that have not\nreceived a confirmation yet have to be moved to unsupported. The INSTALL\nfile needs to be updated to get the list of supported platforms up-to-date\nand the references to 7.3 need to be changed to 7.3.1. A note about the\nSSL incompatibility needs to be added or a fix needs to be mentioned in\nthe list of changes. Someone also wrote me that a translation file is\nstuck in the pgsql-patches queue. It would be good if we could get that\none in as well.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Wed, 18 Dec 2002 20:56:10 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "Peter Eisentraut wrote:\n> Marc G. Fournier writes:\n> \n> > Going to announce later this evening to give the mirrors a chance to catch\n> > up ... let me know if there are any problems ..\n> \n> Plenty...\n> \n> The release notes are missing at least one item and contain at least one\n> factual mistake that needs to be fixed. The HISTORY file needs to be\n> reformatted because the line breaks look pretty ugly. The list of\n\nHow do you do that? Do you manually reformat the whole file after you\ngenerate it, or do you just cut-paste the new release info into\n/HISTORY so the old manual formatting remains? It did line break badly.\n\n> supported platforms needs to be revised and all platforms that have not\n> received a confirmation yet have to be moved to unsupported. The INSTALL\n\nDo we do that during the first minor? I can do that if this is the time.\n\n> file needs to be updated to get the list of supported platforms up-to-date\n> and the references to 7.3 need to be changed to 7.3.1. A note about the\n\nMarc applied the patch after I stamped it. Marc, do you want me to do it?\n\n> SSL incompatibility needs to be added or a fix needs to be mentioned in\n> the list of changes. Someone also wrote me that a translation file is\n> stuck in the pgsql-patches queue. It would be good if we could get that\n> one in as well.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 18 Dec 2002 15:10:43 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On Wed, 18 Dec 2002, Bruce Momjian wrote:\n\n> > file needs to be updated to get the list of supported platforms up-to-date\n> > and the references to 7.3 need to be changed to 7.3.1. A note about the\n>\n> Marc applied the patch after I stamped it. Marc, do you want me to do it?\n\nJust curious as to whether any of this is critical enough to force a\nrebuild of the .tar.gz files, or can they wait until v7.3.2? That is my\nonly concern ... we can do it, and I can do the announce in the morning\ninstead of this evening, just want to make sure that its critical enough\n...\n\n", "msg_date": "Wed, 18 Dec 2002 16:25:39 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": true, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "Marc G. Fournier wrote:\n> On Wed, 18 Dec 2002, Bruce Momjian wrote:\n> \n> > > file needs to be updated to get the list of supported platforms up-to-date\n> > > and the references to 7.3 need to be changed to 7.3.1. A note about the\n> >\n> > Marc applied the patch after I stamped it. Marc, do you want me to do it?\n> \n> Just curious as to whether any of this is critical enough to force a\n> rebuild of the .tar.gz files, or can they wait until v7.3.2? That is my\n> only concern ... we can do it, and I can do the announce in the morning\n> instead of this evening, just want to make sure that its critical enough\n> ...\n\nNot sure. The tarball has 7.3 mentioned in the INSTALL file, not 7.3.1.\nI only today saw how to update that, and I have added it to my\nchecklist.\n\nI think the port changes can wait for 7.3.2. I don't think the SSL\nchange is critical to be mentioned because it only opens up something\nthat was disabled in 7.3.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 18 Dec 2002 15:28:52 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "Bruce Momjian writes:\n\n> How do you do that? Do you manually reformat the whole file after you\n> generate it, or do you just cut-paste the new release info into\n> /HISTORY so the old manual formatting remains? It did line break badly.\n\nI put in the changes I had in mind and reformatted it, so as far as I'm\nconcerned it's good to go now.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Thu, 19 Dec 2002 00:42:47 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "Marc G. Fournier writes:\n\n> Just curious as to whether any of this is critical enough to force a\n> rebuild of the .tar.gz files, or can they wait until v7.3.2? That is my\n> only concern ... we can do it, and I can do the announce in the morning\n> instead of this evening, just want to make sure that its critical enough\n> ...