[00:00.000 --> 00:08.160]  This is actually my first time at FOSDEM.
[00:08.160 --> 00:12.520]  I've been meeting to come here for many, many years,
[00:12.520 --> 00:15.720]  and Saul Lorenzo been bothering me that I should come here and speak.
[00:15.720 --> 00:18.360]  So I'm glad, actually, I finally made it.
[00:18.360 --> 00:23.960]  If you, I run a blog, WebRC hacks for developers.
[00:23.960 --> 00:28.880]  I got a lot of guest authors and I hope readers in the audience here.
[00:28.880 --> 00:33.800]  You might also recognize me if you watch YouTube videos about WebRC stuff.
[00:33.800 --> 00:37.560]  I do an event series and videos at Cranky Geek.
[00:37.560 --> 00:43.880]  But today, I'm here to talk about some trends I did by
[00:43.880 --> 00:46.360]  analyzing GitHub and similar repositories.
[00:46.360 --> 00:47.080]  I'll talk about that in a second.
[00:47.080 --> 00:48.640]  But before we start,
[00:48.640 --> 00:52.600]  what are some things that would be nice to know what's going on in the tree?
[00:52.600 --> 00:56.240]  I'm an analyst like to watch this stuff and write about it.
[00:56.240 --> 00:58.800]  So one, is the community still growing?
[00:58.800 --> 01:00.440]  What are some of the interesting projects?
[01:00.440 --> 01:02.160]  What are some of the new trends?
[01:02.160 --> 01:06.080]  Are people using things like WebCodex yet or not?
[01:06.080 --> 01:08.720]  So how do you go about doing that?
[01:08.720 --> 01:12.360]  Well, I came up with a, I have a methodology.
[01:12.360 --> 01:17.120]  It's largely based on BigQuery and there's a bunch actually,
[01:17.120 --> 01:22.680]  a bunch of providers give data to BigQuery or make their datasets public available there.
[01:22.680 --> 01:23.680]  So you can grab that.
[01:23.680 --> 01:25.720]  GitHub is definitely the best one,
[01:25.720 --> 01:28.560]  basically any time, any public repo,
[01:28.560 --> 01:32.360]  any time you do any kind of commit pull request issue that all gets
[01:32.360 --> 01:36.400]  sent and stored in a BigQuery database.
[01:36.400 --> 01:40.000]  There's actually some other datasets that are cool in there too.
[01:40.000 --> 01:43.480]  I've used in the past and I actually use Stack Overflow in this one.
[01:43.480 --> 01:48.000]  And I often cross-reference that with other sources like the Chrome platform status is
[01:48.000 --> 01:51.760]  a good way to see what actual APIs are being used,
[01:51.960 --> 01:54.280]  at least in the Chrome world.
[01:55.640 --> 01:58.240]  So you get all that data.
[01:58.240 --> 02:03.560]  I actually like to use Colab, which is just a Google's hosted
[02:03.560 --> 02:07.440]  Jupyter notebook type ecosystem to do the analysis and follow things.
[02:07.440 --> 02:11.520]  And then when I get frustrated about coding and doing stuff like
[02:11.520 --> 02:17.200]  making charts with Python, then I fall back to Excel for quick analysis.
[02:17.200 --> 02:19.960]  So some of the hard things about this and important to keep in mind about
[02:19.960 --> 02:21.480]  the analysis, you can't see it there.
[02:21.480 --> 02:26.080]  But all this is really based on regular expressions and pulling out
[02:26.080 --> 02:34.240]  keywords to identify a repo as WebRTC or something as one term or another.
[02:34.240 --> 02:37.920]  So that obviously has some flaws because there'd be false hits in there and
[02:37.920 --> 02:39.400]  you have to be careful with your selection.
[02:39.400 --> 02:41.760]  So a lot of the time I spend is actually just going through looking at
[02:41.760 --> 02:43.400]  the data to make sure it's not junk.
[02:43.400 --> 02:46.240]  And I found in the past, there's a lot of bots.
