text stringlengths 0 2.08k |
|---|
**Andrew Nesbitt:** That's one reason why I haven't started pull download stats for libraries, because you can't compare across different package managers either, because the client may cache really aggressively. RubyGems really aggressively caches every package, whereas if people are kind of blasting away their Node m... |
**Mikeal Rogers:** \[28:01\] I think a lot of the metrics fall into that, though. When you start looking at them across ecosystems, they really don't match up. The one that I think of comparing a lot is Go and npm. GoDoc is actually like a documentation resource, it's not really a package manager, but people essentiall... |
**Andrew Nesbitt:** Yeah, the Go package manager is slightly strange because it's so distributed. It's just, give it a URL and that is the package that it will install, so basically every nested filed inside that package could be considered to be a separate thing, because it's just a URL that points to a file of the in... |
**Mikeal Rogers:** I'd like to get into the human side of this, too. You've mentioned this a little bit earlier when you were talking about the difference between npm and Ruby in terms of locking down your dependencies. That's not enforced by the package manager, it's just now a cultural norm to use Bundler and not npm... |
**Andrew Nesbitt:** I've tried not to look too much into the people yet, partly because I didn't wanna end up pulling a lot of data that could be used by recruiters, and make libraries a source of kind of horrible data that would abuse people's privacy. |
**Mikeal Rogers:** I didn't mean like individuals, I meant like culturally. I didn't mean like, "Be creepy." \[laughs\] |
**Andrew Nesbitt:** \[inaudible 00:29:55.07\] all kinds of horrible things. Nothing springs to mind... I guess you can look at the average number of packages that a developer in a particular language or package manager would potentially publish more, or the size of different packages. Node obviously tends towards small... |
I haven't done too much in comparing the different package managers from that perspective, because it felt like... As you said, you don't get much mileage from going like "What this thing compared to this thing?" It's much better to look at what packages can we highlight as interesting or important within a particular ... |
You definitely see a certain amount of people who live in more than one language as well. It's not often that there's people that are just only doing one particular language. |
**Nadia Eghbal:** I'm curious whether there's - I don't know a whole lot about this, but if there's any way to standardize how package managers work across languages, or just standardize behavior somehow. Because I just sort of think for people that are coming for this from outside of open source, but are really curiou... |
**Andrew Nesbitt:** \[32:14\] Yeah. I've standardized as much as possible with Libraries. The only way I could possibly collect so many things is to kind of go, "Let's treat every package manager as basically the same, and if they don't have a particular feature then that's just 'no' for that particular package manager... |
But you can then kind of go, "Okay, within a given package manager, show me the things that are highly depended on but only have one contributor, or have no license", which is easy to pull out in Go, but then "Order by the number of people that depend on it or the number of releases that it's had" to try and find the p... |
**Mikeal Rogers:** Right. I can see you kind of standardizing the data and some of the people work, but the actual technology - or even the encapsulation - you eventually hit the barrier of the actual module system itself, right? One of the reasons why Node is really good at this is because npm was built and the Node m... |
So there's a general trend I think towards everything getting smaller and packages are getting smaller, but some module systems actually don't support that very well, and you're hitting kind of a bottleneck there. |
**Andrew Nesbitt:** Yeah, I don't think there are many other package managers other than npm that allow you to run multiple versions of a package at the same time, and partly because of the danger of doing that, that you introduce potentially really subtle bugs in the process. But most of the package managers in the la... |
**Mikeal Rogers:** Yeah, it's important to note that's not part of npm, it's part of Node. Node's resolution semantics enable you to do that; it's not actually in npm. npm is just the vehicle by which these things get published and put together. |
I think there's been valiant efforts to make an installer and an npm-like thing in Python, and they eventually hit this problem where you actually need to change the module system a bit. |
**Andrew Nesbitt:** Yeah, I made a shim for RubyGems once that essentially did that and it made a module of the name and the version, and then kind of hijacked the require in Ruby. It was a fun little experiment, but ends up being... You're just fighting against everything else that already exists in the community. So ... |
**Mikeal Rogers:** \[36:00\] In that vein, have you seen any changes across these module systems as they've gone along? Have any really spiked in popularity or fallen? Are there changes that actually happen in these ecosystems once they get established? |
**Andrew Nesbitt:** Not so much. Elixir is making a few small changes, but it's more around how they lock down their dependencies. Usually once there's a few hundred packages - and often it's because I guess there's just not many maintainers that are actually working directly on the package managers; often they're comp... |
**Mikeal Rogers:** It's true... |
**Andrew Nesbitt:** ...and these package managers go round and eventually run into the same problems and solve the same things over again. |
**Nadia Eghbal:** Related to that - I'm curious for both Andrew and Arfon - when we talked about looking at stars versus looking at downloads, and looking at projects that are trending or popular versus ones that are actually being used, for someone who's trying to look through available projects and fair out which one... |
**Andrew Nesbitt:** Yes, tricky one. I've been kind of intentionally avoiding highlighting the new, shiny things in package managers for the moment, and kind of not doing any newsletters of "Here are the latest and greatest things that have been published." I think this mirrors my approach to software at the moment, wh... |
But that's just my point of view. There are lots of people who are looking for employment and want to be able to keep on top of whatever is currently the most likely to get them a job, which is a very different view of "What should I look at? What should I use?" |