\n\nWell the changes are mostly about elimating incorrect information from the\nrelease notes, which is surely critical, and putting in the correct\nversion number, which is also critical. And neither of these things can\nwait for 7.3.2, because that would kind of defeat the point. If you jump\nthe gun with building the tarballs that's the price you have to pay.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Thu, 19 Dec 2002 00:48:04 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "\n'K, I'm going to remove the tar files ... Bruce, can you go through these\nand get them fixed up?\n\nPeter, I have to take part of the blame away from Tom ... I'm on the road\ntomrorow afternoon to Ontario, and won't be back online until *late* Fri,\nso we kinda rushed it all ...\n\nTom, we tried ... I'll do up the tar ball on Friday, if everyone can tak\nthe next day and a bit to make sure we haven't missed anything?\n\nOn Thu, 19 Dec 2002, Peter Eisentraut wrote:\n\n> Marc G. Fournier writes:\n>\n> > Just curious as to whether any of this is critical enough to force a\n> > rebuild of the .tar.gz files, or can they wait until v7.3.2? That is my\n> > only concern ... we can do it, and I can do the announce in the morning\n> > instead of this evening, just want to make sure that its critical enough\n> > ...\n>\n> Well the changes are mostly about elimating incorrect information from the\n> release notes, which is surely critical, and putting in the correct\n> version number, which is also critical. And neither of these things can\n> wait for 7.3.2, because that would kind of defeat the point. If you jump\n> the gun with building the tarballs that's the price you have to pay.\n>\n> --\n> Peter Eisentraut peter_e@gmx.net\n>\n>\n", "msg_date": "Wed, 18 Dec 2002 23:26:28 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": true, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "At 11:26 PM 18/12/2002 -0400, Marc G. Fournier wrote:\n>Tom, we tried ... I'll do up the tar ball on Friday, if everyone can tak\n>the next day and a bit to make sure we haven't missed anything?\n\nSeeing the setting for MAX_FSM_RELATIONS bumped to 1000 would be good \n(patch already sent)\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Thu, 19 Dec 2002 14:44:46 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "\nI don't think we can bump that up in a minor. Should we?\n\n---------------------------------------------------------------------------\n\nPhilip Warner wrote:\n> At 11:26 PM 18/12/2002 -0400, Marc G. Fournier wrote:\n> >Tom, we tried ... I'll do up the tar ball on Friday, if everyone can tak\n> >the next day and a bit to make sure we haven't missed anything?\n> \n> Seeing the setting for MAX_FSM_RELATIONS bumped to 1000 would be good \n> (patch already sent)\n> \n> \n> ----------------------------------------------------------------\n> Philip Warner | __---_____\n> Albatross Consulting Pty. Ltd. |----/ - \\\n> (A.B.N. 75 008 659 498) | /(@) ______---_\n> Tel: (+61) 0500 83 82 81 | _________ \\\n> Fax: (+61) 03 5330 3172 | ___________ |\n> Http://www.rhyme.com.au | / \\|\n> | --________--\n> PGP key available upon request, | /\n> and from pgp5.ai.mit.edu:11371 |/\n> \n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 5: Have you checked our extensive FAQ?\n> \n> http://www.postgresql.org/users-lounge/docs/faq.html\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 18 Dec 2002 22:49:18 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "At 10:49 PM 18/12/2002 -0500, Bruce Momjian wrote:\n>I don't think we can bump that up in a minor.\n\nWhy not? It's a relatively serious problem with the default config.\n\n\n>Should we?\n\nYes.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 03 5330 3172 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n\n", "msg_date": "Thu, 19 Dec 2002 17:32:24 +1100", "msg_from": "Philip Warner <pjw@rhyme.com.au>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On Thu, 19 Dec 2002, Philip Warner wrote:\n\n> At 10:49 PM 18/12/2002 -0500, Bruce Momjian wrote:\n> >I don't think we can bump that up in a minor.\n> \n> Why not? It's a relatively serious problem with the default config.\n> \n> \n> >Should we?\n> \n> Yes.