WEBVTT 00:00.000 --> 00:07.000 People can keep coming in. 00:07.000 --> 00:12.000 All right, hi everyone, my name's Dan Phillips. 00:12.000 --> 00:19.000 I am here to give a talk that's probably quite a bit different from most of the talks in this room today. 00:19.000 --> 00:21.000 And we'll just kind of get into it. 00:21.000 --> 00:23.000 And we can see what we think. 00:23.000 --> 00:25.000 So as people come in, I'll just kind of get started. 00:25.000 --> 00:28.000 But in true fashion, I'm just going to start with a quick demo. 00:28.000 --> 00:34.000 So I work on a project called Boxer. 00:34.000 --> 00:39.000 Boxer is similar in terms of the devics to Docker. 00:39.000 --> 00:42.000 So I'll just give you a quick demo, and then we can talk about it. 00:42.000 --> 00:44.000 All right, so excuse me. 00:44.000 --> 00:46.000 Oh, it's too small. 00:46.000 --> 00:48.000 How's that? 00:48.000 --> 00:49.000 Better? 00:49.000 --> 00:50.000 Okay, great. 00:50.000 --> 00:54.000 All right, so Boxer has the Box Steel I command, 00:54.000 --> 00:56.000 and I'm going to ask you our course. 00:56.000 --> 01:02.000 Boxer allows you to take a Docker file, 01:02.000 --> 01:05.000 just like you would in Docker, and compile it, 01:05.000 --> 01:07.000 and then have a runnable thing. 01:07.000 --> 01:12.000 Okay, so what I'm going to do here is Box Builds. 01:12.000 --> 01:15.000 Docker file. 01:15.000 --> 01:18.000 Okay, one thing I should show you about this. 01:18.000 --> 01:22.000 This is just a small little C binary that we have here. 01:22.000 --> 01:26.000 Actually, what that looks like. 01:26.000 --> 01:28.000 Okay, very, very simple. 01:28.000 --> 01:30.000 Pretty not a bunch of numbers, right? 01:30.000 --> 01:32.000 Excellent for demos. 01:32.000 --> 01:34.000 The most important thing is this line, 01:34.000 --> 01:37.000 which I'll come back to in just a bit. 01:37.000 --> 01:39.000 So I built the image. 01:39.000 --> 01:42.000 Now, I can see what we have. 01:42.000 --> 01:44.000 You'll see that. 01:44.000 --> 01:48.000 This looks a little big for that, but I'll get back into that later. 01:48.000 --> 01:50.000 Cool, so now I can run. 01:50.000 --> 01:51.000 Excellent. 01:51.000 --> 01:52.000 Let's do it. 01:52.000 --> 01:55.000 Now, here's the kicker. 01:55.000 --> 01:58.000 Same exact thing. 01:58.000 --> 02:00.000 Same exact thing runs here. 02:00.000 --> 02:02.000 And if you read the description of the stock, 02:02.000 --> 02:04.000 you'll understand that the reason this is possible 02:04.000 --> 02:08.000 is because everything that was running was in WebAssembly. 02:08.000 --> 02:09.000 Right? 02:09.000 --> 02:14.000 So, back to the talk. 02:14.000 --> 02:17.000 So, just quickly about me, I work at a company called Lupole Labs. 02:17.000 --> 02:19.000 We have a product that is unrelated to this. 02:19.000 --> 02:22.000 I lead all the WebAssembly stuff there, which is one little small part of what we do. 02:22.000 --> 02:25.000 With our products, we have our main thing is architect. 02:25.000 --> 02:28.000 Run, which allows you to migrate live migrate workloads 02:28.000 --> 02:29.000 without downtime. 02:29.000 --> 02:30.000 It's pretty cool to check it out. 02:30.000 --> 02:32.000 This is me on my handles. 02:32.000 --> 02:33.000 I live in Chicago. 02:33.000 --> 02:36.000 And I run the WebAssembly Chicago research group. 02:36.000 --> 02:39.000 And if you're ever in town come through. 02:39.000 --> 02:41.000 We have meetups every couple months. 02:41.000 --> 02:42.000 Cool. 02:42.000 --> 02:45.000 So, most of us have heard of WebAssembly. 