WEBVTT NOTE This file was generated by switchtitle 1 00:00:13.708 --> 00:00:15.708 This is Self-Directed Research. Our 2 00:00:15.708 --> 00:00:17.000 hosts, James and Amos, 3 00:00:17.000 --> 00:00:18.416 host weekly hyper-focused 4 00:00:18.416 --> 00:00:20.750 technical deep dives. In today's episode, 5 00:00:21.125 --> 00:00:22.833 James makes a pitch for "Traceability." 6 00:00:23.541 --> 00:00:24.916 Make sure to like, follow, and subscribe 7 00:00:24.916 --> 00:00:26.750 wherever you find us. And as always, 8 00:00:26.750 --> 00:00:29.625 visit sdr-podcast.com/episodes for 9 00:00:29.625 --> 00:00:29.791 all the 10 00:00:29.791 --> 00:00:31.625 presentations, videos, show notes, 11 00:00:31.625 --> 00:00:34.000 and transcripts. And of course, stay 12 00:00:34.000 --> 00:00:35.041 tuned to the end to hear more 13 00:00:35.041 --> 00:00:36.750 about Depot, the sponsor of this 14 00:00:36.750 --> 00:00:37.166 episode. 15 00:00:42.875 --> 00:00:43.916 Slides, because I haven't looked these in 16 00:00:43.916 --> 00:00:44.583 like a week and a half. 17 00:00:44.958 --> 00:00:45.291 James. 18 00:00:46.041 --> 00:00:49.791 I was like organizing before this, like 19 00:00:49.791 --> 00:00:51.541 all of, like getting all of this audio 20 00:00:51.541 --> 00:00:53.416 and everything into the season two blah, 21 00:00:53.416 --> 00:00:55.083 blah, blah, DaVinci Resolve thing. 22 00:00:55.083 --> 00:00:55.875 It's like you're unionizing. 23 00:00:56.291 --> 00:00:58.666 And then I was like, "Oh shit," like what 24 00:00:58.666 --> 00:01:00.375 was it season two episode three? 25 00:01:01.125 --> 00:01:02.458 And I was like, "What did we even do? 26 00:01:02.458 --> 00:01:03.875 I can't find any slides anywhere. 27 00:01:03.875 --> 00:01:05.583 James, like what did you present about?" 28 00:01:05.583 --> 00:01:07.416 And then like I opened the audio and it 29 00:01:07.416 --> 00:01:08.500 was just like, "And today 30 00:01:08.500 --> 00:01:10.666 I'm not gonna have any slides." 31 00:01:10.666 --> 00:01:12.541 And I'm like, "Okay, good, good." 32 00:01:12.666 --> 00:01:13.833 I remember that. 33 00:01:14.208 --> 00:01:14.916 You were supposed to 34 00:01:14.916 --> 00:01:16.166 make some slides, yeah. 35 00:01:16.375 --> 00:01:18.000 Oh yeah, I should probably go back and 36 00:01:18.000 --> 00:01:19.333 make those slides, yeah. 37 00:01:19.625 --> 00:01:21.791 I have not started editing that one yet, 38 00:01:21.958 --> 00:01:23.708 so I did not yet, but. 39 00:01:23.958 --> 00:01:26.291 (Sighs) 40 00:01:27.041 --> 00:01:29.125 Yeah, I don't think I'm that vain, but 41 00:01:29.125 --> 00:01:31.291 listening to yourself in studio grade 42 00:01:31.291 --> 00:01:32.500 headphones is 43 00:01:32.500 --> 00:01:33.750 fascinating, it's mesmerizing. 44 00:01:34.041 --> 00:01:34.958 I can't do it. 45 00:01:34.958 --> 00:01:36.625 Like I get like one watch of any talk 46 00:01:36.625 --> 00:01:37.958 that I give, then I 47 00:01:37.958 --> 00:01:39.250 have to not listen to it. 48 00:01:39.541 --> 00:01:40.916 Or I have to be in a very good mood when 49 00:01:40.916 --> 00:01:41.541 I listen to it, because 50 00:01:41.541 --> 00:01:43.750 otherwise it's just like, "UGH!" 51 00:01:44.041 --> 00:01:46.291 So you're saying I am a narcissist. 52 00:01:46.625 --> 00:01:47.250 That's what you're saying. 53 00:01:47.708 --> 00:01:49.750 No, you have less anxiety of 54 00:01:49.750 --> 00:01:52.250 your own speaking patterns. 55 00:01:52.250 --> 00:01:52.666 I don't know. 56 00:01:52.791 --> 00:01:54.000 Have I just done it more than 57 00:01:54.000 --> 00:01:54.916 you, is that possible? 58 00:01:55.000 --> 00:01:56.125 That's probably, yeah. 59 00:01:56.125 --> 00:01:57.291 You've listened to yourself a lot more 60 00:01:57.291 --> 00:01:58.583 than me probably, because you've done all 61 00:01:58.583 --> 00:02:00.166 your own editing and production. 62 00:02:00.500 --> 00:02:02.708 I have been working way too much, like 60 63 00:02:02.708 --> 00:02:04.041 hours last week, it's not good. 64 00:02:04.291 --> 00:02:05.708 And I'm not getting startup founder 65 00:02:05.708 --> 00:02:07.291 money, is the sad part. 66 00:02:07.291 --> 00:02:07.958 Oh darn. 67 00:02:08.041 --> 00:02:08.708 I know. 68 00:02:09.000 --> 00:02:09.750 Yeah, I have Poststation shipped. 69 00:02:10.000 --> 00:02:12.541 Yeah, I saw, how did I learn about it 70 00:02:12.666 --> 00:02:14.250 from frickin' Bluesky? 71 00:02:14.750 --> 00:02:16.000 Why didn't you ping me? 72 00:02:16.041 --> 00:02:17.583 Because we haven't sat down and had a 73 00:02:17.583 --> 00:02:18.833 video call in like three weeks. 74 00:02:20.083 --> 00:02:22.125 Welcome back to, "So... the podcast." 75 00:02:22.541 --> 00:02:23.291 So today- 76 00:02:23.625 --> 00:02:24.791 Okay, bye, I'm leaving. 77 00:02:24.833 --> 00:02:25.291 Bye, Amanda. 78 00:02:25.375 --> 00:02:25.916 Get out of here, Amanda. 79 00:02:26.250 --> 00:02:27.000 Anyway, James. 80 00:02:27.000 --> 00:02:28.791 Okay, I have a topic that's a little out 81 00:02:28.791 --> 00:02:30.541 of left field, comes from my background 82 00:02:30.541 --> 00:02:32.541 in safety critical, but it is a thing 83 00:02:32.541 --> 00:02:34.041 that I wish we had more of. 84 00:02:34.041 --> 00:02:35.791 And that thing is traceability. 85 00:02:36.250 --> 00:02:37.666 And really what I wish is that we had 86 00:02:37.666 --> 00:02:39.291 good open source traceability tools. 