\n\nI concur. The problems of a too-low fsm setting are serious, and I can't \nsee setting fsm to 1000 creating problems for anyone. That's actually \nstill a pretty low number for production servers. \n\n", "msg_date": "Thu, 19 Dec 2002 10:08:29 -0700 (MST)", "msg_from": "\"scott.marlowe\" <scott.marlowe@ihs.com>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "scott.marlowe wrote:\n> On Thu, 19 Dec 2002, Philip Warner wrote:\n> \n> > At 10:49 PM 18/12/2002 -0500, Bruce Momjian wrote:\n> > >I don't think we can bump that up in a minor.\n> > \n> > Why not? It's a relatively serious problem with the default config.\n> > \n> > \n> > >Should we?\n> > \n> > Yes.\n> \n> I concur. The problems of a too-low fsm setting are serious, and I can't \n> see setting fsm to 1000 creating problems for anyone. That's actually \n> still a pretty low number for production servers. \n\nOK, I agree it should be bumped up, but the issue is whether to do that\nin a minor release. It will increase shared memory by 36k. Is that\nsafe in a beta? Tom, Mr. FSM, can you comment?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Thu, 19 Dec 2002 12:15:32 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> OK, I agree it should be bumped up, but the issue is whether to do that\n> in a minor release. It will increase shared memory by 36k. Is that\n> safe in a beta? Tom, Mr. FSM, can you comment?\n\nIt seems a reasonably safe change, but I too am concerned about making\nsuch changes in minor releases. For 7.3.1 in particular, since there\nhave already been publicly available tarballs, I think we should avoid\nmaking any more changes other than documentation fixes; otherwise\nthere's too much risk of confusion (\"which 7.3.1 have you got?\").\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 19 Dec 2002 12:29:23 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ... " }, { "msg_contents": "Tom Lane wrote:\n> Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> > OK, I agree it should be bumped up, but the issue is whether to do that\n> > in a minor release. It will increase shared memory by 36k. Is that\n> > safe in a beta? Tom, Mr. FSM, can you comment?\n> \n> It seems a reasonably safe change, but I too am concerned about making\n> such changes in minor releases. For 7.3.1 in particular, since there\n> have already been publicly available tarballs, I think we should avoid\n> making any more changes other than documentation fixes; otherwise\n> there's too much risk of confusion (\"which 7.3.1 have you got?\").\n\nIt would have to be 7.3.2, I think.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Thu, 19 Dec 2002 12:30:51 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "OK, what additional things need to be done for 7.3.1? As far as I know,\nwe have done everything.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Thu, 19 Dec 2002 12:58:47 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "What else needs to be done for 7.3.1?" }, { "msg_contents": "Tom Lane wrote:\n> Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> > OK, I agree it should be bumped up, but the issue is whether to do that\n> > in a minor release. It will increase shared memory by 36k. Is that\n> > safe in a beta? Tom, Mr. FSM, can you comment?\n> \n> It seems a reasonably safe change, but I too am concerned about making\n> such changes in minor releases. For 7.3.1 in particular, since there\n> have already been publicly available tarballs, I think we should avoid\n> making any more changes other than documentation fixes; otherwise\n> there's too much risk of confusion (\"which 7.3.1 have you got?\").\n\nI don't quite understand why there's reluctance to change this,\nthough. What will it break?\n\nIt's probably sufficient to put something in the release notes\nindicating that MAX_FSM_RELATIONS has increased and that you should\nmanually set it back to 100 in the config file if the change causes\nproblems.\n\nWith even relatively old systems having 128 megabytes or more memory\ninstalled, I'd think that a 36k increase in shared memory usage is\nsmall enough to make the change worth the risk.\n\n\nNow, your concerns are probably more justified if you're worried about\nthe change causing some little-used code to suddenly start seeing a\nlot of usage...\n\n\n\n-- \nKevin Brown\t\t\t\t\t kevin@sysexperts.com\n", "msg_date": "Thu, 19 Dec 2002 10:19:05 -0800", "msg_from": "Kevin Brown <kevin@sysexperts.com>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On Thu, 2002-12-19 at 12:58, Bruce Momjian wrote:\n> OK, what additional things need to be done for 7.3.1? As far as I know,\n> we have done everything.\n> \n\nDo we want to coordinate with Lamar or Oliver about having packages\nready to coincide with the release announcement?