[02:46.240 --> 02:50.000]  So you gotta remove those things and these data sets themselves aren't perfect.
[02:50.000 --> 02:52.040]  There's always missing data or junk data.
[02:52.040 --> 02:54.320]  So keep that in mind.
[02:54.320 --> 02:58.440]  But I've been doing this occasionally more frequently now.
[02:58.440 --> 03:01.840]  Since 2015, I've been tweaking the methodology along the way.
[03:01.840 --> 03:06.640]  It's held up so far, but always open to feedback.
[03:06.640 --> 03:08.000]  So let's just dive in.
[03:08.000 --> 03:09.800]  How are we doing here in WebRTC?
[03:11.120 --> 03:13.640]  So this just looks since 2019.
[03:13.640 --> 03:16.840]  I don't think it's a big surprise to anyone.
[03:16.840 --> 03:19.520]  But there was a peak during the pandemic.
[03:19.520 --> 03:23.920]  So you can see here, it's actually April of that year.
[03:23.920 --> 03:27.480]  And this particular graph shows distinct users, right?
[03:27.480 --> 03:32.400]  So anybody like anyone on GitHub, just based on unique GitHub IDs per
[03:32.400 --> 03:35.960]  period in there, so that was over 10,000 in April.
[03:35.960 --> 03:38.760]  But if you look here, actually, we're not doing so bad.
[03:38.760 --> 03:40.760]  This is bad data, I was missing that month.
[03:41.320 --> 03:47.640]  We're not actually doing so bad here, but we're only about 60% of past peak.
[03:47.640 --> 03:52.040]  So it's still pretty above where we were before the pandemic.
[03:53.800 --> 03:58.120]  But another thing I find actually very interesting is also because you can look
[03:58.120 --> 04:00.080]  at these events, who's actually contributing, right?
[04:00.080 --> 04:05.280]  And you can look at the blog, I'll have some links to more on the methodology.
[04:05.280 --> 04:08.560]  But essentially, people doing pull requests, pull requests reviews.
[04:09.560 --> 04:12.680]  That sort of thing that counts as a contributor, right?
[04:12.680 --> 04:17.920]  And actually contributors, if you believe me, are actually up more than ever.
[04:19.000 --> 04:22.840]  So first, one interesting point here you can see.
[04:22.840 --> 04:24.800]  So there was a peak number of users in April, but
[04:24.800 --> 04:27.880]  actually the peak contributions, it was in actually May.
[04:27.880 --> 04:30.040]  And maybe that makes sense, like people during the pandemic,
[04:31.520 --> 04:37.080]  got an issue in WebRTC, start looking at it, they maybe could star a repo,
[04:37.600 --> 04:40.880]  get involved, but then it takes them a few weeks actually before they actually
[04:40.880 --> 04:44.320]  contribute, start adding things into that repo.
[04:44.320 --> 04:49.200]  But as I said here, our most recent peak here was actually in August.
[04:49.200 --> 04:52.240]  And we're not actually too far off for the rest of the year looking at that.
[04:52.240 --> 04:57.440]  So I want to look into see what was going on and compare these two peaks.
[04:57.440 --> 05:03.120]  So one is this is actually from the August peak, I dug into the sea.
[05:03.120 --> 05:07.120]  Where's some of the repos that we're actually having, I had the most activity.
[05:07.120 --> 05:10.840]  And one of them, very popular WebRTC project is Coturn,
[05:10.840 --> 05:15.080]  the open source kind of stone and turn server.
[05:15.080 --> 05:18.800]  And actually one of my co-workers Gustavo took over that project.
[05:18.800 --> 05:21.120]  So I asked him what was going on here, what happened,
[05:21.120 --> 05:26.040]  why do we have such a big spike in this and wrote a whole blog post about it.
[05:26.040 --> 05:31.360]  But essentially, that project was kind of stale, not a lot of activity for
[05:31.360 --> 05:38.120]  a while, he and Pavel took that over and then started really get the community
[05:38.120 --> 05:40.960]  involved and there's a huge spike and things like that.
