text stringlengths 0 2.35k |
|---|
[1705.80 --> 1712.06] But you also cannot get the history of transactions that modify this particular key. |
[1712.06 --> 1715.96] And that is how we implemented temporary capabilities. |
[1715.96 --> 1722.68] So you can go back in time in the database and query the database as it was sometime ago. |
[1722.68 --> 1725.04] And without seeing newer changes. |
[1725.42 --> 1729.16] On top of this, we implemented SQL capabilities. |
[1729.16 --> 1741.08] So when you create an entry thinking in SQL, it ends up being a transaction that consists of key and value entries. |
[1741.08 --> 1749.20] So SQL, all the SQL changes or SQL data model is backed by a key value database. |
[1749.96 --> 1753.94] So actually, the same transaction is what is happening. |
[1754.10 --> 1759.10] So we are using the key value transactions to store transactions that happen in SQL. |
[1759.78 --> 1762.58] So SQL was added afterward. |
[1763.02 --> 1766.08] So it's possible to use both. |
[1766.08 --> 1778.64] So, of course, they are isolated entries that are inserted using the key value are not seeing the internal changes or internal entries that are working with SQL. |
[1779.02 --> 1781.36] But both data models are possible. |
[1781.98 --> 1786.38] The advantage of using SQL, of course, is easier to model your application. |
[1786.38 --> 1797.12] Because you have to think it's easier to work when for later on, define index for is efficiently querying the data, for writing queries, of course. |
[1797.84 --> 1802.44] But we also added the possibility to verify in SQL. |
[1802.66 --> 1804.30] So that is one of the differences. |
[1804.54 --> 1807.88] So you can get a particular row based on the primary key. |
[1808.46 --> 1810.02] And this entry will be very fine. |
[1810.64 --> 1815.74] So you're still able to model your application just like you would in a relational system. |
[1815.74 --> 1816.14] Exactly. |
[1816.38 --> 1821.84] It's just basically the encrypted storage that is used in the verifiability once you pull data out. |
[1822.00 --> 1829.44] All these things you're adding sort of on top of the good old model that most developers who built web applications and whatnot are familiar with for example. |
[1829.74 --> 1832.46] So let's talk about the operability of this. |
[1832.54 --> 1836.26] But before we jump into that, I see John, you've got a burning question you want to ask. |
[1836.58 --> 1837.60] I don't have a burning question. |
[1838.16 --> 1839.02] It's a question. |
[1839.02 --> 1850.40] I was going to say that like the SQL stuff reminds me of the first time I ran into a use case where I didn't necessarily need an immutable database, but I needed to mimic its functionality in some way. |
[1850.82 --> 1853.56] Basically, I was working on like shipping stuff with addresses and everything. |
[1853.56 --> 1858.30] And one of the things that came up where people were like, well, I want to be able to change these addresses I used to ship to things. |
[1858.30 --> 1870.10] And it became very clear that in a relational database, if you have a bunch of previous shipments that all associate with some record and then you change it, then all of a sudden your history is really weird at that point because that's not actually what those shipments were. |
[1870.10 --> 1880.34] And like seeing a database like this, it's kind of interesting to like, I think as developers, we encounter cases where immutable databases or something like that is useful. |
[1880.62 --> 1885.54] I mean, we all use package managers, which are another example of like not really having immutability. |
[1885.72 --> 1889.72] You can release a version, but once it's released to some package manager, you're kind of stuck. |
[1890.04 --> 1891.32] You know, you have to release a new one essentially. |
[1891.62 --> 1892.36] Or you're supposed to be. |
[1893.22 --> 1895.00] I mean, I don't want you to be able to change that. |
[1895.14 --> 1897.22] As I say, most package managers won't let you do it. |
[1897.22 --> 1915.40] So I think as developers, we use immutable systems at times, but we kind of like forget about it because I think a package manager is a great example of something that really benefits from something where you can verify nothing got changed because that would be really bad when you're downloading ... |
[1915.80 --> 1922.28] But it's also like interesting in the sense that I feel like most systems we work with that use immutability have some sort of scapegoat. |
[1922.28 --> 1928.50] The best example I can give is like get we always get where you can have the history and it's supposed to basically be immutable. |
[1929.02 --> 1934.48] But there's always ways to force changes into like to rewrite history, which is not necessarily a great thing, but it's possible. |
[1935.30 --> 1947.16] So knowing that developers at some point want to like rewrite history and stuff, do they have to come into using MUDB like like they can't come into it, I'm assuming, with the same mindset of like I can use this exactly like a SQL database. |