\n\nRobert Treat\n\n", "msg_date": "19 Dec 2002 13:40:00 -0500", "msg_from": "Robert Treat <xzilla@users.sourceforge.net>", "msg_from_op": false, "msg_subject": "Re: What else needs to be done for 7.3.1?" }, { "msg_contents": "Kevin Brown wrote:\n> Tom Lane wrote:\n> > Bruce Momjian <pgman@candle.pha.pa.us> writes:\n> > > OK, I agree it should be bumped up, but the issue is whether to do that\n> > > in a minor release. It will increase shared memory by 36k. Is that\n> > > safe in a beta? Tom, Mr. FSM, can you comment?\n> > \n> > It seems a reasonably safe change, but I too am concerned about making\n> > such changes in minor releases. For 7.3.1 in particular, since there\n> > have already been publicly available tarballs, I think we should avoid\n> > making any more changes other than documentation fixes; otherwise\n> > there's too much risk of confusion (\"which 7.3.1 have you got?\").\n> \n> I don't quite understand why there's reluctance to change this,\n> though. What will it break?\n> \n> It's probably sufficient to put something in the release notes\n> indicating that MAX_FSM_RELATIONS has increased and that you should\n> manually set it back to 100 in the config file if the change causes\n> problems.\n> \n> With even relatively old systems having 128 megabytes or more memory\n> installed, I'd think that a 36k increase in shared memory usage is\n> small enough to make the change worth the risk.\n> \n> \n> Now, your concerns are probably more justified if you're worried about\n> the change causing some little-used code to suddenly start seeing a\n> lot of usage...\n\nI think we have agreed on putting it on 7.3.X. The issue is that 7.3.1\nis already packaged, so it will have to wait for 7.3.2.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Thu, 19 Dec 2002 13:44:12 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On Thu, 2002-12-19 at 18:40, Robert Treat wrote:\n> On Thu, 2002-12-19 at 12:58, Bruce Momjian wrote:\n> > OK, what additional things need to be done for 7.3.1? As far as I know,\n> > we have done everything.\n> > \n> \n> Do we want to coordinate with Lamar or Oliver about having packages\n> ready to coincide with the release announcement?\n\nAll is ready for when the new tar.gz appears. Once I have downloaded\nit, it should only take 10 minutes or so to package. \n\n-- \nOliver Elphick Oliver.Elphick@lfix.co.uk\nIsle of Wight, UK http://www.lfix.co.uk/oliver\nGPG: 1024D/3E1D0C1C: CA12 09E0 E8D5 8870 5839 932A 614D 4C34 3E1D 0C1C\n ========================================\n \"And she shall bring forth a son, and thou shall call \n his name JESUS; for he shall save his people from \n their sins.\" Matthew 1:21 \n\n", "msg_date": "20 Dec 2002 00:02:45 +0000", "msg_from": "Oliver Elphick <olly@lfix.co.uk>", "msg_from_op": false, "msg_subject": "Re: What else needs to be done for 7.3.1?" } ]
[ { "msg_contents": "\n\n> -----Original Message-----\n> From: Marc G. Fournier [mailto:scrappy@hub.org] \n> Sent: 18 December 2002 14:51\n> To: Robert Treat\n> Cc: pgsql-hackers@postgresql.org\n> Subject: Re: [HACKERS] v7.3.1 tar ready ... please check it ...\n> \n> \n> On Wed, 18 Dec 2002, Robert Treat wrote:\n> \n> > Is this going to be announced to a wider press audience? Has anyone \n> > gone over the \"list of things to do when we release\" to make sure \n> > things like the websites getting updated or perhaps getting \n> rpm builds \n> > coordinated has been done?\n> \n> No, we don't do that with minor releases ... nothing has \n> changed that needs to be announced, other then a few bugs fixed ...\n\nMaybe we should? The more publicity the better etc...\n\nRegards, Dave\n", "msg_date": "Wed, 18 Dec 2002 14:53:28 -0000", "msg_from": "\"Dave Page\" <dpage@vale-housing.co.uk>", "msg_from_op": true, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On Wed, 18 Dec 2002, Dave Page wrote:\n\n>\n>\n> > -----Original Message-----\n> > From: Marc G. Fournier [mailto:scrappy@hub.org]\n> > Sent: 18 December 2002 14:51\n> > To: Robert Treat\n> > Cc: pgsql-hackers@postgresql.org\n> > Subject: Re: [HACKERS] v7.3.1 tar ready ... please check it ...\n> >\n> >\n> > On Wed, 18 Dec 2002, Robert Treat wrote:\n> >\n> > > Is this going to be announced to a wider press audience? Has anyone\n> > > gone over the \"list of things to do when we release\" to make sure\n> > > things like the websites getting updated or perhaps getting\n> > rpm builds\n> > > coordinated has been done?\n> >\n> > No, we don't do that with minor releases ... nothing has\n> > changed that needs to be announced, other then a few bugs fixed ...\n>\n> Maybe we should? The more publicity the better etc...