WEBVTT 00:00.000 --> 00:16.000 If you're not coming, and if you can take, try to follow up on several of you and take the chance to give it a drop to all of you. 00:16.000 --> 00:20.000 It should be related to me. 00:20.000 --> 00:30.000 If someone wants to give a talk about something you want to share. 00:31.000 --> 00:37.000 There's something I would like to know about. 00:37.000 --> 00:46.000 Maybe we've over the question, make a choice and capture it with the microphone. 00:46.000 --> 00:50.000 Okay, so I want to talk a bit about what I'm going to do tomorrow. 00:50.000 --> 00:56.000 Since I'm in three years, since three years, more or less now in the matrix community or matrix ecosystem. 00:56.000 --> 01:03.000 And I'm not that deep tech guy. I'm more interested in social things. 01:03.000 --> 01:07.000 I started to organize a couple of things like the matrix community summit. 01:07.000 --> 01:09.000 I'm some matrix meetup in Berlin. 01:09.000 --> 01:15.000 And the newest thing we did just last year as a prototype was the matrix retreat. 01:15.000 --> 01:25.000 Where 10 people which work in different projects companies in the matrix ecosystem spend some time in a place in Austria in Kremsk, close to Vienna. 01:25.000 --> 01:28.000 And they did their usual remote work there. 01:28.000 --> 01:34.000 But then they shared some time and within that time work together on matrix related stuff. 01:34.000 --> 01:37.000 But also on other funny, naughty things. 01:37.000 --> 01:41.000 And I really felt in love with this kind of working atmosphere. 01:41.000 --> 01:44.000 So now I'm going to organize more of them. 01:44.000 --> 01:52.000 And one of them is starting tomorrow because we will go to Thailand and have four weeks matrix retreat in Thailand. 01:52.000 --> 01:54.000 So if you're interested in something like that. 01:54.000 --> 01:58.000 For matrix, then check matrix this week in matrix. 01:58.000 --> 02:03.000 Or check out a vocation, nation, data and often. 02:03.000 --> 02:05.000 There we organize also something. 02:05.000 --> 02:11.000 Or we try to organize something like that for other communities like KDE or norm or maybe others. 02:11.000 --> 02:13.000 That was my announcement. 02:13.000 --> 02:17.000 I could try to do Gabriel's talk. 02:17.000 --> 02:20.000 And I basically know what he would say. 02:20.000 --> 02:22.000 And so on the basis. 02:22.000 --> 02:27.240 So on the basis that I've seen that people are here because they were hoping to hear about how rocket 02:27.240 --> 02:29.000 chap works with matrix. 02:29.000 --> 02:32.000 As opposed to an AMA or whatever. 02:32.000 --> 02:40.000 But I can at least give some overview of bridging and interoperability and where things are out. 02:40.000 --> 02:49.000 So being going since 2016 originally made me to your best kind of flagship app. 02:49.000 --> 02:53.000 For the Meteor community and the JavaScript ecosystem. 02:53.000 --> 02:57.000 Where the whole stick is at your right and app in JavaScript on the front end. 02:57.000 --> 03:00.000 And you do the server code in the front end too. 03:00.000 --> 03:06.000 So you basically have the server as a magic thing that goes and connects to 03:06.000 --> 03:08.000 Gabriel clients and you can focus on the front end. 03:08.000 --> 03:11.000 So I believe that's where rocket chap came from. 03:11.000 --> 03:15.000 And so if people have used it, it's really nice product. 03:15.000 --> 03:17.000 It's a great swim web app. 03:17.000 --> 03:20.000 And I believe it's got native mobile apps as well too. 03:20.000 --> 03:23.000 Lots of communities have used it over the years. 03:23.000 --> 03:26.000 I know the blender used to use it for a while. 03:26.000 --> 03:31.000 I think that rocket chap is a company as similar challenges to the average element has. 03:31.000 --> 03:35.000 In terms of getting very popular with governments and then discovering that. 03:35.000 --> 03:40.000 Large organizations somehow forget from the people writing the software. 03:40.000 --> 03:42.000 And so that might be why he's not here today. 03:42.000 --> 03:50.000 But a few years ago, I guess I first met him probably four or five years ago. 03:50.000 --> 03:52.000 First of all, I think, first time. 03:52.000 --> 03:55.000 And he had a geek out about how rocket chap was going. 03:55.000 --> 03:58.000 And how they should absolutely start talking matrix to. 03:58.000 --> 04:01.000 And years as a common language. 04:01.000 --> 04:04.000 As a way to basically add a feature on. 04:04.000 --> 04:08.000 Because I think they had just on their own Federation, 04:08.000 --> 04:13.000 which was a very, very basic setup where you put SRV, DNS records. 04:13.000 --> 04:17.000 And that would allow the servers to basically just pass messages. 04:17.000 --> 04:19.000 We're between each other. 04:19.000 --> 04:20.000 Didn't have. 04:20.000 --> 04:24.000 It also had end-twent encryption using a shared secret. 