02:45.000 --> 02:46.000 Yeah? 02:46.000 --> 02:47.000 I think, yes. 02:47.000 --> 02:48.000 Okay, good. 02:49.000 --> 02:51.000 So, we're just going to talk briefly about what wasm is. 02:51.000 --> 02:55.000 And then we'll get into how this applies to what I just showed you. 02:55.000 --> 03:00.000 So, WebAssembly is a safe portable, low-level execution code format. 03:00.000 --> 03:01.000 Right? 03:01.000 --> 03:03.000 It's safe sandbox execution environment, 03:03.000 --> 03:06.000 where they deny by default security posture. 03:06.000 --> 03:09.000 It makes no assumption about language or host. 03:09.000 --> 03:10.000 Right? 03:10.000 --> 03:13.000 The best analogy the best way to think about it without getting into the details 03:13.000 --> 03:16.000 is that it is a virtual CPU. 03:17.000 --> 03:19.000 It's a different compilation target. 03:19.000 --> 03:22.000 So, just like you can compile things to native platforms, 03:22.000 --> 03:25.000 you can compile things to wasm. 03:25.000 --> 03:28.000 It's a virtualized ISA. 03:28.000 --> 03:31.000 It uses a bytecode binary format. 03:31.000 --> 03:35.000 It uses a stack machine on abstract stack machine model, 03:35.000 --> 03:39.000 with secure stacks for every single execution environment. 03:39.000 --> 03:42.000 So, what does that really mean? 03:42.000 --> 03:43.000 Right? 03:43.000 --> 03:45.000 In a broad sense, it's just another architecture. 03:45.000 --> 03:47.000 This is the way to think about it. 03:47.000 --> 03:50.000 Key difference is that it is virtualized. 03:50.000 --> 03:53.000 It needs a runtime to translate to machine code. 03:53.000 --> 03:55.000 That doesn't mean that it needs to be interpreted, 03:55.000 --> 03:57.000 however, that means that there needs to be a runtime, 03:57.000 --> 03:59.000 and you can compile it ahead of time. 03:59.000 --> 04:01.000 You can even jit in certain cases. 04:01.000 --> 04:03.000 It's universal. 04:03.000 --> 04:04.000 As in, it's an open standard. 04:04.000 --> 04:06.000 It's in probably all of your pockets, 04:06.000 --> 04:09.000 on at least one web browser and on your computers. 04:09.000 --> 04:12.000 Via the browser, which is where it started. 04:12.000 --> 04:15.000 It literally just needs a runtime that fulfills 04:15.000 --> 04:18.000 the two-page spec, which hasn't changed since 2017. 04:18.000 --> 04:21.000 So, easy to client-side technology. 04:21.000 --> 04:23.000 Usually when I talk to people about this, 04:23.000 --> 04:26.000 and they say, oh, you're working with them. 04:26.000 --> 04:27.000 I don't do a lot of front-end stuff, 04:27.000 --> 04:28.000 and I say, that's interesting, 04:28.000 --> 04:31.000 because actually, I only do server-side web assembly stuff. 04:31.000 --> 04:34.000 And we'll talk about how this applies to what I showed you 04:34.000 --> 04:36.000 in the next few slides. 04:36.000 --> 04:39.000 So, it's a safe sandbox execution environment. 04:39.000 --> 04:41.000 It makes no assumption about languages 04:41.000 --> 04:43.000 or host. 04:43.000 --> 04:46.000 Cold start times are in the nano-to-microseconds, 04:46.000 --> 04:48.000 depending on the size. 04:48.000 --> 04:51.000 It's a universal compilation target, like I mentioned. 04:51.000 --> 04:53.000 And so, wasm the joke in the wasm communities 04:53.000 --> 04:57.000 that wasm is neither web nor is it assembly, really. 