87 00:02:39.916 --> 00:02:41.250 So, Amos, have you heard 88 00:02:41.250 --> 00:02:43.000 of traceability before? 89 00:02:43.041 --> 00:02:44.166 I have heard of the tracing 90 00:02:44.166 --> 00:02:45.500 crate. Are those related? 91 00:02:45.583 --> 00:02:47.000 Okay, nope, this is one of those 92 00:02:47.000 --> 00:02:49.291 traditional software, one word means 30 93 00:02:49.291 --> 00:02:50.000 different things in 94 00:02:50.000 --> 00:02:50.916 30 different contexts. 95 00:02:51.416 --> 00:02:52.916 Is it reproducible build? 96 00:02:54.291 --> 00:02:55.375 No? No. 97 00:02:56.083 --> 00:02:58.250 It does come from safety critical fields. 98 00:02:59.000 --> 00:03:01.375 And generally what it means is it's a way 99 00:03:01.625 --> 00:03:02.916 to link different pieces 100 00:03:02.916 --> 00:03:04.000 of information together. 101 00:03:04.458 --> 00:03:05.625 So whether you're talking about 102 00:03:05.625 --> 00:03:08.333 requirements or documents to your code, 103 00:03:08.333 --> 00:03:10.916 to your tests, to even documents to 104 00:03:10.916 --> 00:03:12.416 documents like, hey, the government says 105 00:03:12.416 --> 00:03:14.625 we have to do this, this and this, and we 106 00:03:14.625 --> 00:03:16.416 promised our customers we were gonna do 107 00:03:16.416 --> 00:03:17.416 this, this and this. 108 00:03:17.708 --> 00:03:19.041 We should probably make sure that we've 109 00:03:19.041 --> 00:03:21.000 done all of those things or like the 110 00:03:21.000 --> 00:03:22.291 superset of all of those things. 111 00:03:22.625 --> 00:03:23.666 And then we should probably make sure 112 00:03:23.666 --> 00:03:25.375 that we've tested that we do all of those 113 00:03:25.375 --> 00:03:26.666 things, that we do them correctly. 114 00:03:27.000 --> 00:03:28.000 And we should probably make sure that 115 00:03:28.000 --> 00:03:30.541 we've tested all of the code that we've 116 00:03:30.541 --> 00:03:32.500 written or that there is a test for 117 00:03:32.500 --> 00:03:34.041 everything or we didn't just invent 118 00:03:34.041 --> 00:03:35.958 something that we're doing for no good 119 00:03:35.958 --> 00:03:37.791 reason if we didn't say we were gonna do 120 00:03:37.791 --> 00:03:39.166 it or we're not required to 121 00:03:39.166 --> 00:03:40.791 do it or things like that. 122 00:03:41.333 --> 00:03:43.375 Right, do you know the classic story 123 00:03:43.375 --> 00:03:45.208 about this where like there's a mom that 124 00:03:45.208 --> 00:03:48.958 always cuts the edges of some meat before 125 00:03:48.958 --> 00:03:51.000 she cooks it in the oven and then 126 00:03:51.000 --> 00:03:52.791 eventually the daughter asks like 'why 127 00:03:52.791 --> 00:03:54.666 you're doing that' and then she doesn't. 128 00:03:54.666 --> 00:03:56.125 So she asks her mother, she doesn't, they 129 00:03:56.125 --> 00:03:57.541 go back all the way and it's like, 'oh, 130 00:03:57.541 --> 00:03:59.000 because the oven was too small. 131 00:03:59.000 --> 00:04:00.083 So we had to cut the 132 00:04:00.083 --> 00:04:01.250 edges so it would fit in.' 133 00:04:01.916 --> 00:04:02.833 They did it for generations. 134 00:04:03.500 --> 00:04:04.625 Exactly that kind of thing. 135 00:04:04.875 --> 00:04:06.958 The example that was always given to me... 136 00:04:07.208 --> 00:04:08.708 so in safety critical fields, usually 137 00:04:08.708 --> 00:04:09.916 traceability is a requirement. 138 00:04:10.291 --> 00:04:11.333 It's one of those things you have to 139 00:04:11.333 --> 00:04:13.458 include in your big piece of paper that 140 00:04:13.458 --> 00:04:15.750 says like: yes, all of my requirements 141 00:04:15.750 --> 00:04:18.333 are documented, all of the code exists to 142 00:04:18.333 --> 00:04:19.916 fulfill a requirement, all the tests 143 00:04:19.916 --> 00:04:22.416 exist to fulfill code and a requirement 144 00:04:22.666 --> 00:04:24.291 and we're not missing anything. 145 00:04:24.291 --> 00:04:25.208 There's no extra code, 146 00:04:25.208 --> 00:04:26.125 there's no extra tests. 147 00:04:26.458 --> 00:04:28.041 The example that was given to us in 148 00:04:28.041 --> 00:04:30.250 safety critical training was: you could put 149 00:04:30.250 --> 00:04:32.583 like a Pac-Man game in a piece of 150 00:04:32.583 --> 00:04:34.625 avionics but you better have requirements 151 00:04:34.625 --> 00:04:36.416 for it and you better have tests for it 152 00:04:36.625 --> 00:04:38.625 because if you just put Easter eggs in 153 00:04:38.625 --> 00:04:40.583 there and they don't map to some 154 00:04:40.583 --> 00:04:42.333 requirement, then they should not be 155 00:04:42.333 --> 00:04:44.291 there because you've never said what 156 00:04:44.291 --> 00:04:46.291 they're supposed to do or not do. And it's 157 00:04:46.291 --> 00:04:48.083 a little silly because you wouldn't wanna 158 00:04:48.083 --> 00:04:50.416 sit down and verify Pac-Man is working 159 00:04:50.416 --> 00:04:51.791 correctly to aviation standards. 