04:24.000 --> 04:26.000 So no perfect forward secrecy. 04:26.000 --> 04:30.000 If somebody uses a password basically to secure the contents of the room. 04:30.000 --> 04:34.000 And if that password leaks, then the entire history of the room goes. 04:34.000 --> 04:36.000 And the Federation was basically proprietary. 04:36.000 --> 04:39.000 So then it's a closed API that they had. 04:39.000 --> 04:43.000 So I said, hey, rather than doing custom Federation, 04:43.000 --> 04:46.000 you should actually talk a common language. 04:46.000 --> 04:49.000 And then you'll be able to bridge free to element via this as well. 04:49.000 --> 04:51.000 And all the other matrix plans. 04:51.000 --> 04:54.000 And you can be a part of our strange matrix code. 04:54.000 --> 04:56.000 Amazingly, he didn't say no. 04:56.000 --> 05:01.000 And he said, actually, this might have some merit to it. 05:01.000 --> 05:05.000 And I can see some of our government customers are really liking this matrix stuff. 05:05.000 --> 05:07.000 Perhaps we should have a look at this. 05:07.000 --> 05:09.000 And interesting what he did at first. 05:09.000 --> 05:15.000 I think was in Google, similar to code 2016, go and do a JSON project 05:15.000 --> 05:19.000 to try to add matrix to Rocket Chat. 05:19.000 --> 05:24.000 And I recommend that the way to do this might be to start off with a bridge. 05:24.000 --> 05:27.000 But you would take the existing materials server, which is backed by MongoDB. 05:27.000 --> 05:29.000 You write a really simple bridge that we go. 05:29.000 --> 05:33.000 And basically copy and paste the messages from Rocket Chat 05:33.000 --> 05:36.000 into a synapse that we would set alongside it. 05:36.000 --> 05:40.000 And the synapse would then bust you out into the great wide matrix. 05:40.000 --> 05:42.000 Beyond it. 05:42.000 --> 05:47.000 I guess he was worried about the hassle of running two servers, 05:47.000 --> 05:49.000 which is something that comes up a lot. 05:49.000 --> 05:52.000 If you're already running a meteor server for Rocket Chat, 05:52.000 --> 05:57.000 do you really want the complication of running a whole bunch of Python alongside 05:57.000 --> 06:00.000 using Postgres when Rocket Chat has already got Mongo, 06:00.000 --> 06:02.000 it doesn't fit very nicely. 06:02.000 --> 06:06.000 And so for the first thing that they tried to do was to take 06:06.000 --> 06:10.000 the Federation Library from Dendrite, go matrix server 06:10.000 --> 06:13.000 lib, and port it, line by line, into time scripts. 06:13.000 --> 06:16.000 Which is quite a bullsey thing to do. 06:16.000 --> 06:19.000 Particularly as Dendrite was pretty alpha at that point, 06:19.000 --> 06:23.000 because again 2017, probably also by this point. 06:23.000 --> 06:27.000 And I think it's worth to say that it was an experiment in the Portuguese 06:27.000 --> 06:29.000 or the Portuguese students might have been a little out of their depth 06:29.000 --> 06:32.000 and the enthusiasm of doing this. 06:32.000 --> 06:34.000 But I think it's still out there. 06:34.000 --> 06:37.000 I'll get some way that I don't think I ever went to production. 06:37.000 --> 06:41.000 So that was the kind of first attempt to join matrix. 06:41.000 --> 06:45.000 Then a couple of years later, it cannot forget. 06:45.000 --> 06:48.000 Thanks to our friends in the Swedish government. 06:48.000 --> 06:51.000 Because Sweden is quite food-facing when it comes 06:51.000 --> 06:54.000 to trying to do the right thing for digital sovereignty. 06:54.000 --> 06:57.000 And I think they've done the numbers on what happens 06:57.000 --> 07:02.000 if too many anchors get dragged over their international submarine cables 07:02.000 --> 07:05.000 or other nasty scenarios. 07:05.000 --> 07:09.000 And so seriously, we'll go through how to communicate 07:09.000 --> 07:13.000 and how to avoid being operationally dependent on the US and SAS 07:13.000 --> 07:15.000 companies in the US for cons. 07:15.000 --> 07:20.000 And they also have a unit of the government called E-SAM 07:21.000 --> 07:24.000 who's full name, I cannot pronounce or remember. 07:24.000 --> 07:28.000 And they are responsible amongst other things for interoperability 07:28.000 --> 07:31.000 and kind of best practices for government things like procurement. 07:31.000 --> 07:36.000 And so what E-SAM saw was the different ministers in Sweden 07:36.000 --> 07:40.000 were picking completely different sovereign channels. 07:40.000 --> 07:43.000 So you had a bunch of people who were running rocket chat, 07:43.000 --> 07:45.000 a bunch of people running matter most, 07:45.000 --> 07:48.000 and a whole bunch of other people running matrix-based solutions. 07:48.000 --> 07:51.000 And so in a very collaborative Swedish manner, 07:51.000 --> 07:55.000 what they did is to invite all the CEOs for these companies 07:55.000 --> 07:57.000 on to a video conference and said, 07:57.000 --> 08:01.000 right, what we want is for you to speak the same thing. 