[05:40.960 --> 05:47.160]  So then I wondered, is basically is this, sorry,
[05:47.160 --> 05:51.720]  is the other peaks in August in here because of spikes?
[05:51.720 --> 05:53.680]  Or is there a lot of regular activity?
[05:53.680 --> 06:00.280]  You can see here, interestingly, you see things like AdGuard is always high in there.
[06:00.280 --> 06:03.560]  Like people actually plan to block WebRTC and its usage, right?
[06:03.560 --> 06:07.960]  But they have a lot of activity every month around that.
[06:07.960 --> 06:12.800]  But actually it wasn't actually, you see some commonality, but some difference here.
[06:12.800 --> 06:18.280]  And, sorry, but when you actually look at the distributions and
[06:18.280 --> 06:23.160]  the change between April of 2020 and August of 2022,
[06:23.160 --> 06:26.760]  the actual distributions across the top 10, top 20, top 100,
[06:26.760 --> 06:29.320]  they're actually not a whole lot different.
[06:29.320 --> 06:31.640]  So what does this all mean?
[06:31.640 --> 06:35.040]  It's like actually the WebRTC development actually is not really getting a lot more
[06:35.040 --> 06:36.080]  concentrated.
[06:36.080 --> 06:37.440]  You can look for a given period of time.
[06:37.440 --> 06:39.640]  Obviously some projects are doing more popular and
[06:39.640 --> 06:42.400]  doing well, have more activity than others.
[06:42.400 --> 06:46.400]  But overall, it's not like we're consolidating down to a few projects, right?
[06:46.400 --> 06:52.120]  It's the same kind of more equal distribution that's existed at least for
[06:52.120 --> 06:52.680]  several years now.
[06:53.520 --> 06:59.360]  So another data set, and this is actually a new one I hadn't looked at before,
[06:59.360 --> 07:01.640]  is Stack Overflow.
[07:01.640 --> 07:05.840]  So I zoomed in to take a look at that.
[07:05.840 --> 07:08.920]  And that's to see if this follows a similar pattern.
[07:08.920 --> 07:12.640]  Now bear in mind compared to the previous charts, this goes back all the way to
[07:12.640 --> 07:15.160]  2012, so it's a much longer data period.
[07:15.160 --> 07:20.360]  And you can see here, this is comments on Stack Overflow questions and
[07:20.560 --> 07:23.240]  actually the questions themselves.
[07:23.240 --> 07:28.480]  And unfortunately, you can't see the font too much of answers within Stack
[07:28.480 --> 07:32.960]  Overflow, but it essentially looks very similar to the questions side of things.
[07:32.960 --> 07:37.360]  And you can see very similar here, peak in April of 2020.
[07:39.480 --> 07:45.680]  Also, unlike the GitHub analysis, this actually shows a peak and
[07:45.680 --> 07:50.800]  is here also in April of 2022.
[07:50.800 --> 07:54.600]  I didn't have a chance to dig into to see what was driving that particular peak
[07:54.600 --> 07:55.720]  this year.
[07:55.720 --> 07:59.480]  But overall, I think you can see it's a little bit harder compared to the other
[07:59.480 --> 08:04.000]  one, but we're still generally up compared to prior to the pandemic in terms of
[08:04.000 --> 08:04.880]  questions and answers.
[08:04.880 --> 08:07.720]  And actually, it's a pretty good sign that there's a lot of activity there.
[08:09.640 --> 08:15.160]  And I also just took a look to see as a percentage of all the questions on
[08:15.160 --> 08:19.560]  Stack Overflow, what percentage of them had at least something that mentioned
[08:19.560 --> 08:21.640]  WebRTC or one of these terms?
[08:22.640 --> 08:25.440]  And very surprising, actually, it's actually very high.