04:57.000 --> 05:02.000 So, server-side, well, server-side wasm. 05:02.000 --> 05:04.000 I like to think of this, 05:04.000 --> 05:07.000 when people saw these characteristics of this technology 05:07.000 --> 05:09.000 when it first came out in 2017, 05:09.000 --> 05:11.000 a lot of people that were working on it had done 05:11.000 --> 05:13.000 a lot of work and cut their teeth on things 05:13.000 --> 05:15.000 like the JVM in the 2000s. 05:15.000 --> 05:18.000 And when people looked at these things, 05:18.000 --> 05:19.000 and they kind of squinted, 05:19.000 --> 05:21.000 they realized that, okay, 05:21.000 --> 05:25.000 it's a small sandbox at the architecture level. 05:25.000 --> 05:27.000 If you squint, 05:27.000 --> 05:30.000 you could think about other use cases for this, right? 05:30.000 --> 05:33.000 And I like to call this the penicillin moment, right? 05:33.000 --> 05:36.000 It was invented so that it could run safely in browsers, 05:36.000 --> 05:38.000 and now people have taken it and applied it 05:38.000 --> 05:40.000 to many different other things. 05:40.000 --> 05:42.000 Some people talk about this progression. 05:42.000 --> 05:44.000 It may or may not be accurate. 05:44.000 --> 05:46.000 We'll see. 05:46.000 --> 05:49.000 And so, when I said smaller safer faster 05:49.000 --> 05:51.000 and much more universal, 05:51.000 --> 05:53.000 usually someone says, no, no, no, no, no, no. 05:53.000 --> 05:55.000 Faster, it can't be faster, right? 05:55.000 --> 05:57.000 Because it has a run time. 05:57.000 --> 05:58.000 This is true. 05:58.000 --> 05:59.000 This is true. 05:59.000 --> 06:01.000 And that's why faster has an asterisk. 06:01.000 --> 06:03.000 Faster startup times by usually several orders 06:03.000 --> 06:05.000 of magnitude for the same code. 06:05.000 --> 06:08.000 And then as far as the absolute performance 06:08.000 --> 06:11.000 difference between compiled code 06:11.000 --> 06:13.000 on a native platform and wasm, 06:13.000 --> 06:15.000 when it's AOT and optimized, 06:15.000 --> 06:21.000 you can typically get it within about 1.2 to 2 times the speed. 06:21.000 --> 06:23.000 So for certain things, 06:23.000 --> 06:25.000 it could be much, much faster. 06:25.000 --> 06:27.000 Everyone knows this quote. 06:27.000 --> 06:29.000 This is kind of, well, maybe not everyone, 06:29.000 --> 06:31.000 but everyone in my world knows this quote. 06:31.000 --> 06:34.000 And this caused a lot of VCs to become very excited. 06:35.000 --> 06:37.000 Back in 2019. 06:37.000 --> 06:41.000 And, you know, there's an interesting thing about this 06:41.000 --> 06:43.000 because whether or not this has come to fruition 06:43.000 --> 06:46.000 in the last five years is debatable. 06:46.000 --> 06:50.000 So, today, I'm going to talk about Boxer. 06:50.000 --> 06:52.000 Boxer is what I just showed you. 06:52.000 --> 06:54.000 Boxer is a tool that allows you to take 06:54.000 --> 06:57.000 container definitions right now and compile them 06:57.000 --> 06:58.000 to lasm. 06:58.000 --> 07:00.000 But the problem with lasm is that it's just an architecture. 07:00.000 --> 07:01.000 It doesn't have anything. 07:01.000 --> 07:02.000 It doesn't have a kernel. 07:02.000 --> 07:04.000 It doesn't have a LIDC. 07:04.000 --> 07:06.000 It doesn't have any sort of system. 07:06.000 --> 07:08.000 It doesn't know anything about what's going on 07:08.000 --> 07:10.000 outside of the guest. 07:10.000 --> 07:12.000 In WebAssembly, we have these concepts. 07:12.000 --> 07:14.000 The host is the run time. 07:14.000 --> 07:16.000 The guest is where the wasm actually runs. 07:16.000 --> 07:19.000 So, what Boxer does is, 07:19.000 --> 07:22.000 it kind of makes this dev X a little bit easier. 