08:01.000 --> 08:03.000 We don't want to be vendor-locks or element 08:03.000 --> 08:05.000 or we're all matters to rocket chat or whatever. 08:05.000 --> 08:07.000 We want to be able to buy whoever we want, 08:07.000 --> 08:10.000 but we also want to talk to each other, go fight. 08:10.000 --> 08:12.000 And it was a really fun call. 08:12.000 --> 08:15.000 I think it's the only time that basically all the competitors 08:15.000 --> 08:20.000 had to be in the same place and told to justify what they're doing. 08:20.000 --> 08:26.000 So the outcome of that is it was all very diplomatic and political, 08:26.000 --> 08:27.000 as you mentioned. 08:27.000 --> 08:29.000 And all the CEOs basically said, 08:29.000 --> 08:30.000 yes, yes, absolutely. 08:30.000 --> 08:32.000 We obviously want to do the right thing and say yes to the government 08:32.000 --> 08:33.000 to make this happen. 08:33.000 --> 08:35.000 And yes, I can see the benefits of doing this. 08:35.000 --> 08:38.000 Obviously, I was quite biased in favor of it 08:38.000 --> 08:40.000 because there was only one solution on the table, 08:40.000 --> 08:42.000 which was matrix as an mechanism to go and link 08:42.000 --> 08:44.000 the different departments together. 08:44.000 --> 08:47.000 But afterwards, I think we've got Gabriel offline 08:47.000 --> 08:51.000 and said, hey, do you want to do this for real? 08:51.000 --> 08:53.000 So yeah, why not? 08:53.000 --> 08:55.000 I mean, we'll be stronger together. 08:55.000 --> 08:56.000 And it was like, yeah, he gets it. 08:56.000 --> 08:59.000 There is absolutely opportunity to, 08:59.000 --> 09:02.000 know for him, for some more rocket chat instances 09:02.000 --> 09:03.000 and likewise element. 09:03.000 --> 09:07.000 So we started talking about it more seriously again 09:07.000 --> 09:11.000 and started to do this time, right to bridge. 09:11.000 --> 09:15.000 And this must be about 2021 or so. 09:15.000 --> 09:20.000 And it was just after element had taken getter 09:20.000 --> 09:24.000 off the hands of get lab, just as get lab was going to IPO. 09:24.000 --> 09:27.000 Get lab was having to focus all of its books 09:27.000 --> 09:30.000 and get rid of any weird stuff and poor old getter 09:30.000 --> 09:33.000 who they had bought so many years earlier 09:33.000 --> 09:35.000 was a bit of an anonymity, whereas we of course 09:35.000 --> 09:37.000 element did an awful lot of chat. 09:37.000 --> 09:42.000 And so basically get lab suggested that we could go 09:42.000 --> 09:45.000 and move getter into matrix and I thought, wow, 09:45.000 --> 09:48.000 that would be great because getter can be a showcase 09:48.000 --> 09:51.000 of how you're taking existing app and how do you integrate 09:51.000 --> 09:53.000 it to make it natively speak matrix. 09:53.000 --> 09:55.000 And we had just done that. 09:55.000 --> 09:57.000 And as it happens, getter has a very similar start 09:57.000 --> 09:58.000 to rocket chat. 09:58.000 --> 10:00.000 It's a whole bunch of touch groups 10:00.000 --> 10:03.000 and both from the front and then the back end from memory. 10:03.000 --> 10:06.000 So being able to say to Gabriel, yeah. 10:06.000 --> 10:09.000 Sure, you can get rocket chat talking matrix. 10:09.000 --> 10:12.000 And hey, look at the tutorial we wrote when we went 10:12.000 --> 10:15.000 and wrote a bridge to make getter natively speak matrix. 10:15.000 --> 10:17.000 You should do the same thing. 10:17.000 --> 10:20.000 And he puts in full time employees on it and went off. 10:20.000 --> 10:24.000 And we also as elements and provided consulting 10:24.000 --> 10:28.000 to basically answer any of the weird matrix bits. 10:28.000 --> 10:30.000 And it was really fun. 10:30.000 --> 10:31.000 It worked. 10:31.000 --> 10:32.000 They got written. 10:32.000 --> 10:33.000 And it went live. 10:33.000 --> 10:38.000 I believe in rocket chat 6, which is about a year or two. 10:38.000 --> 10:41.000 Well, possibly that sort of time frame. 10:41.000 --> 10:44.000 Now, the story gets big murky here because I'm not 10:44.000 --> 10:46.000 in fact Gabriel Angles, he a rocket chat. 10:46.000 --> 10:48.000 I'm Matthew Hodgson, he a relevant. 10:48.000 --> 10:53.000 But last I checked, as I said, they were marketing on it 10:53.000 --> 10:55.000 as a differentiator for rocket chat. 10:55.000 --> 10:58.000 The hey one of the good things is that we've had a regular matrix 10:58.000 --> 11:00.000 too and really leaning into it, which is cool. 11:00.000 --> 11:04.000 It's worth noting that Matthew most also said, yes, we will talk 11:04.000 --> 11:05.000 matrix. 11:05.000 --> 11:07.000 Did get somebody to write a matrix bridge for it. 11:07.000 --> 11:10.000 And then kind of didn't yell about it too much. 11:10.000 --> 11:14.000 But if you do look very deep in the maths community projects, 11:14.000 --> 11:16.000 I think you will find there is a mathematics 11:16.000 --> 11:19.000 connector or matrix connector for mathematics. 