**Nadia Eghbal:** Something I really struggle with software in general, you often hear people saying, "Oh, this project should just die, because it's not following modern development practices, or it's just kind of hopeless and we should just focus on whatever is new." I think it's because it's comparatively easier to ... |
**Andrew Nesbitt:** Yeah, and sometimes it's a case of people very loudly saying, "I'm not gonna use this anymore", whereas there are a number of people that are just using it and not telling anyone, just getting on with what they're doing. They still require that stuff. Often you see companies will have their own priv... |
**Arfon Smith:** \[39:54\] Right. I actually think this is one of the things where conventions can really help. I still recommend Rails to people who are getting into web development, because when you do Rails new, it comes with an opinion on what your web server should be, what your testing framework should be, what J... |
I relatively recently started doing some Node stuff and I wanted to find a testing framework; I just wanted to write some tests, and I ended up going through about six in about five hours and it seemed by my assessment of what's going on, the community was moving so quickly - three of the frameworks are all written by ... |
I consider myself reasonably technical and pretty good at using GitHub hopefully, and I found it hard to find a good set of defaults. I don't know, I think finding the right thing, it's... |
**Andrew Nesbitt:** It's very similar in the browser at the moment. It's hard to know - is this library the right thing anymore? I find myself going to, and I use DotCom to work out, like "Is this mirroring and API that now is a standard, or has it moved on?" because the browser has been evergreen mix, everything reall... |
**Arfon Smith:** Yeah, I don't know... The other thing is if you actually went out, stick your neck out and say "You should use these things" then somebody's obviously gonna shout at you on the internet and say "You're an idiot. You should use this thing." I think it's hard for the individual to have a strong preferenc... |
**Mikeal Rogers:** The scary thing to me is that there is no correlation that I can find between the health of a project and a popularity of a project. |
**Nadia Eghbal:** Yes! |
**Mikeal Rogers:** It's totally fine if it's not the coolest thing, but people are still working on it and it's still maintained. But things actually die off and the maintainer leaves and it's still popular and still out there, and still being heavily used because it's that thing that people find. But as you said, that... |
**Nadia Eghbal:** And making that argument to someone that... They might not care about the health, because they're like, "Well, it's popular and everyone's using it." I struggle with sort of like what is a good argument for saying "You should care about this" to a user. |
**Arfon Smith:** Yeah, it's a very long-term thing as well, because if you get an instant result and you can ship it and be done, you're like "Oh, that's fine, I don't need to come back and look at this again", whereas in six months, a year's time you might come back to it and be like "Oh, I wish I didn't do this." But... |
**Andrew Nesbitt:** Yeah, I feel like there was a thing relatively recently on Hacker News, like "Commiserations, you've now got a popular open source project", or something like that. It was this really well-articulated overview of, so you publish something on GitHub; now a bunch of people are using it, and now you've... |
\[44:06\] For me that's just a good demonstration of, you know, lots of people publish open source code, and they're doing that because that's just normal, or maybe they're doing that because that's the free side of GitHub, or whatever the reason is they're doing that; or they're solving probably their own problems - t... |
I think a lot of people who publish something that then becomes popular maybe don't want to maintain it, or maybe don't have the time to maintain it. Money helps, I think, but I think funding open source is hard; for lots of people it isn't their day job to work on these things, and I think there's not a good way yet -... |
Nadia, you've written a bunch about this, I'm sure you've got a bunch of opinions on this as well. |
**Nadia Eghbal:** Yeah. |
**Mikeal Rogers:** I think that you've really highlighted the basis for the shift in open source, which is that we've gone to a more traditional peer production model. If you read anything from Clay Shirky about peer production, it's like you publish first and then you filter, and the culture around how you filter and ... |
And in older open source, in order to get involved at all it was so hard, that you basically internalized all of that culture and then basically became a maintainer waiting in the wings, and that's just not the world anymore. |
People publish that have no interest in maintaining things at all, because everybody just publishes, that's the culture now. I think we're actually gonna come into a break now, but when we get back we're gonna dive into what are those metrics of success, what are those metrics of health and how can we better define thi... |
**Break:** \[46:36\] |
**Mikeal Rogers:** \[48:49\] And we're back. Alright, so let's dive right into this. What are the metrics that we can use for success? How can we use this data to show what the health of an open source project might be and expose that to people? Let's start with Arfon, since we have so many new metrics coming out of th... |
**Arfon Smith:** Yeah, so I'll start by not answering your question directly, if you don't mind. One thing I would love to see is... There are things that I can do, and anybody who's looked at an enough open source software... If you give somebody ten minutes, "Tell me if this project is doing well", you can answer tha... |
What that actually means in terms of heuristics, the ways in which we could codify that in terms of hard metrics, I think that's a reasonably tough problem. I don't think it's impossible by any stretch, but it's things like - we could make some up right now. Like, are there commits coming and landing in master? Are pul... |
**Mikeal Rogers:** Yes, yes! |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.