07:22.000 --> 07:24.000 Has anyone actually used WebAssembly 07:24.000 --> 07:25.000 for anything in here? 07:25.000 --> 07:26.000 Okay, cool. 07:26.000 --> 07:28.000 So, you all know how hard it is. 07:28.000 --> 07:30.000 Or maybe what the pain points are. 07:30.000 --> 07:33.000 It's not quite as easy as downloading the 07:33.000 --> 07:35.000 Docker CLI and getting started. 07:35.000 --> 07:38.000 So, I'm going to talk about Boxer a little bit 07:38.000 --> 07:39.000 and how it works. 07:39.000 --> 07:41.000 So, what is in a Box? 07:41.000 --> 07:44.000 Much like a container or the idea of a container 07:44.000 --> 07:46.000 you have a base layer. 07:46.000 --> 07:49.000 The base layer has interfaces. 07:49.000 --> 07:51.000 Those interfaces are very similar to 07:51.000 --> 07:52.000 Pozac's interfaces. 07:52.000 --> 07:55.000 Then, we add several other layers including 07:55.000 --> 07:58.000 a virtualized file system and system code layers. 07:58.000 --> 08:01.000 This is a broad term which I will get into. 08:01.000 --> 08:03.000 It comes with the run time. 08:03.000 --> 08:07.000 And then we have the huge source code. 08:07.000 --> 08:10.000 The way that WebAssembly modules can communicate 08:10.000 --> 08:12.000 with each other and with the outside world 08:12.000 --> 08:15.000 are with the concept of simple imports and exports. 08:15.000 --> 08:18.000 The only thing that a WebAssembly module 08:18.000 --> 08:21.000 knows about is what is exported to it 08:21.000 --> 08:23.000 and what it also imports. 08:23.000 --> 08:26.000 This changes things on many spectrums 08:26.000 --> 08:28.000 and what it comes to trying to rewrite 08:28.000 --> 08:30.000 how system layers can work. 08:30.000 --> 08:32.000 Compared with things like processes 08:32.000 --> 08:35.000 and interprocess communication 08:35.000 --> 08:38.000 and communicating over sockets. 08:38.000 --> 08:40.000 It's that right. 08:40.000 --> 08:41.000 Thanks. 08:41.000 --> 08:44.000 So, how could this be possible? 08:44.000 --> 08:49.000 Here, we start here with the from alpine latest. 08:49.000 --> 08:50.000 And you'll say wait a second. 08:50.000 --> 08:52.000 That isn't actually alpine latest. 08:52.000 --> 08:54.000 That is in your right. 08:54.000 --> 08:57.000 Literally just the interfaces that the alpine image 08:57.000 --> 08:58.000 will give you. 08:58.000 --> 09:01.000 If you want anything else, then you can compile that 09:01.000 --> 09:05.000 and expose it as another WebAssembly module. 09:05.000 --> 09:08.000 Similar thing to what maybe Docker will do. 09:08.000 --> 09:10.000 You can create something inside the environments. 09:10.000 --> 09:11.000 You can copy to it. 09:11.000 --> 09:13.000 You can set the working directory. 09:13.000 --> 09:16.000 And then you can execute it. 09:16.000 --> 09:19.000 A big part of this is the WebAssemblyFS. 09:19.000 --> 09:23.000 The WebAssemblyFS is a possex compliant virtualized file system 09:23.000 --> 09:25.000 that mounts with every single box. 09:25.000 --> 09:27.000 Every single WebAssembly box. 09:27.000 --> 09:28.000 It's written in Rust. 09:28.000 --> 09:30.000 So obviously it's incredible. 09:30.000 --> 09:34.000 And it implements all file system related 09:34.000 --> 09:36.000 Cisco's and then it exposes them. 09:36.000 --> 09:38.000 So, specifically in this project, 09:38.000 --> 09:42.000 what we do is we wrap this raw-space file system 09:42.000 --> 09:45.000 in C and then in C we have what's basically 09:45.000 --> 09:48.000 a lib C which is an amalgamation of 09:48.000 --> 09:51.000 G lib C and muscle in a very 09:51.000 --> 09:53.000 more closely done fashion. 09:53.000 --> 09:56.000 Artfully, also, maybe meaning hacky. 