11:19.000 --> 11:21.000 But rocket chat properly leaned into it. 11:21.000 --> 11:26.000 And I believe that they then decided that they weren't wanting 11:26.000 --> 11:31.000 to bundle synapses next to rocket chat because synapses 11:31.000 --> 11:35.000 is in Python and they wanted something shiny. 11:35.000 --> 11:38.000 And so they decided they would use dendroy. 11:38.000 --> 11:41.000 Which in some ways would make sense if it wasn't for the fact 11:41.000 --> 11:44.000 which Eric Glost over completely in his synapses 11:44.000 --> 11:46.000 later the nation just now. 11:46.000 --> 11:51.000 The poor dendroy has basically got stuck in a eternal beta 11:51.000 --> 11:56.000 system where a first the intention was to move synapses 11:56.000 --> 11:58.000 entirely over to dendroy. 11:58.000 --> 12:01.000 But then I think France turned up and rolled out synapses 12:01.000 --> 12:06.000 at scale everywhere and suddenly become accredited 12:06.000 --> 12:09.000 and audited and was actually being run for real. 12:09.000 --> 12:12.000 And we had to really focus on making sure that it worked. 12:12.000 --> 12:15.000 At which point all of the time it should be spent making dendroy 12:15.000 --> 12:18.000 amazing with being spent maintaining synapses. 12:18.000 --> 12:21.000 And then rapidly we started stealing the architecture 12:21.000 --> 12:24.000 of dendroy like workers and replication 12:24.000 --> 12:27.000 and backporting it into synapses. 12:27.000 --> 12:30.000 When we first wrote dendroy synapses had no workers. 12:30.000 --> 12:33.000 It was a Python long lift and if you run out of CPU on a call 12:33.000 --> 12:34.000 you were done. 12:34.000 --> 12:37.000 But dendroy, on the other hand, is a microservices architecture 12:37.000 --> 12:40.000 that uses Kafka at the time in nowadays 12:40.000 --> 12:43.000 and that's as an event bus to go and link together 12:43.000 --> 12:45.000 all these different instances. 12:45.000 --> 12:47.000 The work is using radius as an event bus 12:47.000 --> 12:50.000 and synapses is basically a port of dendroy 12:50.000 --> 12:51.000 architecture to synapses. 12:51.000 --> 12:53.000 Long story short, we ended up with 12:53.000 --> 12:56.000 and still have two server implementations written by the 12:56.000 --> 12:59.000 core team who now does hang out our element. 12:59.000 --> 13:03.000 And port and tried is great. 13:03.000 --> 13:06.000 And there's now being maintained or at least 13:06.000 --> 13:08.000 mainly contributed to by the community. 13:08.000 --> 13:10.000 I'm hoping the element will get profit all 13:10.000 --> 13:12.000 some day before we run out of money ideally. 13:12.000 --> 13:15.000 And we will be able to invest some of the profits 13:15.000 --> 13:18.000 into maintaining dendroy properly again 13:18.000 --> 13:22.000 so that there is a second totally independent 13:22.000 --> 13:26.000 implementation done by the folks who created matrix 13:26.000 --> 13:28.000 in the first place to keep ourselves on 13:28.000 --> 13:30.000 as much as anything else. 13:30.000 --> 13:32.000 But so how dendroy is used today 13:32.000 --> 13:36.000 is basically for experimentation like peer-to-peer, 13:36.000 --> 13:39.000 like crypto IDs and pseudo IDs. 13:39.000 --> 13:43.000 There's no privacy work happening 13:43.000 --> 13:45.000 particularly in synapses right now 13:45.000 --> 13:47.000 because it was all happening in dendroy 13:47.000 --> 13:52.000 at least once dendroy had cash applied to it. 13:52.000 --> 13:55.000 So basically I told some Gabriel, 13:55.000 --> 13:57.000 I'm not sure you should be using dendroy 13:57.000 --> 13:59.000 because it's kind of in beta. 13:59.000 --> 14:01.000 And he said, oh, my devs love it. 14:01.000 --> 14:03.000 It's so much fun working on this experimental code. 14:03.000 --> 14:06.000 Yeah, we'll go with this and ship to anyway. 14:06.000 --> 14:09.000 At least I think they ship it anyway. 14:09.000 --> 14:11.000 So I'm not sure how that's going 14:11.000 --> 14:14.000 because obviously dendroy has not been massively 14:14.000 --> 14:16.000 maintained now for a year and a bit. 14:16.000 --> 14:18.000 There are PLs coming in. 14:18.000 --> 14:20.000 We'd still do critical updates, 14:20.000 --> 14:23.000 but you probably should not be using it in production. 14:23.000 --> 14:25.000 And if they are using it as the 14:25.000 --> 14:28.000 Dockerized home server that you get 14:28.000 --> 14:30.000 when you do a Docker compose rocket chat 14:30.000 --> 14:33.000 and still then that would not be ideal. 14:34.000 --> 14:37.000 However, I'm hoping that they also have 14:37.000 --> 14:39.000 it set up so you can use an app so you can use your 14:39.000 --> 14:41.000 existing home server alongside it. 14:41.000 --> 14:43.000 Or perhaps they've gone Polish dendroy enough 14:43.000 --> 14:45.000 that it is production great. 