text
stringlengths
0
1.49k
**Sam Boyer:** I mean frankly, I'm pretty roses and unicorns and rainbows about the possibility of this whole problem space anyway; I think there's a ton of different things we can do, and looking at using some sort of distributed storage for tracking people's releases is definitely on that list of cool stuff.
**Erik St. Martin:** Alright, so I think it's time for our first sponsor break. Our first sponsor for today is Toptal.
**Break:** \[20:10\] to \[\\00:20:43.13\]
**Erik St. Martin:** And we are back, talking to Sam Boyer about the GPS tool and some of the new dependency management stuff that the community's been working on. We've talked a bit about how GPS works and the motivation for that... What I'm interested to talk about a little bit is the dep tool itself. Is this created...
**Sam Boyer:** The single thing that GPS and therefore dep does that other tools don't really do is it's a real solver, which means that it's possible for us to have this useful pattern where each different project in the dependency graph get to say, "These are the constraints on the versions of my dependencies that I ...
This is a constraint-solving problem, it is NP-complete, it's nasty, but having that basic capability is -- the essential reason that that's so important is it lets each different author who's working on their own project make choices about like "These are the versions that I know that I work with, and I, as the person...
It doesn't force us to respect what everyone else says about their versions of dependencies to use - there is override behavior in everything - but it gives this sort of same default where we get to all collaborate on the hard problem in general that is "Which version of my dependencies do I use?" It gives us a fully-f...
**Erik St. Martin:** Yeah, I was gonna ask that same question, because a lot of people, especially if you're using packages that aren't constantly maintained, some people will put hard versions rather than loose for semver, and when you get up to your level, it can't solve whether there was an easy way to just be like,...
**Sam Boyer:** Yes, there's an override behavior. If it exists today, it works today. The rest of it to say though that we've aggressively invested effort in trying to make this as simple an experience as possible, which is to say we're trying to pair out as much of the command set as we can, we're trying to keep the d...
\[24:14\] There's a lot of commonalities between these systems, but the thing that is most unique about Go and dep in particular is that in these other systems their manifest files declare both the constraints concept and the requires concept. You have to list something in the manifest in order for it to show up, and i...
In dep it's different. The thing that determines whether or not something shows up is the import graph. We statically analyze the code, because this is Go and we can do this. We statically analyze the code, we see which imports are there, and that's what determines whether or not something shows up. You can then... Wha...
**Brian Ketelsen:** I hate fiddling. So I'm gonna beat a dead horse, because it's what I do... This tool sounds pretty awesome, and I have to admit that the GopherCon website code, which is on GitHub, is now vendored with dep, and the first time I used it, it worked beautifully. I didn't have any issues, it just worked...
**Sam Boyer:** Cool!
**Brian Ketelsen:** ...so kudos for that. That's kind of awesome, considering how new it is. But it occurs to me that unless people tag and make real releases, we're not gonna be any better off even with an official dep tool; we're still out in the wild, wild west.
**Sam Boyer:** Yeah, I think things are incrementally better, even with it. There's the possibility of doing -- well, I think we're incrementally better, but for the most part, yes, you're right... If people are not actually making releases, then there's not a lot of magic that we can do, which is part of the reason th...
**Brian Ketelsen:** Well, it reminds me of Dave Cheney's proposal (was it a year or two ago?) where he suggested that Go adopt officially a versioning standard. The response from the community was overwhelmingly "Why?" But as usual, Dave was forward-thinking, and it's kind of disappointing that we're so far behind the ...
**Sam Boyer:** My take on that long, painful thread was that we had a chicken or egg problem. There was no reason for people to start using tagging until there was a tool that actually really worked effectively with semver. Then there's no reason to write a tool, unless people have already made the release. So we're in...
**Brian Ketelsen:** \[28:03\] So if we were to give somebody a platform at GopherCon, say a stage in front of 1,500 people, live stream to the world, do you think we could make some change?
**Sam Boyer:** I think we could make some change.
**Brian Ketelsen:** Interesting. Not that we have a platform to give, but just hypothetically... \[laughter\]
**Sam Boyer:** This sounds totally hypothetical and not at all resembling the universe that we live in, yeah.
**Carlisia Thompson:** Sam, I wanted to ask you to talk about the vendor directory. I'm looking through Ed Muller's [blog post](http://icanhazdowntime.org/post/dep-101/) "I can haz downtime", and he describes here how you do `dep ensure`, and that basically creates a vendor directory which is mostly what most people do...
**Sam Boyer:** I think that the actual structure of files on disk is not gonna change. We're still gonna need a vendor directory; those semantics are not gonna change any time soon. The thing that's worth noting about it - and this is in at least one of the issues... I wanna write a script that can go and count the num...