09:56.000 --> 09:59.000 And we expose all the same possex semantics with an 09:59.000 --> 10:01.000 esterist also. 10:01.000 --> 10:04.000 Some key differences with the WebAssemblyFS is that things 10:04.000 --> 10:08.000 like flushing, locking, thinking work like they might 10:08.000 --> 10:12.000 work in the Linux VFS, but not quite as sophisticated yet. 10:12.000 --> 10:14.000 So, there's an interesting thing there. 10:14.000 --> 10:17.000 The other thing about this project is when people 10:17.000 --> 10:20.000 say when we talk about trying to deploy things with 10:20.000 --> 10:23.000 this sort of set of tools, people say, well, it's a virtualized 10:23.000 --> 10:24.000 file system. 10:24.000 --> 10:26.000 So, I'm not sure if I can use that for my project. 10:26.000 --> 10:29.000 And then usually what I say is that you are already 10:29.000 --> 10:30.000 using a virtualized file system. 10:30.000 --> 10:32.000 You just don't know it because it's in the kernel. 10:32.000 --> 10:35.000 And so, the idea is that we just have these 10:35.000 --> 10:36.000 perceptions of this. 10:36.000 --> 10:38.000 A lot of people wonder if this is another fused file system. 10:38.000 --> 10:39.000 It's also not. 10:39.000 --> 10:42.000 It's just made for Wazom specifically with 10:42.000 --> 10:45.000 the communication between modules, which is one of 10:45.000 --> 10:47.000 the key points. 10:47.000 --> 10:51.000 Yeah, I guess I'll do that quickly. 10:51.000 --> 10:54.000 So, right. 10:54.000 --> 10:57.000 So, the Wazom VFS repo, you can see all of the 10:57.000 --> 11:02.000 supported, all of the supported functions. 11:02.000 --> 11:04.000 Great. 11:04.000 --> 11:06.000 Okay. 11:06.000 --> 11:08.000 I love them. 11:08.000 --> 11:10.000 Okay, I've got some big ones. 11:10.000 --> 11:11.000 Okay. 11:11.000 --> 11:12.000 This is a working progress. 11:12.000 --> 11:13.000 And this is my side project. 11:13.000 --> 11:15.000 So, I've been working on this for about three years. 11:15.000 --> 11:16.000 And it's just me. 11:16.000 --> 11:18.000 But with that, we still have a 11:18.000 --> 11:21.000 POSIX sort of surface area that we have worked on. 11:21.000 --> 11:22.000 It's not just me. 11:22.000 --> 11:23.000 It's me and a couple other people. 11:23.000 --> 11:25.000 If you want to get involved, please let me know. 11:25.000 --> 11:26.000 It's all open source. 11:26.000 --> 11:30.000 We have about 70% of POSIX implemented, 11:30.000 --> 11:32.000 which, wow, that seems like a lot. 11:32.000 --> 11:37.000 But the thing with POSIX is that the idea about 11:37.000 --> 11:39.000 POSIX is not necessarily the implementation, 11:39.000 --> 11:41.000 rather than the interface, right? 11:41.000 --> 11:43.000 So, POSIX maybe is just interfaces. 11:43.000 --> 11:47.000 So, we've exposed interfaces with much of the sort of 11:47.000 --> 11:48.000 system functionality. 11:48.000 --> 11:53.000 But how it works, it may look like something that you would see in 11:53.000 --> 11:56.000 Unix or POSIX, but maybe it works slightly differently. 11:56.000 --> 11:57.000 Networking. 11:57.000 --> 11:59.000 This is a whole other talk, right? 11:59.000 --> 12:01.000 So, networking does work. 12:01.000 --> 12:03.000 You can expose networking to the host, right? 12:03.000 --> 12:06.000 If you're running on a server, you can also have this thing 12:06.000 --> 12:09.000 that we're working on now, which is in the browser. 