14:45.000 --> 14:48.000 Metal known fact is a Reddit chat is built entirely 14:48.000 --> 14:49.000 on dendroyts. 14:49.000 --> 14:51.000 Who knew that? 14:51.000 --> 14:53.000 Yeah, about four people. 14:53.000 --> 14:55.000 It's pretty depressing in that 14:55.000 --> 14:58.000 about the point that we ran after money 14:58.000 --> 15:01.000 to work on dendroyts was also the 15:01.000 --> 15:04.000 point that read it which is a relatively major 15:04.000 --> 15:08.000 organisation of hundreds of millions of users 15:08.000 --> 15:10.000 turned it on to power their system. 15:10.000 --> 15:13.000 Apparently they did a proprietary fork of it 15:13.000 --> 15:15.000 thanks to the matcha curve Apache licensing 15:15.000 --> 15:18.000 and replaced the backend to use 15:18.000 --> 15:22.000 Cassandra, Rav, then Postgres 15:22.000 --> 15:24.000 because all of their infrastructure is Cassandra. 15:24.000 --> 15:26.000 And I believe to this day if you go to chat. 15:26.000 --> 15:28.000 Reddit.com right now and look at the network 15:28.000 --> 15:30.000 perspective you'll see a whole bunch of 15:30.000 --> 15:32.000 matrix traffic flying around as it 15:32.000 --> 15:34.000 talks dendroyts. 15:34.000 --> 15:36.000 In fact as that counts as production. 15:36.000 --> 15:38.000 The power dendroyts is production great. 15:38.000 --> 15:41.000 Perhaps for rocket chat using it isn't 15:41.000 --> 15:42.000 such a crazy idea. 15:42.000 --> 15:44.000 But for my side it's a little bit weird 15:44.000 --> 15:46.000 and I hope that they have the option 15:46.000 --> 15:48.000 of synapses as well. 15:48.000 --> 15:50.000 I think the Gabriel for I was trying to 15:50.000 --> 15:52.000 have no good at 15:52.000 --> 15:54.000 disweight him from using dendroyts because I 15:54.000 --> 15:56.000 wanted to keep dendroyts as a special 15:56.000 --> 15:59.000 differentiated competitive advantage for 15:59.000 --> 16:01.000 element which is not the case. 16:01.000 --> 16:03.000 I told him that this is probably stuck in 16:03.000 --> 16:05.000 beta I really meant it. 16:05.000 --> 16:08.000 But that is where things are right there. 16:08.000 --> 16:11.000 And actually I did speak to him 16:11.000 --> 16:13.000 at Postom 2 years ago. 16:13.000 --> 16:15.000 And had a meal to go for and talk about 16:15.000 --> 16:17.000 how things going and he I think said 16:17.000 --> 16:19.000 that they were starting to look fairly 16:19.000 --> 16:22.000 seriously as the solution to the dendroyts thing 16:22.000 --> 16:25.000 to finish off the TS matrix 16:25.000 --> 16:27.000 server lab idea and get their 16:27.000 --> 16:29.000 node code based talking proper 16:29.000 --> 16:31.000 federation. 16:31.000 --> 16:33.000 And based on the good experience 16:33.000 --> 16:35.000 that they had with the bridging and the 16:35.000 --> 16:37.000 fact that the architecture had worked 16:37.000 --> 16:38.000 relatively well. 16:38.000 --> 16:40.000 So I don't know if somebody has 16:40.000 --> 16:41.000 got a laptop in front of them 16:41.000 --> 16:43.000 perhaps open up get up and get 16:43.000 --> 16:45.000 to rocket chat and do some search terms to 16:45.000 --> 16:47.000 see if they actually have server 16:47.000 --> 16:50.000 federation API work happening there. 16:50.000 --> 16:52.000 But I believe their intention was to 16:52.000 --> 16:54.000 bring it native so you don't have to 16:54.000 --> 16:56.000 just have one stack and it will be written 16:56.000 --> 16:58.000 in the same language they would maintain 16:58.000 --> 17:00.000 themselves and it would work great. 17:00.000 --> 17:02.000 So perhaps his talk was going to be about 17:02.000 --> 17:04.000 that and what it was working. 17:04.000 --> 17:07.000 And in my dream of it would be kind of 17:07.000 --> 17:09.000 fun if there was a whole new home server 17:09.000 --> 17:10.000 implementation written in 17:10.000 --> 17:13.000 node done by the rocket chat team. 17:13.000 --> 17:15.000 But unless someone has actually looked 17:15.000 --> 17:17.000 and found it I honestly don't know what 17:17.000 --> 17:18.000 the answer is. 17:18.000 --> 17:23.000 So. 17:23.000 --> 17:26.000 But may or may not have been in 17:26.000 --> 17:28.000 the talk he knows. 17:28.000 --> 17:31.000 Well, big applause anyway. 17:31.000 --> 17:36.000 Thank you very much. 17:36.000 --> 17:38.000 I think there are a lot of interesting 17:38.000 --> 17:39.000 insights that actually were 17:39.000 --> 17:41.000 probably would have never got back to the 17:41.000 --> 17:42.000 group. 17:42.000 --> 17:44.000 Yeah, it gets all questions would be a bit 17:44.000 --> 17:45.000 hard. 17:45.000 --> 17:47.000 Now I mean honestly people have questions 17:47.000 --> 17:49.000 and ways of gluing took different 17:49.000 --> 17:50.000 technologies together. 