[08:27.120 --> 08:33.120]  So it's something like one in, during the pandemic,
[08:33.120 --> 08:37.400]  it was one out of every 1400 questions on Stack Overflow had something that
[08:37.400 --> 08:40.640]  mentioned WebRTC, which that seems like quite a bit because I still consider
[08:40.640 --> 08:43.560]  WebRTC kind of a very niche sort of thing.
[08:43.560 --> 08:49.680]  And even if you look today, just in the last data point in this one is November,
[08:49.680 --> 08:55.720]  at that point it was still one in, I'm sorry, it was one in 900 during the peak
[08:55.720 --> 08:58.560]  of April of 2020.
[08:58.560 --> 09:01.880]  It's still one in 1400 today, which was still actually very high.
[09:01.880 --> 09:04.400]  So you can see, you can interpret this two ways.
[09:04.400 --> 09:07.760]  One, WebRTC is very confusing and people have a lot of questions.
[09:07.760 --> 09:10.240]  So you need to comment on it, or you can also see there's a lot of people involved.
[09:10.240 --> 09:12.960]  I think both are good, right?
[09:15.460 --> 09:15.960]  But yeah.
[09:17.760 --> 09:22.920]  So also very interesting that can we look at this data set to understand
[09:22.920 --> 09:26.600]  development trends, what's going on in the market.
[09:26.600 --> 09:29.720]  And one of the very interesting things I always like to look at is what are some of
[09:29.720 --> 09:33.840]  the language trends, programming languages that people are using.
[09:33.840 --> 09:38.360]  And this is a jumble and hard to see, so let's actually zoom in.
[09:38.360 --> 09:43.320]  So one of the ones I've been trying to track for a while is JavaScript versus TypeScript.
[09:43.320 --> 09:50.760]  I've been delaying, converting to TypeScript and I'm kind of wondering, do I need to, is
[09:50.760 --> 09:54.240]  it time for me to really switch over or can I wait a little bit longer?
[09:54.240 --> 09:57.840]  You can see here, well, obviously TypeScript's been getting more popular.
[09:57.840 --> 10:03.160]  We just, you know, just in December reached the 50-50 point, right?
[10:03.160 --> 10:06.720]  Where, you know, all these repos where TypeScript's half.
[10:06.720 --> 10:09.480]  So I think I'm probably behind it and need to switch.
[10:13.600 --> 10:19.560]  So there's also, at this conference, a bunch of exciting new languages that are coming out.
[10:19.560 --> 10:24.080]  So I wanted to zoom in and kind of take a look to see what's going on with them.
[10:24.080 --> 10:27.920]  So, you know, Go, Kotlin and Rust in particular.
[10:27.920 --> 10:32.920]  So I will say one of the challenges, I didn't get any chance to filter this out, but
[10:32.920 --> 10:37.400]  this Go jump from November to December is some bots.
[10:37.400 --> 10:39.720]  So that's just bot activity.
[10:39.720 --> 10:44.680]  So you can, yeah, I thought originally maybe it's just Christmas and Go developers don't
[10:44.680 --> 10:45.680]  have anything better to do.
[10:45.680 --> 10:49.200]  So over the holidays, you know, they're just programming a lot and starting a lot of new
[10:49.200 --> 10:50.200]  repos.
[10:50.200 --> 10:53.120]  That wasn't, it was actually, it was some bugs.
[10:53.120 --> 10:55.240]  But you can see here, you know, steadily increasing.
[10:55.240 --> 11:01.600]  It's not a huge, huge spike for these other, these other two, but it is going up.
[11:01.600 --> 11:07.360]  But as a new language that's getting popular, I guess you'd expect to see more of that.
[11:07.360 --> 11:12.440]  So in addition to languages, also there's a bunch of the new APIs, some that were referenced
[11:12.440 --> 11:13.680]  earlier today.
[11:13.680 --> 11:20.360]  So Insertable Streams is one such API, and that's actually two sub-APIs, a media stream
[11:20.360 --> 11:23.840]  track processor and track generator.