160 00:04:52.250 --> 00:04:54.041 But I think this is one of those things 161 00:04:54.041 --> 00:04:55.750 where a lot of the tools that exist for 162 00:04:55.750 --> 00:04:58.250 this are outrageously expensive and 163 00:04:58.250 --> 00:05:00.000 usually their UX is terrible because 164 00:05:00.000 --> 00:05:02.208 their market is like 50 automotive 165 00:05:02.208 --> 00:05:05.250 companies and 10 aviation companies and 166 00:05:05.250 --> 00:05:07.250 some medical device companies and the 167 00:05:07.250 --> 00:05:09.875 iconic document capture tool for this is 168 00:05:09.875 --> 00:05:13.458 IBM DOORS which is like a tool straight 169 00:05:13.458 --> 00:05:16.375 out of the 90s and it's like a very fancy 170 00:05:16.375 --> 00:05:18.458 version of Word with way-- I don't know, 171 00:05:18.458 --> 00:05:20.000 I'm probably casting aspersions on it. 172 00:05:20.000 --> 00:05:22.833 I haven't used DOORS in 10, 15 years but 173 00:05:22.833 --> 00:05:24.916 it's a tool that you only use if you have 174 00:05:24.916 --> 00:05:25.958 to and you pay a lot of 175 00:05:25.958 --> 00:05:28.041 money to a few set of companies. 176 00:05:28.041 --> 00:05:28.375 I bet. 177 00:05:28.791 --> 00:05:29.958 Because you have to-- 178 00:05:29.958 --> 00:05:31.625 I thought I was edgy with Apple scripts and 179 00:05:31.625 --> 00:05:34.000 you bring out IBM stuff, okay. 180 00:05:34.000 --> 00:05:35.208 And there are some companies, like I've 181 00:05:35.208 --> 00:05:36.666 worked at companies that rolled their own, 182 00:05:36.958 --> 00:05:38.166 but they were companies that were big 183 00:05:38.166 --> 00:05:40.333 enough where the cost benefit ratio, it 184 00:05:40.333 --> 00:05:42.041 was cheaper to pay an in-house dev tool 185 00:05:42.041 --> 00:05:44.333 team, like 10 people full time to work on 186 00:05:44.333 --> 00:05:45.583 that tool than it was to pay the 187 00:05:45.583 --> 00:05:46.958 licensing cost for that. 188 00:05:47.208 --> 00:05:48.500 Whereas if you're a smaller company or 189 00:05:48.500 --> 00:05:50.291 you're only making a couple products, you 190 00:05:50.291 --> 00:05:52.375 pay through the nose to have a tool that 191 00:05:52.375 --> 00:05:54.083 when you show it to a regulator, they're 192 00:05:54.083 --> 00:05:55.833 gonna go, "Oh yeah, you use that tool, I 193 00:05:55.833 --> 00:05:57.750 know how to check your configuration, I 194 00:05:57.750 --> 00:05:58.375 know how to check your 195 00:05:58.375 --> 00:05:59.708 reports, no surprises." 196 00:06:00.041 --> 00:06:01.750 A lot of engineering in general but 197 00:06:01.750 --> 00:06:02.916 especially like safety critical 198 00:06:02.916 --> 00:06:03.708 engineering is just 199 00:06:03.708 --> 00:06:05.750 avoiding surprises to regulators. 200 00:06:06.083 --> 00:06:08.041 So just making sure that they go, "Yeah, I 201 00:06:08.041 --> 00:06:08.875 know what I'm seeing and 202 00:06:08.875 --> 00:06:10.083 I know how to verify it." 203 00:06:10.166 --> 00:06:11.833 But I've also worked in teams where this 204 00:06:11.833 --> 00:06:13.916 wasn't just a checkbox, this was like 205 00:06:13.916 --> 00:06:15.166 part of the whole process. 206 00:06:15.625 --> 00:06:18.208 And there's some real like development 207 00:06:18.208 --> 00:06:21.750 flow stuff that I desperately miss from 208 00:06:21.750 --> 00:06:23.125 having that tooling set up. 209 00:06:23.125 --> 00:06:24.708 And I've not found any open source 210 00:06:24.708 --> 00:06:26.666 tooling... There's some open source tooling 211 00:06:26.666 --> 00:06:27.833 that gets near this and there's some 212 00:06:27.833 --> 00:06:28.416 people who have built 213 00:06:28.416 --> 00:06:29.500 their own on top of it. 214 00:06:29.500 --> 00:06:31.208 For example, like Ferrocene is written, 215 00:06:31.708 --> 00:06:33.250 their requirements document is written in 216 00:06:33.250 --> 00:06:34.500 Sphinx which is a Python 217 00:06:34.500 --> 00:06:36.000 tool for specify requirement. 218 00:06:36.208 --> 00:06:38.166 I think they either use some extensions 219 00:06:38.166 --> 00:06:40.291 or built some extensions and they can 220 00:06:40.291 --> 00:06:42.000 like annotate requirements and that's 221 00:06:42.000 --> 00:06:42.791 where like the Ferrocene 222 00:06:42.791 --> 00:06:44.875 language spec is specified in. 223 00:06:45.250 --> 00:06:46.750 It's some dialect of Sphinx. 224 00:06:47.125 --> 00:06:48.125 So it's just a way of 225 00:06:48.125 --> 00:06:49.208 marking up documents. 226 00:06:50.208 --> 00:06:50.416 Interesting. 227 00:06:50.625 --> 00:06:52.208 Python would not be where I would start 228 00:06:52.208 --> 00:06:52.958 for something like that 229 00:06:52.958 --> 00:06:54.833 but I also don't do it. 230 00:06:54.833 --> 00:06:56.125 There's a lot of these tools. 231 00:06:56.125 --> 00:06:57.416 The other one that I can think of is from 232 00:06:57.416 --> 00:07:00.708 the, like the web world, WHATWG I think, 233 00:07:00.708 --> 00:07:02.250 they use a tool which is hilariously 234 00:07:02.250 --> 00:07:04.291 called Bikeshed which is a requirements 235 00:07:04.291 --> 00:07:06.041 capture tool, which is excellent. 236 00:07:06.125 --> 00:07:07.000 It's self-aware. 237 00:07:07.375 --> 00:07:08.041 Yeah, exactly. 238 00:07:08.375 --> 00:07:10.250 And it's made to make these requirements 239 00:07:10.250 --> 00:07:11.500 documents and things like that, but-- 240 00:07:12.000 --> 00:07:13.500 The WHATWG is the version 241 00:07:13.500 --> 00:07:15.041 of the W3C that matters. 242 00:07:15.250 --> 00:07:16.083 Well, it's a different 243 00:07:16.083 --> 00:07:17.666 consortium that actually works. 244 00:07:18.375 --> 00:07:21.208 From what I know, I just offended like 12 245 00:07:21.208 --> 00:07:21.708 different people 246 00:07:21.708 --> 00:07:22.583 listening to this podcast. 