17:50.000 --> 17:51.000 You could also gossip about the 17:51.000 --> 17:52.000 disaster that is DMA. 17:52.000 --> 17:53.000 To the have a question. 17:53.000 --> 17:56.000 Okay, what's the cause? 17:56.000 --> 17:58.000 I'll relate to the audience in the second. 17:58.000 --> 18:01.000 Is called home server? 18:01.000 --> 18:04.000 Why are those the best name for a home server? 18:04.000 --> 18:05.000 I don't know. 18:05.000 --> 18:06.000 I don't know. 18:06.000 --> 18:09.000 Okay, what's the cause? 18:09.000 --> 18:11.000 I'll relate to the audience in the second. 18:11.000 --> 18:14.000 Is called home server? 18:14.000 --> 18:16.000 Why are those the best name for a home server? 18:16.000 --> 18:18.000 I've ever heard. 18:18.000 --> 18:20.000 Okay, so to relate to there's been 18:20.000 --> 18:22.000 researching and it turns out that 18:22.000 --> 18:24.000 indeed they have been working on 18:24.000 --> 18:26.000 an active note home server 18:26.000 --> 18:28.000 implementation for integration 18:28.000 --> 18:30.000 into rocket share and it is 18:30.000 --> 18:32.000 GitHub.com slash rocket dot 18:32.000 --> 18:34.000 chat slash home server. 18:34.000 --> 18:36.000 And it was being worked on until 18:36.000 --> 18:37.000 a couple of months ago. 18:37.000 --> 18:38.000 It doesn't look complete. 18:38.000 --> 18:40.000 But I imagine when rocket share 18:40.000 --> 18:42.000 submitted this talk they wanted to talk 18:42.000 --> 18:44.000 about that project and the guy who was 18:44.000 --> 18:46.000 working on it and then the success 18:46.000 --> 18:48.000 or not that they had had with it. 18:48.000 --> 18:50.000 The fact that it hasn't had commits in a few months 18:50.000 --> 18:51.000 and Gabriel isn't here. 18:51.000 --> 18:52.000 It implies that it may not have gone 18:52.000 --> 18:53.000 amazing you are. 18:53.000 --> 18:54.000 But you know, it's going to look at the 18:54.000 --> 18:56.000 codes and helps it's great. 18:56.000 --> 18:58.000 But yeah, people do have questions 18:58.000 --> 19:00.000 about bridging or interrupt 19:00.000 --> 19:02.000 existing systems into matrix. 19:02.000 --> 19:04.000 I mean, bridging is a 19:04.000 --> 19:05.000 double-edged sword. 19:05.000 --> 19:07.000 It's a great way to get rocket chat 19:07.000 --> 19:09.000 or get a raw mattermaster or 19:09.000 --> 19:11.000 WhatsApp into matrix. 19:11.000 --> 19:12.000 But on the other hand, 19:12.000 --> 19:15.000 it's always an impedance mismatch. 19:15.000 --> 19:17.000 It's not benefiting from typically 19:17.000 --> 19:19.000 not benefiting from matrixes and to 19:19.000 --> 19:20.000 encryption. 19:20.000 --> 19:21.000 It's a bit of a dirty hat. 19:21.000 --> 19:24.000 So, President Kant, if people 19:24.000 --> 19:26.000 have questions or thoughts, this is your 19:26.000 --> 19:27.000 time to have an improvised bridging 19:27.000 --> 19:28.000 Q&A. 19:28.000 --> 19:30.000 If not, I will stop one morning. 19:30.000 --> 19:31.000 There was a hand. 19:31.000 --> 19:32.000 No, that was Kevin's scratching his ear. 19:32.000 --> 19:33.000 Sorry, Kevin. 19:33.000 --> 19:34.000 Yeah. 19:34.000 --> 19:38.000 So, how scare boys 19:39.000 --> 19:41.000 are going to be? 19:41.000 --> 19:44.000 So, scalability for synapse. 19:44.000 --> 19:47.000 So, 19:47.000 --> 19:50.000 yeah, yep. 19:50.000 --> 19:52.000 So, the Eric had a sort of 19:52.000 --> 19:53.000 state-of-the-nation moment. 19:53.000 --> 19:54.000 He didn't really target 19:54.000 --> 19:55.000 specifically scalability. 19:55.000 --> 19:57.000 First of all, how do you define 19:57.000 --> 19:58.000 scalability? 19:58.000 --> 20:00.000 Is this like messages per second? 20:00.000 --> 20:02.000 Is this number of uses in the room? 20:02.000 --> 20:04.000 Is this number of service 20:04.000 --> 20:05.000 participating in a room? 20:05.000 --> 20:08.000 If we assume that it is the number 20:08.000 --> 20:11.000 of users, let me think. 20:11.000 --> 20:14.000 That number of locals uses on a server. 20:14.000 --> 20:16.000 So, say that you're running just one server. 20:16.000 --> 20:18.000 It's not federated and you've got 20:18.000 --> 20:20.000 a million users on it. 20:20.000 --> 20:22.000 You're going to need to run lots of work 20:22.000 --> 20:24.000 processes, which is not 20:24.000 --> 20:25.000 Kafka. 20:25.000 --> 20:26.000 It's synapse. 20:26.000 --> 20:28.000 You use radius to basically 20:28.000 --> 20:30.000 find out data from the main process. 20:30.000 --> 20:32.000 Nowadays, the main process 20:32.000 --> 20:34.000 doesn't do any messaging. 20:34.000 --> 20:36.000 It is just for coordination. 20:36.000 --> 20:37.000 So, it is not on the hot path 20:37.000 --> 20:39.000 of any actual messaging. 