[11:23.840 --> 11:29.760]  First took a look on Chrome status, actually credit to Fippo, you know, Phil Hankey for
[11:29.760 --> 11:35.480]  having a, he built a custom viewer of the Chrome status information that you can see
[11:35.480 --> 11:37.880]  or so compare them both at the same time.
[11:37.880 --> 11:41.200]  You can see that, you know, these are actually peaking, you know, quite a bit towards the
[11:41.200 --> 11:42.200]  end of the year going up quite a bit.
[11:42.200 --> 11:46.440]  So I was curious, like, who, you know, can we see our open source repos actually using
[11:46.440 --> 11:49.040]  these or is it somebody else?
[11:49.040 --> 11:55.720]  And looking at it, you know, there's a big spike here, but it doesn't look like much.
[11:55.720 --> 11:57.240]  So what's going on there?
[11:57.280 --> 12:01.920]  Zoom in a little bit more, and again, apologies, it's really small, but like that initial spike,
[12:01.920 --> 12:10.200]  a lot of that was just pure standards related activity of the W3C repos and WebKit and others
[12:10.200 --> 12:13.400]  that are just basically adopting, you know, adopting those APIs in the first place.
[12:13.400 --> 12:15.440]  So you see a big jump.
[12:15.440 --> 12:20.280]  After that, it's really kind of hit or miss, and I was, I mean, I love working with Insertable
[12:20.280 --> 12:22.520]  Streams, you know, let's see, do a lot of fun stuff.
[12:22.840 --> 12:26.120]  So hoping to see more, but it's kind of just spotty.
[12:26.120 --> 12:28.440]  So, you know, going back to the Chrome status, what does that mean?
[12:28.440 --> 12:34.520]  Well, at least people that are using it are probably someone like Google Meet, sort of
[12:34.520 --> 12:36.400]  thing that don't have public repos, right?
[12:36.400 --> 12:41.520]  Or there's something else outside of the public GitHub data set that's driving that usage.
[12:43.800 --> 12:46.640]  So another one is Web Codex.
[12:46.640 --> 12:47.320]  It's another one.
[12:47.320 --> 12:48.680]  This one doesn't have quite the same peak.
[12:48.720 --> 12:54.640]  It's a little bit, you know, Web Codex is not quite as far along, but they're still driving up.
[12:54.640 --> 12:57.320]  You want to see if there's something going on here too.
[12:57.320 --> 13:02.000]  And again, you see gradual increase, not a ton, except for this one spike.
[13:02.000 --> 13:08.560]  And this one spike, again, was largely related to, you know, the initial standards release
[13:08.560 --> 13:13.040]  of WebKit and W3C type repos and related once to deploy that.
[13:13.040 --> 13:17.680]  So we see some uptick, but nothing all that significant yet.
[13:19.680 --> 13:27.320]  And for the last section, I was also wondering, is WebRC winning?
[13:27.320 --> 13:31.760]  Like, we don't talk a whole lot about WebRC having competition so much anymore, at least I haven't.
[13:31.760 --> 13:34.800]  But in the early days, it was always, you know, WebRC versus SIP.
[13:34.800 --> 13:37.000]  And is it, you know, should SIP, you know, those SIP type developers,
[13:37.000 --> 13:39.600]  that ecosystems, should they shift over to WebRTC?
[13:39.600 --> 13:43.840]  And we haven't seen that a whole lot, but I think in reality, there still is competition.
[13:43.840 --> 13:48.560]  And that is certainly during the pandemic, you know, well, it's Zoom, right?
[13:48.560 --> 13:52.160]  And I actually presented this a couple of years ago at Dan's conference,
[13:52.160 --> 13:57.160]  an interesting fact where, you know, there was a month in time where Zoom was worth more
[13:57.160 --> 14:00.640]  than the seventh largest Analyze put together, at least our market capitalization,
[14:00.640 --> 14:02.440]  which is still insane when you think about it, right?
[14:02.440 --> 14:03.800]  But, you know, that was the reality.