\n\nThe problem is that there is nothing to announce ... \"Hi, we fixed some\nbugs\"? :) minor releases don't have any features added to them, so isn't\nreally news worthy ... :(\n", "msg_date": "Wed, 18 Dec 2002 11:25:32 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On Wed, 18 Dec 2002, Marc G. Fournier wrote:\n\n> On Wed, 18 Dec 2002, Dave Page wrote:\n>\n> >\n> >\n> > > -----Original Message-----\n> > > From: Marc G. Fournier [mailto:scrappy@hub.org]\n> > > Sent: 18 December 2002 14:51\n> > > To: Robert Treat\n> > > Cc: pgsql-hackers@postgresql.org\n> > > Subject: Re: [HACKERS] v7.3.1 tar ready ... please check it ...\n> > >\n> > >\n> > > On Wed, 18 Dec 2002, Robert Treat wrote:\n> > >\n> > > > Is this going to be announced to a wider press audience? Has anyone\n> > > > gone over the \"list of things to do when we release\" to make sure\n> > > > things like the websites getting updated or perhaps getting\n> > > rpm builds\n> > > > coordinated has been done?\n> > >\n> > > No, we don't do that with minor releases ... nothing has\n> > > changed that needs to be announced, other then a few bugs fixed ...\n> >\n> > Maybe we should? The more publicity the better etc...\n>\n> The problem is that there is nothing to announce ... \"Hi, we fixed some\n> bugs\"? :) minor releases don't have any features added to them, so isn't\n> really news worthy ... :(\n>\n\nI think changing major version of libpq is an important thing to be\nmentioned.\n\n\n> ---------------------------(end of broadcast)---------------------------\n> TIP 4: Don't 'kill -9' the postmaster\n>\n\n\tRegards,\n\t\tOleg\n_____________________________________________________________\nOleg Bartunov, sci.researcher, hostmaster of AstroNet,\nSternberg Astronomical Institute, Moscow University (Russia)\nInternet: oleg@sai.msu.su, http://www.sai.msu.su/~megera/\nphone: +007(095)939-16-83, +007(095)939-23-83\n\n", "msg_date": "Wed, 18 Dec 2002 18:45:52 +0300 (MSK)", "msg_from": "Oleg Bartunov <oleg@sai.msu.su>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On Wed, 18 Dec 2002, Marc G. Fournier wrote:\n\n> The problem is that there is nothing to announce ... \"Hi, we fixed some\n> bugs\"? :) minor releases don't have any features added to them, so isn't\n> really news worthy ... :(\n\nI don't know, if you're a postgresql user and you don't read these lists, \nyou might find out about a bug in a release note and upgrade when you \notherwise might not.\n\nwww.linuxtoday.com has weekly updates from many gnu / OSS projects which \nare far less interesting than our 7.3.1 release is. I could see posting a \nminor upgrade release notice there and on other OSS news web site \n(freshmeat, slashdot, etc...)\n\n", "msg_date": "Wed, 18 Dec 2002 08:54:51 -0700 (MST)", "msg_from": "\"scott.marlowe\" <scott.marlowe@ihs.com>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On Wed, 18 Dec 2002, scott.marlowe wrote:\n\n> On Wed, 18 Dec 2002, Marc G. Fournier wrote:\n>\n> > The problem is that there is nothing to announce ... \"Hi, we fixed some\n> > bugs\"? :) minor releases don't have any features added to them, so isn't\n> > really news worthy ... :(\n>\n> I don't know, if you're a postgresql user and you don't read these lists,\n> you might find out about a bug in a release note and upgrade when you\n> otherwise might not.\n>\n> www.linuxtoday.com has weekly updates from many gnu / OSS projects which\n> are far less interesting than our 7.3.1 release is. I could see posting a\n> minor upgrade release notice there and on other OSS news web site\n> (freshmeat, slashdot, etc...)\n\nOkay, that I agree on ... note that my press list contains alot more then\nthat, most of which wouldn't be appropriate for a minor release ... maybe\nwe need to setup a seperate 'minor release' list?\n\nPlease note, I have no problems sending it out to a bunch of email\naddresses, I'm just questioning whether it makes sense to do so ...\nfreshmeat not withstanding, since I don't see it as much of an annoucne\nsite as a centralized list of OSS software ...\n\n\n", "msg_date": "Wed, 18 Dec 2002 11:58:29 -0400 (AST)", "msg_from": "\"Marc G. Fournier\" <scrappy@hub.org>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "scott.marlowe wrote:\n> On Wed, 18 Dec 2002, Marc G. Fournier wrote:\n> \n> \n>>The problem is that there is nothing to announce ... \"Hi, we fixed some\n>>bugs\"? :) minor releases don't have any features added to them, so isn't\n>>really news worthy ... :(\n> \n> \n> I don't know, if you're a postgresql user and you don't read these lists, \n> you might find out about a bug in a release note and upgrade when you \n> otherwise might not.