20:39.000 --> 20:41.000 So, what you do is to run lots of 20:41.000 --> 20:42.000 what we call work processes, 20:42.000 --> 20:44.000 which go and handle 20:44.000 --> 20:46.000 I know sending messages to clients, 20:46.000 --> 20:47.000 receiving messages from clients, 20:47.000 --> 20:49.000 also sending receiving to servers 20:49.000 --> 20:51.000 and a bunch of other ones. 20:51.000 --> 20:52.000 And you can run as many of 20:52.000 --> 20:53.000 those as you like. 20:53.000 --> 20:54.000 And as we say, 20:54.000 --> 20:56.000 matrix.org is about 20:56.000 --> 20:57.000 170 of these. 20:57.000 --> 20:59.000 So, basically 170 calls 20:59.000 --> 21:01.000 worth maximum of CPU. 21:02.000 --> 21:03.000 And that does a 21:03.000 --> 21:05.000 bout 21:05.000 --> 21:06.000 all around 21:06.000 --> 21:07.000 two million simultaneous 21:07.000 --> 21:08.000 connected users, 21:08.000 --> 21:09.000 I think, 21:09.000 --> 21:11.000 on that server across all of those 21:11.000 --> 21:12.000 workers. 21:12.000 --> 21:14.000 And two two physical 21:14.000 --> 21:15.000 servers, 21:15.000 --> 21:17.000 each with a 21:17.000 --> 21:19.000 whole bunch of 64 21:19.000 --> 21:21.000 called CPUs and lots 21:21.000 --> 21:22.000 of NVMe 21:22.000 --> 21:23.000 and disks about 10 21:23.000 --> 21:24.000 terabytes of 21:24.000 --> 21:25.000 disk, 21:25.000 --> 21:26.000 specifically. 21:26.000 --> 21:27.000 In fact, 21:27.000 --> 21:28.000 there are two application 21:28.000 --> 21:30.000 servers, two database 21:30.000 --> 21:31.000 servers. 21:31.000 --> 21:32.000 And then a bunch of 21:32.000 --> 21:33.000 BPSs 21:33.000 --> 21:34.000 who do some of the more 21:34.000 --> 21:35.000 boring 21:35.000 --> 21:36.000 workers 21:36.000 --> 21:37.000 and setting alongside it. 21:37.000 --> 21:38.000 So, that is how far it 21:38.000 --> 21:39.000 can scale. 21:39.000 --> 21:40.000 But you 21:40.000 --> 21:41.000 right now, 21:41.000 --> 21:42.000 or at least until 21:42.000 --> 21:43.000 very recently, 21:43.000 --> 21:44.000 and all 21:44.000 --> 21:45.000 thought of those workers, 21:45.000 --> 21:46.000 did nothing but 21:46.000 --> 21:47.000 keep up with traffic. 21:47.000 --> 21:48.000 So, they weren't actually doing 21:48.000 --> 21:49.000 much useful work. 21:49.000 --> 21:50.000 They were basically 21:50.000 --> 21:51.000 consuming, 21:51.000 --> 21:52.000 replicated 21:52.000 --> 21:53.000 fire hose, 21:53.000 --> 21:54.000 filtering it, 21:54.000 --> 21:55.000 pulling stuff out of the database, 21:55.000 --> 21:57.000 trying to decide what to do with it. 21:57.000 --> 21:59.000 So, 21:59.000 --> 22:01.000 I'm going to be a 22:01.000 --> 22:04.000 project to go and improve 22:04.000 --> 22:05.000 the scalability, 22:05.000 --> 22:07.000 but I can't talk about it because 22:07.000 --> 22:08.000 this is awesome. 22:08.000 --> 22:10.000 Any other questions? 22:10.000 --> 22:12.000 Did I answer the question? 22:12.000 --> 22:13.000 Excellent. 22:13.000 --> 22:15.000 I can obviously am a 22:15.000 --> 22:17.000 on nature extra from 22:17.000 --> 22:18.000 an employee in the face, 22:18.000 --> 22:20.000 so, if it is flat the room or 22:20.000 --> 22:21.000 friendly themselves, 22:21.000 --> 22:22.000 I have to win the 22:22.000 --> 22:24.000 one. 22:24.000 --> 22:27.000 So, this one, 22:28.000 --> 22:29.000 for sure. 22:29.000 --> 22:30.000 So, the question is, 22:30.000 --> 22:32.000 make tricks on IoT. 22:32.000 --> 22:34.000 So, 22:34.000 --> 22:35.000 I mean, 22:35.000 --> 22:36.000 Matrix is very flexible. 22:36.000 --> 22:37.000 It's just an invent bus 22:37.000 --> 22:39.000 with persistence and encryption 22:39.000 --> 22:40.000 for anything. 22:40.000 --> 22:41.000 And 22:41.000 --> 22:42.000 we've done a lot of 22:42.000 --> 22:43.000 weird experiments over the 22:43.000 --> 22:45.000 years to try to persuade people to 22:45.000 --> 22:46.000 build 22:46.000 --> 22:47.000 weird things on Matrix and use it 22:47.000 --> 22:48.000 from open chat. 22:48.000 --> 22:49.000 In practice, 22:49.000 --> 22:50.000 in retrospect, 22:50.000 --> 22:51.000 this might have been more about 22:51.000 --> 22:52.000 me wanting to do fun 22:52.000 --> 22:54.000 than making the world a better place. 22:54.000 --> 22:55.000 But, as a result, 22:56.000 --> 22:57.000 we've done some IoT 22:57.000 --> 22:58.000 ones. 22:58.000 --> 22:59.000 A very early one was actually 22:59.000 --> 23:00.000 done by the community, 23:00.000 --> 23:01.000 by the guy called Ryan 23:01.000 --> 23:02.000 Rex. 23:02.000 --> 23:03.000 He was a Uber at the 23:03.000 --> 23:04.000 time. 23:04.000 --> 23:05.000 He connected his entire 23:05.000 --> 23:07.000 home assistant set up of 23:07.000 --> 23:08.000 LED strip lighting. 23:08.000 --> 23:09.000 This was like 10 years ago, 23:09.000 --> 23:10.000 so the LED strip lighting 23:10.000 --> 23:12.000 was pretty cool. 23:12.000 --> 23:13.000 With 23:13.000 --> 23:15.000 non-due in the Raspberry Pi 23:15.000 --> 23:16.000 and Matrix, 23:16.000 --> 23:18.000 so that he could basically use 23:18.000 --> 23:19.000 Matrix as an invent bus 23:19.000 --> 23:21.000 to get all the telemetry of both 23:21.000 --> 23:23.000 conjunction lights and other stuff in the 23:23.000 --> 23:24.000 room. 23:24.000 --> 23:25.000 Another example is 23:25.000 --> 23:26.000 one 23:26.000 --> 23:28.000 a demo that I did at WebRTC 23:28.000 --> 23:29.000 World 2016 23:29.000 --> 23:30.000 in Miami, 23:30.000 --> 23:31.000 Florida. 