12:09.000 --> 12:11.000 Like I showed you with this. 12:11.000 --> 12:16.000 We have a networking relay, which you can do all 12:16.000 --> 12:19.000 BSD sockets from the browser, just like you would, 12:19.000 --> 12:20.000 and you don't have to think about it. 12:20.000 --> 12:23.000 Using a sort of edge network that we've come up with, 12:23.000 --> 12:26.000 it's kind of experimental, but it can work, which is interesting. 12:26.000 --> 12:29.000 So, the interesting thing about that is you could do the same work 12:29.000 --> 12:30.000 in the browser. 12:30.000 --> 12:32.000 It could even be your coding environments. 12:32.000 --> 12:34.000 You could take that, and then you can move it to a server. 12:34.000 --> 12:36.000 You could also move it to an embedded device, 12:36.000 --> 12:39.000 as long as there was one times everywhere. 12:39.000 --> 12:46.000 So, lower layers. 12:46.000 --> 12:48.000 How does this actually work? 12:48.000 --> 12:51.000 It actually works with a project called Markot. 12:51.000 --> 12:55.000 Markot, we're in Belgium, so my French is not very good, 12:55.000 --> 12:57.000 but in old French, this means like substrate, right? 12:57.000 --> 12:58.000 Like a lower layer. 12:58.000 --> 12:59.000 Right? 12:59.000 --> 13:03.000 So, Markot basically is what I told you about the Frankenstein 13:03.000 --> 13:07.000 lib C operation that I do, where we have G lib C and Muscle. 13:07.000 --> 13:09.000 And we sort of expose all of those things. 13:09.000 --> 13:15.000 And we have a matrix of every single positive function 13:15.000 --> 13:18.000 and how well supported it is, right? 13:18.000 --> 13:23.000 And some of the more challenging ones are at the end, 13:23.000 --> 13:27.000 where it's coming to things like process and thread management, 13:27.000 --> 13:30.000 signals, time and date, and it's not too hard. 13:30.000 --> 13:32.000 Memory management also not very hard. 13:32.000 --> 13:35.000 But we've about 30% left that we're trying to implement. 13:35.000 --> 13:40.000 We're open to ideas, but the idea is that this sort of thesis for this 13:40.000 --> 13:45.000 was that with many containers, the average container user 13:45.000 --> 13:48.000 may not care about what the architecture is. 13:48.000 --> 13:50.000 Maybe 5% do, right? 13:50.000 --> 13:53.000 There's also parts of the puzzle, because the puzzle is a little bit 13:53.000 --> 13:56.000 of a leafy abstraction in places, like if you've ever had the 13:56.000 --> 13:59.000 bad fortune to use a set jump and long jump, 13:59.000 --> 14:03.000 you will understand that it is tied to specific architecture. 14:04.000 --> 14:07.000 There are no registers in WebAssembly, right? 14:07.000 --> 14:09.000 So there's quite a few challenges there. 14:09.000 --> 14:14.000 But for 95% of things, it's very possible that you could run a container 14:14.000 --> 14:20.000 in a smaller, safer, faster, and more universal environments. 14:20.000 --> 14:23.000 This is a picture that I had to chat to IPT make for me, 14:23.000 --> 14:28.000 and I think it kind of fits kind of a fun thing. 14:28.000 --> 14:30.000 And some metrics. 14:30.000 --> 14:33.000 So there are in the AI revolution that we're in. 14:33.000 --> 14:36.000 There's a lot of Python users, a lot of people want to run 14:36.000 --> 14:37.000 Python in the browser. 14:37.000 --> 14:40.000 There's already works with WebAssembly things like 14:40.000 --> 14:43.000 Piedide and with Jupiter light, Jupiter notebooks. 14:43.000 --> 14:46.000 So this is just a small matrix of the differences, right? 14:46.000 --> 14:49.000 The same app with, and this doesn't also include the 14:49.000 --> 14:51.000 kernel and the runtime for the container. 14:51.000 --> 14:53.000 This is just the container, right? 