\n> \n> www.linuxtoday.com has weekly updates from many gnu / OSS projects which \n> are far less interesting than our 7.3.1 release is. I could see posting a \n> minor upgrade release notice there and on other OSS news web site \n> (freshmeat, slashdot, etc...)\n\nAt the very least the PostgreSQL website team should be loudly notified so we can confirm the needed links have been \nupdated prior to the release announcement.\n\nIt probably wouldn't hurt to go through a proper release process, but determine which steps are optional or not needed \nfor smaller releases.\n\n:-)\n\nRegards and best wishes,\n\nJustin clift\n\n-- \n\"My grandfather once told me that there are two kinds of people: those\nwho work and those who take the credit. He told me to try to be in the\nfirst group; there was less competition there.\"\n- Indira Gandhi\n\n", "msg_date": "Thu, 19 Dec 2002 03:05:04 +1100", "msg_from": "Justin Clift <justin@postgresql.org>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On 18 Dec 2002 at 8:54, scott.marlowe wrote:\n> \n> www.linuxtoday.com has weekly updates from many gnu / OSS projects which \n> are far less interesting than our 7.3.1 release is. I could see posting a \n> minor upgrade release notice there and on other OSS news web site \n> (freshmeat, slashdot, etc...)\n\nI read linuxtoday and /. daily. Linuxtoday is OK but /. would be an almost \nwaste of another 800 comments of postgresql v/s mysql.\n\nJust a thought..\n\nBye\n Shridhar\n\n--\nManly's Maxim:\tLogic is a systematic method of coming to the wrong conclusion\t\nwith confidence.\n\n", "msg_date": "Wed, 18 Dec 2002 21:37:01 +0530", "msg_from": "\"Shridhar Daithankar\" <shridhar_daithankar@persistent.co.in>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On Wed, 2002-12-18 at 08:53, Dave Page wrote:\n> > -----Original Message-----\n> > From: Marc G. Fournier [mailto:scrappy@hub.org] \n> > Sent: 18 December 2002 14:51\n> > To: Robert Treat\n> > Cc: pgsql-hackers@postgresql.org\n> > Subject: Re: [HACKERS] v7.3.1 tar ready ... please check it ...\n> > \n> > \n> > On Wed, 18 Dec 2002, Robert Treat wrote:\n> > \n> > > Is this going to be announced to a wider press audience? Has anyone \n> > > gone over the \"list of things to do when we release\" to make sure \n> > > things like the websites getting updated or perhaps getting \n> > rpm builds \n> > > coordinated has been done?\n> > \n> > No, we don't do that with minor releases ... nothing has \n> > changed that needs to be announced, other then a few bugs fixed ...\n> \n> Maybe we should? The more publicity the better etc...\n> \n> Regards, Dave\n> \n> ---------------------------(end of broadcast)---------------------------\n> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org\n\nI agree it should be considered. It helps build mind share, which I\nthink we can all agree is somewhat lacking for PostgreSQL compared to\nMySQL. It helps build the impression that PostgreSQL doesn't just sit\nidle between major releases. It allows a potential user base to see\n\"PostgreSQL\" more frequently and build interest. It let's people know\nthat PostgreSQL is constantly being improved.\n\nMind share is a powerful thing and as any advertiser will tell you,\npress releases is one of the best ways to get the word out.\n\nGreg\n\n\n-- \nGreg Copeland <greg@copelandconsulting.net>\nCopeland Computer Consulting\n\n", "msg_date": "18 Dec 2002 10:15:12 -0600", "msg_from": "Greg Copeland <greg@CopelandConsulting.Net>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "\"Marc G. Fournier\" <scrappy@hub.org> writes:\n> On Wed, 18 Dec 2002, Dave Page wrote:\n>> Maybe we should? The more publicity the better etc...\n\n> The problem is that there is nothing to announce ... \"Hi, we fixed some\n> bugs\"? :) minor releases don't have any features added to them, so isn't\n> really news worthy ... :(\n\nI think this is exactly the difference between \"press release\" and\n\"technical announcement\" that Marc was getting beat up on just a couple\nweeks ago. A bug-fix-only update is not worthy of a press release.\nIt is worthy of a technical announcement --- which is exactly what Marc\nplans to push out to pgsql-announce as soon as the FTP mirrors are up\nto date.\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 18 Dec 2002 11:33:44 -0500", "msg_from": "Tom Lane <tgl@sss.pgh.pa.us>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ... " }, { "msg_contents": "Oleg Bartunov wrote:\n> > > > No, we don't do that with minor releases ... nothing has\n> > > > changed that needs to be announced, other then a few bugs fixed ...