247 00:07:22.833 --> 00:07:23.458 I will send you 248 00:07:23.458 --> 00:07:25.000 individually penned apologies. 249 00:07:25.291 --> 00:07:26.458 These are things where I'd have to go in 250 00:07:26.458 --> 00:07:29.375 like DM Simon Sapin and like Manish to 251 00:07:29.375 --> 00:07:30.916 figure out like: okay, how 252 00:07:30.916 --> 00:07:32.166 should I feel about these groups? 253 00:07:32.166 --> 00:07:33.250 Because those are the people who have the 254 00:07:33.250 --> 00:07:35.125 opinions that I trust of 255 00:07:35.125 --> 00:07:36.791 those groups and things like that. 256 00:07:36.833 --> 00:07:39.583 The logo for the WHATWG is just a question 257 00:07:39.583 --> 00:07:41.250 mark which I feel is very appropriate. 258 00:07:41.583 --> 00:07:42.416 It's how I feel about it. 259 00:07:43.458 --> 00:07:45.000 So let me explain why I miss this. 260 00:07:45.291 --> 00:07:46.875 So at least in Rust, we have a pretty 261 00:07:46.875 --> 00:07:47.791 good culture of having 262 00:07:47.791 --> 00:07:49.333 docs, code and tests. 263 00:07:49.625 --> 00:07:52.166 You usually have a book written in like 264 00:07:52.166 --> 00:07:56.250 MD book or you have doc tests or you just 265 00:07:56.250 --> 00:07:58.083 have like module level docs, so you have 266 00:07:58.083 --> 00:08:00.916 some level of docs and those are 267 00:08:00.916 --> 00:08:02.875 basically requirements like... or they can 268 00:08:02.875 --> 00:08:03.916 be think of to take the 269 00:08:03.916 --> 00:08:04.791 place of requirements. 270 00:08:04.791 --> 00:08:06.708 They're saying what the code should do in 271 00:08:06.708 --> 00:08:09.000 a non-code form through examples or 272 00:08:09.000 --> 00:08:10.750 written prose or things like that. 273 00:08:10.791 --> 00:08:12.208 But they're not like preconditions like 274 00:08:12.208 --> 00:08:13.500 you got in some other languages. 275 00:08:13.916 --> 00:08:16.000 Like... I think people coming from maybe Ada 276 00:08:16.000 --> 00:08:17.583 or something would think that what we 277 00:08:17.583 --> 00:08:19.583 have in Rust is not good enough. 278 00:08:19.583 --> 00:08:20.916 It doesn't go far enough I think. 279 00:08:21.083 --> 00:08:22.500 There's all different kinds of things and 280 00:08:22.500 --> 00:08:24.000 like this is like formal verification, 281 00:08:24.291 --> 00:08:25.791 not formal in the same way safety 282 00:08:25.791 --> 00:08:27.208 critical is, but like formally 283 00:08:27.208 --> 00:08:29.500 mathematically verified stuff is- it's 284 00:08:29.500 --> 00:08:31.416 kind of like using code to write your 285 00:08:31.416 --> 00:08:33.125 docs or there's even like, what is it? 286 00:08:33.333 --> 00:08:35.000 Cucumber tests and things like that, 287 00:08:35.000 --> 00:08:36.625 where you say like: as a user, I want my 288 00:08:36.625 --> 00:08:39.125 session to be retained when I logged in 289 00:08:39.208 --> 00:08:40.541 and you write prose to write 290 00:08:40.541 --> 00:08:41.666 tests and things like that. 291 00:08:41.958 --> 00:08:42.875 So I mean, there's different ways to 292 00:08:42.875 --> 00:08:45.208 slice all of this, but in general you 293 00:08:45.208 --> 00:08:46.666 have some kind of documentation, some 294 00:08:46.666 --> 00:08:47.541 kind of description of what 295 00:08:47.541 --> 00:08:48.791 your code's supposed to do. 296 00:08:48.791 --> 00:08:50.166 You have your code where you actually do 297 00:08:50.166 --> 00:08:51.666 it and then you probably have some kind 298 00:08:51.666 --> 00:08:53.458 of unit tests or integration tests where 299 00:08:53.458 --> 00:08:55.458 you say like, yes, it actually does what 300 00:08:55.458 --> 00:08:58.250 I intended and... You just kind of maintain 301 00:08:58.250 --> 00:09:00.416 those all as a bag of things in a repo. 302 00:09:00.791 --> 00:09:02.416 And like these days in open source, 303 00:09:02.541 --> 00:09:04.208 usually the split is better because most 304 00:09:04.208 --> 00:09:05.416 people are putting all three of those 305 00:09:05.416 --> 00:09:06.750 things in one Git repo. 306 00:09:07.208 --> 00:09:08.083 So they do travel 307 00:09:08.083 --> 00:09:09.333 together, which is nice. 308 00:09:09.333 --> 00:09:11.000 Like they travel as a crate or as a 309 00:09:11.000 --> 00:09:13.291 workspace or whatever, which is not 310 00:09:13.291 --> 00:09:15.750 always what like commercial or corporate 311 00:09:15.750 --> 00:09:17.291 projects used to do at least where you'd 312 00:09:17.291 --> 00:09:19.416 have like whatever Excel docs or Word 313 00:09:19.416 --> 00:09:21.083 docs that were captured somewhere and you 314 00:09:21.083 --> 00:09:23.166 might have your code in one weird source 315 00:09:23.166 --> 00:09:25.458 control thing and tests might be some 316 00:09:25.458 --> 00:09:26.500 paperwork that you give 317 00:09:26.500 --> 00:09:28.083 people to do on every release. 318 00:09:28.083 --> 00:09:29.833 At least like in open source, we tend to 319 00:09:29.833 --> 00:09:32.708 have a fairly grouped set of all of these 320 00:09:32.708 --> 00:09:33.958 assets in one place. 