14:53.000 --> 14:56.000 So 80 to 900 megabytes, start of speed, 14:56.000 --> 14:58.000 could be in the seconds. 14:58.000 --> 15:01.000 The difference is that the same box with all of the functionality 15:01.000 --> 15:03.000 in it, six megabytes. 15:03.000 --> 15:05.000 And if you extract just the source code, 15:05.000 --> 15:09.000 500 kilobytes, you can move that around to different environments 15:09.000 --> 15:13.000 where those system functions are snapshot it. 15:13.000 --> 15:15.000 I'm going to move a little quicker. 15:15.000 --> 15:17.000 So paradigm shift. 15:17.000 --> 15:19.000 Maybe this is possible, maybe it's crazy, 15:19.000 --> 15:22.000 but this would allow for some really interesting things in the future. 15:22.000 --> 15:24.000 Like a composable universal, 15:24.000 --> 15:28.000 based operating environment in a more distributed fashion. 15:28.000 --> 15:29.000 True isomorphism. 15:29.000 --> 15:32.000 If you remember the Node.js craze from a few years ago, 15:32.000 --> 15:34.000 this is actually isomorphic, right? 15:34.000 --> 15:36.000 You're actually running the same exact code everywhere, 15:36.000 --> 15:38.000 but in a polyglot fashion. 15:38.000 --> 15:41.000 And then finally, if we have any fans with plan 9 in here, 15:41.000 --> 15:45.000 the idea of a secure CPU is a secure stack for every single execution. 15:45.000 --> 15:48.000 This is something that comes a little closer to that. 15:48.000 --> 15:51.000 If you have more questions, I would love to talk with you after. 15:51.000 --> 15:52.000 So thank you. 15:52.000 --> 15:53.000 So thank you. 15:53.000 --> 16:02.000 So thank you. 16:02.000 --> 16:04.000 So thank you. 16:04.000 --> 16:06.000 We actually have time for one question, I guess. 16:06.000 --> 16:08.000 Is there one question? 16:08.000 --> 16:09.000 Yeah, there's one. 16:09.000 --> 16:10.000 Sure. 16:10.000 --> 16:12.000 Just a general question. 16:12.000 --> 16:14.000 I didn't hear you mention anything about the component model. 16:14.000 --> 16:16.000 I just wondered how this ties in. 16:16.000 --> 16:17.000 $1,000. 16:17.000 --> 16:18.000 It's an excellent question. 16:18.000 --> 16:19.000 The component model. 16:19.000 --> 16:20.000 Yeah. 16:20.000 --> 16:21.000 I don't know if anyone else has heard of the component model. 16:21.000 --> 16:22.000 Yeah. 16:22.000 --> 16:23.000 Okay, cool, cool. 16:23.000 --> 16:25.000 So the component model is, that's with Wazzy. 16:25.000 --> 16:27.000 I didn't talk about Wazzy at all. 16:27.000 --> 16:30.000 The component model is a really cool idea. 16:30.000 --> 16:34.000 And it's basically strongly typed function wrappers around Wazzy modules. 16:34.000 --> 16:37.000 And it's composable at different levels. 16:37.000 --> 16:41.000 But they are many people in that group. 16:41.000 --> 16:44.000 And I'm actually part of the standard group with Wazzy. 16:44.000 --> 16:47.000 That group is not as interested in providing 16:47.000 --> 16:48.000 physics compatibility. 16:48.000 --> 16:53.000 In fact, they kind of see that as a separate thing. 16:53.000 --> 16:54.000 Yeah. 16:54.000 --> 16:58.000 But the thing is, with these interfaces, at the lower levels, those could be components. 16:58.000 --> 17:00.000 And you could have components all the way down. 17:00.000 --> 17:03.000 And so certain interfaces could be componentized. 17:03.000 --> 17:05.000 And that's just a different implementation detail. 17:05.000 --> 17:06.000 So, yeah. 17:06.000 --> 17:07.000 So, thank you. 17:07.000 --> 17:08.000 Thank you. 17:08.000 --> 17:11.000 Thank you. 17:17.000 --> 17:19.000 Thank you.