\n> > >\n> > > Maybe we should? The more publicity the better etc...\n> >\n> > The problem is that there is nothing to announce ... \"Hi, we fixed some\n> > bugs\"? :) minor releases don't have any features added to them, so isn't\n> > really news worthy ... :(\n> >\n> \n> I think changing major version of libpq is an important thing to be\n> mentioned.\n\nIt is mentioned in the release notes text:\n\n Migration to version 7.3.1\n \n A dump/restore is *not* required for those running 7.3. However, it\n should be noted that the main PostgreSQL interface library, libpq, has\n a new major version number for this release, which may require\n recompilation of client code in certain cases.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 18 Dec 2002 14:33:54 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "Peter Eisentraut wrote:\n> Bruce Momjian writes:\n> \n> > A dump/restore is *not* required for those running 7.3. However, it\n> > should be noted that the main PostgreSQL interface library, libpq, has\n> > a new major version number for this release, which may require\n> > recompilation of client code in certain cases.\n> \n> s/certain/all/\n\nI was unclear on that. If they install right over their existing\npgsql/lib directory, the old libpq will still be there, so a recompile\nwill not be required.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 18 Dec 2002 14:52:40 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "Bruce Momjian writes:\n\n> A dump/restore is *not* required for those running 7.3. However, it\n> should be noted that the main PostgreSQL interface library, libpq, has\n> a new major version number for this release, which may require\n> recompilation of client code in certain cases.\n\ns/certain/all/\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Wed, 18 Dec 2002 20:59:16 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "If it's all, perhaps we should reword as:\n\n... has a new major version number for this release and will require\nrecompilation of client code.\n\nRobert Treat\n\nOn Wed, 2002-12-18 at 14:59, Peter Eisentraut wrote:\n> Bruce Momjian writes:\n> \n> > A dump/restore is *not* required for those running 7.3. However, it\n> > should be noted that the main PostgreSQL interface library, libpq, has\n> > a new major version number for this release, which may require\n> > recompilation of client code in certain cases.\n> \n> s/certain/all/\n> \n\n\n\n", "msg_date": "18 Dec 2002 15:11:47 -0500", "msg_from": "Robert Treat <xzilla@users.sourceforge.net>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "On Wed, 18 Dec 2002, Bruce Momjian wrote:\n\n> Peter Eisentraut wrote:\n> > Bruce Momjian writes:\n> >\n> > > A dump/restore is *not* required for those running 7.3. However, it\n> > > should be noted that the main PostgreSQL interface library, libpq, has\n> > > a new major version number for this release, which may require\n> > > recompilation of client code in certain cases.\n> >\n> > s/certain/all/\n>\n> I was unclear on that. If they install right over their existing\n> pgsql/lib directory, the old libpq will still be there, so a recompile\n> will not be required.\n>\n\nIt's not always safe to install over existing previous installation.\nI think special mention about recompiling DBD::Pg would be important\n\n\n>\n\n\tRegards,\n\t\tOleg\n_____________________________________________________________\nOleg Bartunov, sci.researcher, hostmaster of AstroNet,\nSternberg Astronomical Institute, Moscow University (Russia)\nInternet: oleg@sai.msu.su, http://www.sai.msu.su/~megera/\nphone: +007(095)939-16-83, +007(095)939-23-83\n\n", "msg_date": "Wed, 18 Dec 2002 23:55:54 +0300 (MSK)", "msg_from": "Oleg Bartunov <oleg@sai.msu.su>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "Oleg Bartunov wrote:\n> On Wed, 18 Dec 2002, Bruce Momjian wrote:\n> \n> > Peter Eisentraut wrote:\n> > > Bruce Momjian writes:\n> > >\n> > > > A dump/restore is *not* required for those running 7.3. However, it\n> > > > should be noted that the main PostgreSQL interface library, libpq, has\n> > > > a new major version number for this release, which may require\n> > > > recompilation of client code in certain cases.\n> > >\n> > > s/certain/all/\n> >\n> > I was unclear on that. If they install right over their existing\n> > pgsql/lib directory, the old libpq will still be there, so a recompile\n> > will not be required.\n> >\n> \n> It's not always safe to install over existing previous installation.