321 00:09:33.958 --> 00:09:35.041 But if you go through and you're doing 322 00:09:35.041 --> 00:09:37.666 some like, optimization or just fixing a 323 00:09:37.666 --> 00:09:39.541 bug or something like that, you probably 324 00:09:39.541 --> 00:09:42.916 change some stuff and you probably make 325 00:09:42.916 --> 00:09:43.875 sure your tests work. 326 00:09:44.125 --> 00:09:46.333 And if you have doc tests and you did 327 00:09:46.333 --> 00:09:47.833 change something fundamental enough or 328 00:09:47.833 --> 00:09:49.458 your doc tests now break, because in 329 00:09:49.458 --> 00:09:51.375 Rust, the examples that you put in your 330 00:09:51.375 --> 00:09:53.458 docs will get run as unit tests, at least 331 00:09:53.458 --> 00:09:54.708 that they compile and 332 00:09:54.708 --> 00:09:55.916 sometimes that they run as well. 333 00:09:56.250 --> 00:09:58.541 But outside of that, if you rearchitected 334 00:09:58.541 --> 00:10:00.458 something major or if you fundamentally 335 00:10:00.458 --> 00:10:02.791 added or changed something, nothing's 336 00:10:02.791 --> 00:10:04.625 really gonna tell you if you have a whole 337 00:10:04.625 --> 00:10:06.250 section of your documents that are 338 00:10:06.250 --> 00:10:09.208 totally not relevant anymore or totally 339 00:10:09.208 --> 00:10:12.166 incorrect or you don't know if you 340 00:10:12.166 --> 00:10:13.833 changed something and all of your tests 341 00:10:13.833 --> 00:10:15.500 pass, but you just added a bunch of new 342 00:10:15.500 --> 00:10:17.875 code that is completely untested and 343 00:10:17.875 --> 00:10:18.958 maybe doesn't work the 344 00:10:18.958 --> 00:10:19.791 way you think it does. 345 00:10:20.291 --> 00:10:21.708 So when I've worked in these environments 346 00:10:21.708 --> 00:10:23.208 before where you had real traceability 347 00:10:23.208 --> 00:10:25.458 tools and you had linked all of your docs 348 00:10:25.458 --> 00:10:27.166 to your code and your code to your tests 349 00:10:27.541 --> 00:10:28.958 and then usually your tests back to your 350 00:10:28.958 --> 00:10:31.708 docs, what if instead of just when you do 351 00:10:31.708 --> 00:10:34.166 a PR, the things that changed pop up, 352 00:10:34.541 --> 00:10:37.500 what if also everything one step away 353 00:10:37.500 --> 00:10:40.041 from that thing also came up for 354 00:10:40.041 --> 00:10:41.833 re-review to make sure 355 00:10:41.833 --> 00:10:42.625 it was still accurate? 356 00:10:43.000 --> 00:10:45.625 So if you said this paragraph of docs 357 00:10:45.916 --> 00:10:47.708 goes with this chunk of code, whenever 358 00:10:47.708 --> 00:10:49.708 you change that code, it would present to 359 00:10:49.708 --> 00:10:50.833 you: hey, here's this paragraph. 360 00:10:51.041 --> 00:10:52.208 Is this paragraph still right? 361 00:10:52.500 --> 00:10:53.958 Or hey, there are these tests. 362 00:10:54.333 --> 00:10:56.750 You didn't tell me where the new tests 363 00:10:56.750 --> 00:10:59.083 for the new functionality are or are all 364 00:10:59.083 --> 00:11:01.041 these tests still testing something 365 00:11:01.041 --> 00:11:03.125 worthwhile or are they just trivial or 366 00:11:03.125 --> 00:11:05.500 they're passing for the wrong reason or 367 00:11:05.500 --> 00:11:06.666 something like that now. 368 00:11:06.875 --> 00:11:08.416 I would imagine the value in that is only 369 00:11:08.416 --> 00:11:10.375 as good as the annotations that you leave 370 00:11:10.375 --> 00:11:12.083 in your docs sources though. 371 00:11:12.416 --> 00:11:15.541 Yeah, there's no magic here. 372 00:11:15.541 --> 00:11:16.875 At least in these different tools, 373 00:11:16.875 --> 00:11:20.666 usually you get a unique ID for functions 374 00:11:20.958 --> 00:11:23.083 and lines of your docs and 375 00:11:23.083 --> 00:11:24.250 your tests and things like that. 376 00:11:24.250 --> 00:11:26.666 So you have some like project global 377 00:11:26.666 --> 00:11:29.166 numeric store and they act as like 378 00:11:29.166 --> 00:11:30.750 permalinks for all of these things. 379 00:11:31.041 --> 00:11:32.250 When you write your tests, you might say: 380 00:11:32.250 --> 00:11:34.625 verifies either this function name or 381 00:11:34.625 --> 00:11:37.291 verifies some ID that goes with that. 382 00:11:37.291 --> 00:11:39.291 Or when you write the code, you might say: 383 00:11:39.291 --> 00:11:41.166 this goes with this text 384 00:11:41.166 --> 00:11:42.500 section and things like that. 385 00:11:42.500 --> 00:11:43.166 And you don't usually 386 00:11:43.166 --> 00:11:45.083 do like section 2.2f. 387 00:11:45.333 --> 00:11:48.083 You have some ID that is consistent even 388 00:11:48.083 --> 00:11:49.875 if you reorder the docs or you reorder 389 00:11:49.875 --> 00:11:50.791 the code or whatever. 390 00:11:51.125 --> 00:11:52.958 And those tools like DOORS or whatever 391 00:11:52.958 --> 00:11:54.208 will, they have essentially like a 392 00:11:54.208 --> 00:11:56.291 numeric allocator where you can assign 393 00:11:56.291 --> 00:11:58.083 these numbers and they only ever go up. 394 00:11:58.083 --> 00:11:59.833 And it's a large enough space that even 395 00:11:59.833 --> 00:12:01.791 over a very long project, it's like eight 396 00:12:01.791 --> 00:12:03.291 or nine digits worth of numbers. 397 00:12:03.291 --> 00:12:04.375 So you're not gonna run out of them. 398 00:12:04.916 --> 00:12:07.375 To be fair, it's really hard to, if you 399 00:12:07.375 --> 00:12:09.333 have tests, you can run them and you can 400 00:12:09.333 --> 00:12:11.625 instrument the code to measure code 401 00:12:11.625 --> 00:12:12.416 coverage essentially. 402 00:12:12.750 --> 00:12:15.208 But measuring lines is garbage. 