So in one of these issues that is giving me brain fog, at the top of it we note that you should think of the vendor directory as an implementation detail, not as the actual target that we're going for. This is sort of way off into the future, so I don't wanna dwell too much on it, but we think it's possible to have an ...
**Erik St. Martin:** Is that a new verb now? We're just gonna call it left-padding?
**Sam Boyer:** Left-padding... I mean, I've given presentations where I certainly use it as a verb, so I think it counts. Does that answer your question, Carlisia?
**Carlisia Thompson:** Yeah, and also the vendor directory - is it going to be flattened?
**Sam Boyer:** Yes. Yes, always, and aggressively. Vendor is volatile, and if you put stuff in there that the tool didn't put in there, it will blow it away and not apologize.
**Erik St. Martin:** See, that's always good too, because people like to patch things.
**Sam Boyer:** No, there's no apologies for that. The only thing with treating Vendor as volatile that I have concerns about is some things related to code generation. I've seen a couple of issues raised for the last year with this in Glide - I can't remember it offhand. I have a list that might have the issue in it, b...
**Carlisia Thompson:** \[32:06\] I don't wanna jump ahead too much and I don't even know if you can answer this question, but when are we gonna have this tool? \[laughs\]
**Sam Boyer:** Right! This is the most important question, right? The committee's been talking with a bunch of people; we've been talking with [Russ](https://twitter.com/_rsc) as well, and the goal is -- and I have this roadmap that I've been working on, which I was hoping to publish by today; I'm still sort of tweakin...
Other stuff is important, but we can sort of continue to work on it post-merge into the toolchain. That is gonna be a hard process. Despite the fact that this is at [GitHub/golang/dep](https://github.com/golang/dep), that does not mean that this is blessed or that is foreordained that this will make it into the toolcha...
**Erik St. Martin:** Yeah, I think anything at GitHub is production-worthy, so you shouldn't have put it there. \[laughter\]
**Sam Boyer:** You and half of the rest of the world. So it's fair... There is a reason we waited a while to release this. We wanted it to be at least runnable, despite the big letter warning that says "Don't commit the manifest and lock files that are generated by this!" People have already done it, including people o...
**Carlisia Thompson:** And is there a date for 1.10?
**Sam Boyer:** That will be like the end of this year, I believe.
**Carlisia Thompson:** Cool.
**Erik St. Martin:** I'm trying to remember their releases... One's in August, and the other one is -- I can't remember.
**Sam Boyer:** It will be this time next year.
**Brian Ketelsen:** Sorry, I was muted... It's February 2018, because we have six-month release cycles.
**Sam Boyer:** There you go.
**Carlisia Thompson:** And talking about help, talk to us about how is the process of people getting involved going to be like? For example, are you going to want to have issues opened before people jump in and start doing things? Because how are people even gonna know what is there to do? Do you have a process, do you...
**Sam Boyer:** Yes. The answer is yes to all of the possible things that I could need; I need all of them, that would be great. We need people to help with kicking the tires, we need people to help with writing docs, we need people to help with figuring out some of the design issues that we have, we need people to help...
The roadmap that I'm trying to get out there is intended to provide a generalized picture, so that people can read it and say, "Alright, I have a sense of where this is going and I have a sense of where I might direct my effort." From there we've got like a "Help wanted" and a good First PR label for our issues on GitH...
**Carlisia Thompson:** And this is probably the best way for people to get in contact for people who want to start helping...?
**Sam Boyer:** \[35:54\] Yeah. Right now the three ways in are 1) just install it, run dep on a project and do some things, run into some stuff, have some questions, post an issue... I mean, it's runnable enough now that you can really try it; you can kick the tires and just post an issue if you have a problem. You wil...
**Carlisia Thompson:** Cool.
**Erik St. Martin:** What do you envision for the future of GoDep? Do you think that will be in a place as a community with semver and things like that, and kind of agreement on the dep tool where if it comes as part of 1.10, community adoption would be mostly there before it hits the Go release? Or do you envision the...
**Sam Boyer:** Well, eventually the word "dep" is gonna go away. It's gonna become like -- now you do `dep init`, `dep ensure`, and it will probably become `go init`, `go ensure`. That's the current plan, anyway. We went back and forth over naming within the committee a lot. We've tried out some alternates, but ultimat...
I think it's gonna be a little of both. We're trying very hard to get this out there and to give people plenty of lead time to be able to test it out. A first major milestone for us really is the stabilizing of the manifest and mod files. Once we do that, people can start kind of using it for real, and the guarantee th...