[14:03.800 --> 14:09.360]  So I did want to check to see if that's still the case, and it's not, right?
[14:09.360 --> 14:14.720]  So, yeah, Zoom is, you know, using next to the same data point and just extending it out,
[14:14.720 --> 14:18.920]  you know, a little bit further, you know, Zoom's down near 80% where they were back
[14:18.920 --> 14:20.920]  in February of 2020.
[14:20.920 --> 14:23.280]  The airlines, though, aren't actually doing all that much better, right?
[14:23.280 --> 14:27.600]  So still relative, Zoom's not doing some bad relative to the airlines,
[14:27.600 --> 14:29.400]  at least those same seven.
[14:29.400 --> 14:37.120]  But anyway, Zoom, not quite what it was, but they still really are competition, right?
[14:37.120 --> 14:40.480]  And particularly because Zoom now has released a Zoom SDK,
[14:40.480 --> 14:42.120]  and they have a Web version of this SDK.
[14:42.120 --> 14:46.320]  So as a developer, you do have a choice.
[14:46.320 --> 14:49.520]  Hey, I want to go build a real-time communications application.
[14:49.520 --> 14:52.600]  You can choose to use the WebRC and, you know, all the vendors in the ecosystem,
[14:52.600 --> 14:55.040]  or you can go to choose Zoom for this.
[14:55.040 --> 14:57.920]  And I was curious, in Zoom's marketing, it's a lot.
[14:57.920 --> 15:02.800]  I'll probably have a post on WebRC hacks with football, hopefully in a few weeks
[15:02.800 --> 15:05.720]  that, you know, where Zoom's been promoting the benefits of this.
[15:05.720 --> 15:11.560]  And it's a, I'll go into why that's not completely true.
[15:12.000 --> 15:12.640]  During the post.
[15:12.640 --> 15:17.400]  But I wanted to see our developers actually choosing Zoom over,
[15:17.400 --> 15:22.920]  or at least mentioning the Zoom SDK over WebRC.
[15:22.920 --> 15:27.440]  It was going to take me a while to dig into all this on the GitHub analysis.
[15:27.440 --> 15:30.160]  It wasn't clear, so I didn't include that part yet.
[15:30.160 --> 15:32.000]  But on Stack Overflow, it's pretty actually clear.
[15:32.000 --> 15:37.000]  There's a distinct Zoom SDK tag or label that they have there.
[15:37.000 --> 15:38.960]  And you can see here, actually, at least for now,
[15:38.960 --> 15:42.360]  WebRC is still way more popular than the Zoom SDK.
[15:44.200 --> 15:45.840]  Two minutes, okay.
[15:45.840 --> 15:48.120]  And actually, I am done.
[15:48.120 --> 15:53.000]  So I guess what we've learned here, I mean, part of it is what are your expectations here?
[15:53.000 --> 15:55.880]  I didn't necessarily go into any expectations other than I was interested
[15:55.880 --> 16:00.680]  to see what are some of the trends and can we find or like learn things about new APIs
[16:00.680 --> 16:01.760]  or new repos.
[16:01.760 --> 16:05.240]  And I do go through the list, actually, is interesting to see new projects.
[16:05.240 --> 16:06.640]  Didn't have time to fit all that stuff in there.
[16:06.640 --> 16:09.360]  But again, you can reference some of the blog posts on this.
[16:09.360 --> 16:11.840]  But overall, my impression of WebRC is still doing pretty well.
[16:11.840 --> 16:13.520]  Obviously, it's not pandemic well.
[16:13.520 --> 16:18.400]  But given the circumstances, we're better than it was before the pandemic.
[16:18.400 --> 16:21.760]  We have more developers involved and it seems that developers that are involved,
[16:21.760 --> 16:24.080]  it is a lot of measures to say that they're more mature.
[16:24.080 --> 16:25.000]  They're better developers, right?
[16:25.000 --> 16:27.480]  They're contributing more than in the past.
[16:27.480 --> 16:28.720]  And I think that's a great thing.