403 00:12:15.791 --> 00:12:17.666 Measuring branches is not much better 404 00:12:17.791 --> 00:12:19.500 because yeah, you can make sure that in 405 00:12:19.500 --> 00:12:22.125 all tests taken together, each branch is 406 00:12:22.125 --> 00:12:24.166 at least taken once, but I don't think it 407 00:12:24.166 --> 00:12:26.083 tests for like all possible combinations 408 00:12:26.083 --> 00:12:27.208 of all possible branches. 409 00:12:27.791 --> 00:12:29.916 So there may be behavior that's described 410 00:12:29.916 --> 00:12:31.000 in the docs that is not 411 00:12:31.000 --> 00:12:32.041 actually tested at all. 412 00:12:32.041 --> 00:12:32.875 And there's no automated 413 00:12:32.875 --> 00:12:34.125 way to really find that out. 414 00:12:34.583 --> 00:12:36.666 Yeah, so I mean, there's two things in 415 00:12:36.666 --> 00:12:38.666 safety critical, depending on how safety 416 00:12:38.666 --> 00:12:40.000 critical you're talking, where you'll get 417 00:12:40.000 --> 00:12:41.708 into something that's called MC/DC 418 00:12:41.708 --> 00:12:43.791 coverage or modified condition/decision 419 00:12:43.791 --> 00:12:45.291 coverage, where kind of like you're 420 00:12:45.291 --> 00:12:49.083 talking about if A and and B or C, you 421 00:12:49.083 --> 00:12:51.041 don't need to just exercise getting in 422 00:12:51.041 --> 00:12:53.416 that if and out of that if, you have to 423 00:12:53.416 --> 00:12:55.833 exercise all the preconditions and the 424 00:12:55.833 --> 00:12:58.708 MC/DC versus DC means that you could at 425 00:12:58.708 --> 00:13:00.791 least cheat and or not cheat, but you 426 00:13:00.791 --> 00:13:02.291 don't have to prove that the ones that 427 00:13:02.291 --> 00:13:04.208 shortcut because if you have like A and 428 00:13:04.208 --> 00:13:07.916 and B or C, if A is false, you can just 429 00:13:07.916 --> 00:13:09.333 take it for granted that you don't have 430 00:13:09.333 --> 00:13:10.166 to come up with 431 00:13:10.166 --> 00:13:11.708 different test cases for B and C. 432 00:13:12.458 --> 00:13:13.625 So that's the MC/DC versus 433 00:13:13.625 --> 00:13:15.166 just raw decision coverage. 434 00:13:16.125 --> 00:13:18.000 But the other thing is, at least in 435 00:13:18.000 --> 00:13:20.208 avionics, you get into something called 436 00:13:20.208 --> 00:13:22.208 requirements based testing in that you 437 00:13:22.208 --> 00:13:23.666 can't just write tests 438 00:13:23.666 --> 00:13:25.166 to hit code coverage. 439 00:13:25.666 --> 00:13:27.125 Every test case has to be 440 00:13:27.125 --> 00:13:28.958 mapped to some requirement. 441 00:13:29.458 --> 00:13:31.291 And so you end up in this case where 442 00:13:31.291 --> 00:13:32.083 like, you can't just be 443 00:13:32.083 --> 00:13:33.375 like, "Ah, I need to cover this." 444 00:13:33.666 --> 00:13:35.125 It kind of forces you to make sure that 445 00:13:35.125 --> 00:13:37.208 your docs are accurate in that like, you 446 00:13:37.208 --> 00:13:39.041 need to explain why there's an edge case 447 00:13:39.041 --> 00:13:41.541 here and what you do in response to those 448 00:13:41.541 --> 00:13:44.041 edge cases or why you do something 449 00:13:44.041 --> 00:13:45.583 differently when there are more than 64 450 00:13:45.583 --> 00:13:47.541 items where you might just do it linearly 451 00:13:47.541 --> 00:13:49.750 less than that, but you might do binary 452 00:13:49.750 --> 00:13:51.250 search or something for more than that. 453 00:13:51.583 --> 00:13:54.541 So what you're getting at is that you 454 00:13:54.541 --> 00:13:56.416 think we should not write 455 00:13:56.416 --> 00:13:58.333 markdown for documentation. 456 00:13:58.708 --> 00:13:59.875 We should not write MDBook. 457 00:13:59.875 --> 00:14:01.666 We should write a bunch of tests and then 458 00:14:01.666 --> 00:14:03.791 in the doc tests for those tests, that 459 00:14:03.791 --> 00:14:04.750 should be the entire documentation. 460 00:14:05.041 --> 00:14:06.083 No, actually the opposite. 461 00:14:06.333 --> 00:14:07.291 I know that's not what you're actually 462 00:14:07.291 --> 00:14:08.416 saying, but if we did that. 463 00:14:09.083 --> 00:14:10.916 Yeah, I think that's one way of doing it, 464 00:14:10.916 --> 00:14:13.000 but I think what I really wish is that we 465 00:14:13.000 --> 00:14:16.166 had a way, regardless of syntax, like if 466 00:14:16.166 --> 00:14:17.708 I write it in Typst or if I write it in 467 00:14:17.708 --> 00:14:20.041 markdown or restructured text or 468 00:14:20.041 --> 00:14:22.750 whatever, what I want is a cargo tracing 469 00:14:22.750 --> 00:14:26.208 or cargo trace or some tool that wasn't 470 00:14:26.208 --> 00:14:27.583 necessarily tied to any language 471 00:14:27.583 --> 00:14:29.000 ecosystem where you'd be able to put 472 00:14:29.000 --> 00:14:31.000 these annotations in and some way of 473 00:14:31.000 --> 00:14:32.583 understanding between them. 474 00:14:32.583 --> 00:14:35.333 So actually when I advise teams, I tell 475 00:14:35.333 --> 00:14:36.083 them to keep doing 476 00:14:36.083 --> 00:14:37.083 whatever's effective for them. 477 00:14:37.083 --> 00:14:38.416 If they use a wiki for 478 00:14:38.416 --> 00:14:40.041 docs, keep using the wiki. 479 00:14:40.041 --> 00:14:41.458 It has change management, 480 00:14:41.458 --> 00:14:42.375 it has things like that. 481 00:14:42.625 --> 00:14:44.333 If you use markdown files, great. 482 00:14:44.708 --> 00:14:47.208 Use whatever is useful to you, but I 483 00:14:47.208 --> 00:14:49.458 wanna be able to add that linking of 484 00:14:49.458 --> 00:14:51.500 metadata in a way that 485 00:14:51.500 --> 00:14:53.000 you can get that awareness. 