\n\nUh, that's what we recommend right now. How is it wrong?\n\n> I think special mention about recompiling DBD::Pg would be important\n\nUh, we don't distribute DBD:Pg.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 18 Dec 2002 16:25:23 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "Bruce Momjian writes:\n\n> I was unclear on that. If they install right over their existing\n> pgsql/lib directory, the old libpq will still be there, so a recompile\n> will not be required.\n\nThat's kind of like saying, if you keep using PostgreSQL 7.2 then a\ndump/restore will not be required. ;-) Installing new code directly over\nold one without deleting is an extremely bad strategy.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Thu, 19 Dec 2002 00:41:51 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "Peter Eisentraut wrote:\n> Bruce Momjian writes:\n> \n> > I was unclear on that. If they install right over their existing\n> > pgsql/lib directory, the old libpq will still be there, so a recompile\n> > will not be required.\n> \n> That's kind of like saying, if you keep using PostgreSQL 7.2 then a\n> dump/restore will not be required. ;-) Installing new code directly over\n> old one without deleting is an extremely bad strategy.\n\nWell, our release notes say:\n\n\tA dump/restore is *not* required for those running 7.3. \n\nand INSTALL says:\n\n 4. Installing The Files\n Note: If you are upgrading an existing system and are going to\n install the new files over the old ones, then you should have\n backed up your data and shut down the old server by now, as\n explained in the Section called If You Are Upgrading above.\n\nIn fact, we are a unclear about exactly when you need a dump/reload and\nwhen you don't. Seems the INSTALL file should have a clearer setup for\nfolks upgrading from 7.3 to 7.3.1.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n pgman@candle.pha.pa.us | (610) 359-1001\n + If your life is a hard drive, | 13 Roberts Road\n + Christ can be your backup. | Newtown Square, Pennsylvania 19073\n", "msg_date": "Wed, 18 Dec 2002 18:44:27 -0500 (EST)", "msg_from": "Bruce Momjian <pgman@candle.pha.pa.us>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "Shridhar Daithankar wrote:\n> On 18 Dec 2002 at 8:54, scott.marlowe wrote:\n> > \n> > www.linuxtoday.com has weekly updates from many gnu / OSS projects which \n> > are far less interesting than our 7.3.1 release is. I could see posting a \n> > minor upgrade release notice there and on other OSS news web site \n> > (freshmeat, slashdot, etc...)\n> \n> I read linuxtoday and /. daily. Linuxtoday is OK but /. would be an almost \n> waste of another 800 comments of postgresql v/s mysql.\n\nThat's okay, another 800 comments of PostgreSQL vs MySQL on Slashdot\nwould increase the load on Slashdot's MySQL server, thus making it\nslower and illustrating the point that they should be using PostgreSQL\ninstead. [ ducks ]\n\n:-)\n\n\n-- \nKevin Brown\t\t\t\t\t kevin@sysexperts.com\n", "msg_date": "Wed, 18 Dec 2002 20:25:44 -0800", "msg_from": "Kevin Brown <kevin@sysexperts.com>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." }, { "msg_contents": "Bruce Momjian writes:\n\n> Note: If you are upgrading an existing system and are going to\n> install the new files over the old ones, then you should have\n> backed up your data and shut down the old server by now, as\n> explained in the Section called If You Are Upgrading above.\n>\n> In fact, we are a unclear about exactly when you need a dump/reload and\n> when you don't. Seems the INSTALL file should have a clearer setup for\n> folks upgrading from 7.3 to 7.3.1.\n\nThere's a hidden \"...backed up your data, if required, ...\" in there that\nonly the author can see. ;-) But I think the section called If You Are\nUpgrading makes it rather clear when a backup is required.\n\n-- \nPeter Eisentraut peter_e@gmx.net\n\n", "msg_date": "Fri, 20 Dec 2002 00:45:15 +0100 (CET)", "msg_from": "Peter Eisentraut <peter_e@gmx.net>", "msg_from_op": false, "msg_subject": "Re: v7.3.1 tar ready ... please check it ..." } ]
[ { "msg_contents": "> I have made the change and am just building v7.3.1 right now ...\n> should be\n> available in a few minutes, and I'll announce it this evening as being\n> available ... can you grab a copy and make sure that it works as\n> expected?\n\nIt works fine for me.\n\n --Nate\n", "msg_date": "Wed, 18 Dec 2002 10:27:05 -0600 (CST)", "msg_from": "Nathan Mueller <nmueller@cs.wisc.edu>", "msg_from_op": true, "msg_subject": "Re: 7.3.1 stamped" } ]