23:31.000 --> 23:32.000 I turned up with a 23:32.000 --> 23:33.000 parent drone 23:33.000 --> 23:34.000 back before drones 23:34.000 --> 23:36.000 with many used to kill people. 23:36.000 --> 23:38.000 It was appropriate to turn 23:38.000 --> 23:39.000 up at a conference with a 23:39.000 --> 23:40.000 little drone, 23:40.000 --> 23:41.000 which I'd written a bridge 23:41.000 --> 23:43.000 between the C++ 23:43.000 --> 23:44.000 led parent library that 23:44.000 --> 23:45.000 it came with 23:45.000 --> 23:46.000 and Matrix, 23:46.000 --> 23:47.000 so that I could video 23:47.000 --> 23:48.000 call it. 23:48.000 --> 23:49.000 And get hardware 23:49.000 --> 23:51.000 264 encoded 1080p video 23:51.000 --> 23:52.000 into a normal matrix 23:53.000 --> 23:54.000 cool, so just use it as a 23:54.000 --> 23:55.000 remote webcam, 23:55.000 --> 23:56.000 but I also did 23:56.000 --> 23:57.000 flight by 23:57.000 --> 23:58.000 while over 23:58.000 --> 23:59.000 IoT. 23:59.000 --> 24:00.000 It's a very sophisticated 24:00.000 --> 24:01.000 boat school, 24:01.000 --> 24:02.000 where you would type 24:02.000 --> 24:03.000 launch, return, 24:03.000 --> 24:04.000 and this would 24:04.000 --> 24:05.000 make the thing go up 24:05.000 --> 24:06.000 and flip, 24:06.000 --> 24:07.000 and it would do 24:07.000 --> 24:08.000 360, 24:08.000 --> 24:09.000 and all this sort of thing. 24:09.000 --> 24:10.000 But joking inside, 24:10.000 --> 24:11.000 it was really a basic way 24:11.000 --> 24:12.000 of doing 24:12.000 --> 24:13.000 non-human 24:13.000 --> 24:14.000 messaging there. 24:14.000 --> 24:15.000 Then 24:15.000 --> 24:16.000 the vendors there 24:16.000 --> 24:18.000 initiated a project 24:18.000 --> 24:19.000 to do 24:20.000 --> 24:21.000 an IoT telemetry 24:21.000 --> 24:22.000 through 24:22.000 --> 24:23.000 vehicles, 24:23.000 --> 24:24.000 so specifically 24:24.000 --> 24:26.000 ODB2 24:26.000 --> 24:27.000 telemetry, 24:27.000 --> 24:28.000 so this is the 24:28.000 --> 24:29.000 diagnostic port that all 24:29.000 --> 24:30.000 vehicles have. 24:30.000 --> 24:31.000 You can set the data 24:31.000 --> 24:32.000 out of it, 24:32.000 --> 24:33.000 which is 24:33.000 --> 24:35.000 hilariously fragmented and 24:35.000 --> 24:36.000 non-standardized, 24:36.000 --> 24:37.000 is every vendor just 24:37.000 --> 24:39.000 puts whatever the 24:39.000 --> 24:40.000 hell they like in it. 24:40.000 --> 24:41.000 And then 24:41.000 --> 24:42.000 sucking it all up, 24:42.000 --> 24:43.000 putting it into 24:43.000 --> 24:44.000 matrix, aggregating it, 24:44.000 --> 24:45.000 building 24:45.000 --> 24:46.000 dashboards of 24:46.000 --> 24:47.000 engine pressure, 24:47.000 --> 24:48.000 and 24:48.000 --> 24:49.000 there would be a bunch of other ones 24:49.000 --> 24:50.000 like it. 24:50.000 --> 24:51.000 But it's always been a bit 24:51.000 --> 24:52.000 fringe. 24:52.000 --> 24:53.000 I mean, it works, 24:53.000 --> 24:54.000 but there's never been a 24:54.000 --> 24:55.000 kind of 24:55.000 --> 24:56.000 element like company out there saying, 24:56.000 --> 24:57.000 you know, 24:57.000 --> 24:58.000 we're going to do 24:58.000 --> 24:59.000 kick-ass IoT 24:59.000 --> 25:00.000 with matrix. 25:00.000 --> 25:01.000 And also, 25:01.000 --> 25:02.000 the IoT folks like 25:02.000 --> 25:03.000 MQT, 25:03.000 --> 25:04.000 they like 25:04.000 --> 25:06.000 the co-app 25:06.000 --> 25:07.000 style protocols, 25:07.000 --> 25:08.000 which are very, 25:08.000 --> 25:09.000 very simple, 25:09.000 --> 25:10.000 small, 25:10.000 --> 25:11.000 and can put on your 25:11.000 --> 25:12.000 ESP, 25:12.000 --> 25:13.000 86 or whatever 25:13.000 --> 25:15.000 your favorite micro controller 25:15.000 --> 25:16.000 is a matrix client, 25:16.000 --> 25:18.000 is a bit chunky. 25:18.000 --> 25:19.000 First of all, 25:19.000 --> 25:20.000 let's do TLS, 25:20.000 --> 25:21.000 then we want to do 25:21.000 --> 25:22.000 end-to-end encryption, 25:22.000 --> 25:23.000 and here is a 25:23.000 --> 25:25.000 massive E2E stack. 25:25.000 --> 25:26.000 And so, 25:26.000 --> 25:27.000 I think there is 25:27.000 --> 25:29.000 basically a gap between 25:29.000 --> 25:30.000 the lightness that you 25:30.000 --> 25:31.000 expect on an IoT device, 25:31.000 --> 25:32.000 and the chunkiness, 25:32.000 --> 25:33.000 the matrix, 25:33.000 --> 25:34.000 can give you. 25:34.000 --> 25:35.000 And I hope 25:35.000 --> 25:36.000 ran out of time. 25:36.000 --> 25:37.000 I'm not 25:37.000 --> 25:38.000 existing to thank you so much. 25:38.000 --> 25:39.000 Thank you.