486 00:14:53.000 --> 00:14:54.500 Because like I said, safety critical does 487 00:14:54.500 --> 00:14:55.458 this because you have to have what's 488 00:14:55.458 --> 00:14:57.250 called a traceability matrix, which shows 489 00:14:57.250 --> 00:14:59.500 that you've got like 100% traceability 490 00:14:59.500 --> 00:15:00.708 coverage more or less. 491 00:15:01.333 --> 00:15:03.541 But just how useful it is when you go 492 00:15:03.541 --> 00:15:05.541 like: I'm changing this chunk of code, 493 00:15:05.916 --> 00:15:07.416 why is this code like this? 494 00:15:07.791 --> 00:15:11.125 I wish that language servers existed, so 495 00:15:11.125 --> 00:15:12.666 you could like right click on a function 496 00:15:13.208 --> 00:15:15.583 and jump to the docs or jump to the tests 497 00:15:15.583 --> 00:15:17.833 or even like within a function, like why 498 00:15:17.833 --> 00:15:18.750 does this change 499 00:15:18.750 --> 00:15:20.666 behavior above 64 elements? 500 00:15:21.083 --> 00:15:23.041 And you'd be able to jump to that and say 501 00:15:23.041 --> 00:15:25.125 like: ah, it's like this for this. 502 00:15:25.375 --> 00:15:27.125 In the same way that we have just that 503 00:15:27.125 --> 00:15:28.916 ability to jump back and forth, and once 504 00:15:28.916 --> 00:15:31.166 you have that metadata, the ability to 505 00:15:31.166 --> 00:15:33.916 build tools that help you with that is a 506 00:15:33.916 --> 00:15:36.875 stunningly useful piece of kit. 507 00:15:36.875 --> 00:15:39.083 Because you always know like, if I change 508 00:15:39.083 --> 00:15:39.916 something, what are the 509 00:15:39.916 --> 00:15:41.000 ripple effects of this? 510 00:15:41.458 --> 00:15:43.166 And you don't end up changing something 511 00:15:43.166 --> 00:15:45.291 six months ago and only realizing six 512 00:15:45.291 --> 00:15:46.708 months later that it changed, and you go, 513 00:15:46.708 --> 00:15:47.833 "Why are these docs out of date?" 514 00:15:48.500 --> 00:15:51.166 So you mentioned wikis earlier, but 515 00:15:51.166 --> 00:15:53.666 usually those have a separate revision 516 00:15:53.666 --> 00:15:56.916 control system than like-- I would think 517 00:15:56.916 --> 00:15:57.916 that with the tooling that you're 518 00:15:57.916 --> 00:15:59.500 describing, you would need to store the 519 00:15:59.500 --> 00:16:01.541 docs alongside the source code so that 520 00:16:01.541 --> 00:16:03.666 you can travel back in time six months 521 00:16:03.666 --> 00:16:05.458 and see what like things were then. 522 00:16:05.458 --> 00:16:07.708 Otherwise, you only get the picture of 523 00:16:07.708 --> 00:16:10.208 what things are now, you need to refer to 524 00:16:10.208 --> 00:16:11.791 timestamps, which gets complicated with 525 00:16:11.791 --> 00:16:13.166 DBCS and branches and whatnot. 526 00:16:13.583 --> 00:16:15.916 Yeah, I think, yeah, paid tools do that. 527 00:16:15.958 --> 00:16:18.458 They have integrations with Jira for wiki 528 00:16:18.666 --> 00:16:19.791 or the integrations with 529 00:16:19.791 --> 00:16:20.916 other tools and stuff like that. 530 00:16:21.333 --> 00:16:22.458 I guess what I'm saying is in open 531 00:16:22.458 --> 00:16:23.666 source, we actually sort of have it 532 00:16:23.666 --> 00:16:24.833 easier because they 533 00:16:24.833 --> 00:16:26.250 do all travel together. 534 00:16:26.458 --> 00:16:27.875 So I do think that if you were building 535 00:16:27.875 --> 00:16:29.708 something, you'd wanna lean on that 536 00:16:29.708 --> 00:16:31.375 requirement that like: okay, we're just 537 00:16:31.375 --> 00:16:32.583 gonna say that they all travel in the 538 00:16:32.583 --> 00:16:35.166 same repo and so when branches happen, 539 00:16:35.666 --> 00:16:37.458 it's only coherent within like a single 540 00:16:37.458 --> 00:16:39.500 commit or maybe through some 541 00:16:39.500 --> 00:16:41.208 like history of that as well. 542 00:16:41.666 --> 00:16:43.291 But yeah, I think you definitely make 543 00:16:43.291 --> 00:16:45.583 your life way easier when all three of 544 00:16:45.583 --> 00:16:46.708 those travel together. 545 00:16:47.208 --> 00:16:48.666 And yeah, the other thing you can just do 546 00:16:48.666 --> 00:16:51.375 is in the same way that like on Docs.rs, 547 00:16:51.625 --> 00:16:52.625 you can say like, what 548 00:16:52.625 --> 00:16:55.166 percentage of my code is documented? 549 00:16:55.375 --> 00:16:56.833 So you can say like: on all your public 550 00:16:56.833 --> 00:16:58.875 modules or public functions and things 551 00:16:58.875 --> 00:17:02.041 like that, now on Docs.rs, you can poke 552 00:17:02.041 --> 00:17:04.000 the dropdown and it'll say like "70% of 553 00:17:04.000 --> 00:17:06.375 functions have documentation" or whatever. 554 00:17:06.708 --> 00:17:07.208 Wait, is that new? 555 00:17:08.208 --> 00:17:10.375 In the last year or so, I think, but 556 00:17:10.375 --> 00:17:11.208 yeah, it shows you like 557 00:17:11.208 --> 00:17:12.208 documentation coverage. 558 00:17:12.708 --> 00:17:14.416 Oh yeah, wow, I never noticed that. 559 00:17:14.416 --> 00:17:15.458 It's relatively new. 560 00:17:15.458 --> 00:17:16.458 Yeah, it is in the dropdown. 561 00:17:16.458 --> 00:17:16.666 Yeah. 562 00:17:16.916 --> 00:17:17.333 Cool. 563 00:17:17.375 --> 00:17:18.458 But like, what if you 564 00:17:18.458 --> 00:17:19.666 could have the same thing? 