[1947.16 --> 1954.16] So are there any like tips or advice that sort of like help them get out of that mindset that you see people struggling with when they're starting? |
[1954.16 --> 1960.92] So in MUDB what actually you could think of is that you can change the data, right? |
[1960.96 --> 1967.20] You can do corrections, but what you will still get, you have this auditability of the history. |
[1967.60 --> 1971.08] So it's like not lying to anybody that I did not make a mistake. |
[1971.16 --> 1971.82] I did a mistake. |
[1972.08 --> 1973.46] I just corrected it right now. |
[1973.56 --> 1974.28] This is the state. |
[1974.62 --> 1975.86] But let me be clear. |
[1976.04 --> 1977.02] This is what we see. |
[1977.38 --> 1978.44] This is the current state. |
[1978.66 --> 1980.38] And previously it was something different. |
[1980.38 --> 1991.04] And also this example with changing the address, I think this is something very interesting because on the key value level inside MUDB, we have something like a reference to other key. |
[1991.68 --> 1998.66] So instead of getting some specific value, you just try to read it from other key and just forward it back. |
[1998.96 --> 2004.48] But what you can do is you can say that this reference is for that key at that transaction. |
[2005.02 --> 2009.02] So what it means is that it like freezes the value inside the history. |
[2009.02 --> 2018.48] So then you could create, let's say, a record that there was this kind of shipment to that person under that address at that point in time. |
[2018.92 --> 2020.08] So that is something unique. |
[2021.16 --> 2025.50] Also, I need to comment about one thing, this package managers. |
[2026.42 --> 2027.10] Let me say that. |
[2027.34 --> 2027.62] Please do. |
[2028.18 --> 2034.86] We have actually been using this immutable databases, but we just don't know it or just forgot about this. |
[2034.86 --> 2038.98] And a very good example is actually GoMod proxy. |
[2039.66 --> 2044.60] And actually the technology behind GoMod proxy is very similar to what we have. |
[2044.68 --> 2046.16] It's this kind of immutable ledger. |
[2046.76 --> 2052.96] And actually we had the situation where we released some kind of tag, some version of MUDB. |
[2053.28 --> 2056.94] And once somebody just fetches it through the proxy, it's set in stone. |
[2057.12 --> 2058.66] You cannot change it. |
[2058.66 --> 2061.14] You cannot switch the tag to something else. |
[2061.38 --> 2063.40] And very weird situations happen. |
[2063.62 --> 2066.04] And actually this is for the security reasons. |
[2066.30 --> 2073.90] So if you release something, then everybody who downloads this particular version will only get this version of the code. |
[2074.22 --> 2077.02] You have to release a new patch version. |
[2077.40 --> 2078.42] And that's actually really good. |
[2078.42 --> 2080.74] And that's good about the security. |
[2081.24 --> 2082.12] I agree that that's good. |
[2082.30 --> 2095.16] I guess I would imagine it would make adoption harder in the sense that developers are just weird about, like if somebody releases an invalid package and they want to pull it back real quickly, they're still weird about like now I have to increment the version and they don't want to do that. |
[2095.34 --> 2099.44] For whatever reason, mentally, they're like, I don't want to admit I made a mistake and show that to people. |
[2099.76 --> 2100.12] Too bad. |
[2100.12 --> 2103.88] So like, does that make adoption harder when like you're basically forcing them to do that? |
[2104.10 --> 2109.34] In this case, in MUDB, you have to convince every other client. |
[2110.18 --> 2120.86] If you want to roll back the history in MUDB, you will have to convince every auditor or client that already have that register, that state locally. |
[2121.44 --> 2122.50] That's the only option. |
[2122.74 --> 2124.00] Measure twice, cut once. |
[2124.00 --> 2129.18] But I think that's really making a mistake is not something huge. |
[2129.26 --> 2131.08] And we all made mistakes. |
[2131.58 --> 2135.46] And like in real life, there is always an option to correct the mistake. |
[2135.62 --> 2136.10] For example? |
[2137.12 --> 2137.54] Yeah. |
[2137.80 --> 2138.52] Like releasing. |
[2138.74 --> 2139.24] Let's hear it. |
[2139.72 --> 2142.04] Releasing a package that contains some bug. |
[2142.42 --> 2142.98] Right, right. |
[2143.14 --> 2145.24] And then why should we be ashamed of that? |
[2145.24 --> 2154.16] And actually, I see that people who can say that they made a mistake and they corrected that, they tend to deal with those issues better than trying to hide it. |
[2154.66 --> 2155.98] So I would go that way. |
[2156.52 --> 2157.16] That makes sense. |
[2157.36 --> 2161.62] I mean, I guess there are definitely cases where it makes sense to want to delete things. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.