565 00:17:19.958 --> 00:17:22.000 What if it was just easy to say like: how 566 00:17:22.000 --> 00:17:24.875 much of my code has docs? Or how much of 567 00:17:24.875 --> 00:17:27.208 my docs match my code? Or 568 00:17:27.208 --> 00:17:29.458 what is my test coverage? 569 00:17:29.458 --> 00:17:30.875 Not necessarily like what is my code 570 00:17:30.875 --> 00:17:33.375 coverage, but like how many places did I 571 00:17:33.375 --> 00:17:35.125 say: hey, I'm testing this function or 572 00:17:35.125 --> 00:17:36.875 I'm testing this module or I'm testing 573 00:17:36.875 --> 00:17:39.041 the whole system or something like that. 574 00:17:39.333 --> 00:17:40.541 And you can see that linkage. 575 00:17:40.833 --> 00:17:42.625 And then kind of to complete the circle 576 00:17:42.625 --> 00:17:45.083 to say like, how many tests are backing 577 00:17:45.083 --> 00:17:46.625 up what I claim in my docs? 578 00:17:46.875 --> 00:17:48.375 How confident can someone be that like, 579 00:17:48.375 --> 00:17:50.208 if I write a big paragraph of how this 580 00:17:50.208 --> 00:17:52.375 works, that didn't change six months ago 581 00:17:52.375 --> 00:17:54.125 when I totally changed the algorithm or 582 00:17:54.125 --> 00:17:55.166 something like that. 583 00:17:55.166 --> 00:17:56.791 And this, like I mentioned, this is what 584 00:17:56.791 --> 00:17:58.041 gets reported and is required 585 00:17:58.041 --> 00:17:59.583 in safety critical projects. 586 00:17:59.583 --> 00:18:00.916 This is the traceability matrix. 587 00:18:01.541 --> 00:18:04.500 And in a really cool way, the Ferrocene 588 00:18:04.500 --> 00:18:05.875 Project publishes their 589 00:18:05.875 --> 00:18:06.833 traceability matrix. 590 00:18:06.833 --> 00:18:08.583 So you can see, hey, here's a version of 591 00:18:08.583 --> 00:18:10.541 the Rust language specification. 592 00:18:11.000 --> 00:18:12.000 And then what is the 593 00:18:12.000 --> 00:18:13.333 traceability matrix of that? 594 00:18:13.375 --> 00:18:14.708 How does that link to code? 595 00:18:14.708 --> 00:18:16.583 And how does that link to validation of 596 00:18:16.583 --> 00:18:17.625 that in the form of 597 00:18:17.625 --> 00:18:18.625 tests and things like that? 598 00:18:18.916 --> 00:18:20.583 And I think you have to pay for the real 599 00:18:20.583 --> 00:18:22.541 one that you can use when you submit to 600 00:18:22.541 --> 00:18:24.750 whatever your governing authority is. 601 00:18:25.000 --> 00:18:26.833 But just as an open source artifact, you 602 00:18:26.833 --> 00:18:28.625 can go and look at it and you can tell 603 00:18:28.625 --> 00:18:30.750 like, "Ah, they said that this unit test 604 00:18:30.750 --> 00:18:33.625 goes to verify this portion of the 605 00:18:33.625 --> 00:18:35.708 language specification," which is like a 606 00:18:35.708 --> 00:18:37.541 really validating and useful thing to 607 00:18:37.541 --> 00:18:39.833 have because you know, otherwise it's 608 00:18:39.833 --> 00:18:42.291 hard to tell when docs get stale because 609 00:18:42.291 --> 00:18:44.041 they aren't tested nearly as thoroughly 610 00:18:44.208 --> 00:18:46.541 as code is and things like that. 611 00:18:46.583 --> 00:18:48.958 Yes, I'm looking at one of the UI tests 612 00:18:49.208 --> 00:18:50.583 from the Ferrocene test suite. 613 00:18:50.583 --> 00:18:52.416 And it has-- they basically just use a 614 00:18:52.416 --> 00:18:53.833 bunch of normal commands. 615 00:18:53.833 --> 00:18:55.333 They're not even code comments. 616 00:18:55.333 --> 00:18:56.958 They're just a Ferrocene annotations, 617 00:18:57.208 --> 00:18:59.583 colon, and then an identifier starting 618 00:18:59.583 --> 00:19:01.708 with FLS_, and then 619 00:19:01.708 --> 00:19:03.458 some random bunch of letters. 620 00:19:03.458 --> 00:19:05.708 And those are all listed in, yeah, in the 621 00:19:05.708 --> 00:19:07.125 traceability matrix report, 622 00:19:07.125 --> 00:19:09.958 which I can see without paying. 623 00:19:10.208 --> 00:19:10.666 I don't know which 624 00:19:10.666 --> 00:19:11.708 part is paid, but yeah. 625 00:19:11.750 --> 00:19:12.875 I think it's one of 626 00:19:12.875 --> 00:19:13.791 those licensing things. 627 00:19:13.791 --> 00:19:16.166 I think it's free to view, but you can't 628 00:19:16.166 --> 00:19:17.416 use it as part of your like, 629 00:19:17.500 --> 00:19:18.583 certification argument 630 00:19:18.583 --> 00:19:19.916 for those kinds of things. 631 00:19:19.916 --> 00:19:21.333 This is all stuff that I started when I 632 00:19:21.333 --> 00:19:23.125 was at Ferrous, but the actual execution 633 00:19:23.125 --> 00:19:25.250 was done entirely after I left. 634 00:19:25.250 --> 00:19:27.041 So I remember what the plans were, and I 635 00:19:27.041 --> 00:19:28.583 always have sort of a blurry time of what 636 00:19:28.583 --> 00:19:30.083 we were originally planning versus the 637 00:19:30.083 --> 00:19:32.041 folks like Jorge and Pietro and things 638 00:19:32.041 --> 00:19:32.625 like that, that actually 639 00:19:32.625 --> 00:19:34.541 executed on it after I left. 640 00:19:35.000 --> 00:19:36.208 But I give them a ton of credit because 641 00:19:36.208 --> 00:19:37.708 they did it in the way that we wanted it 642 00:19:37.708 --> 00:19:38.666 to, which was not 643 00:19:38.666 --> 00:19:40.000 intrusive to the Rust language. 644 00:19:40.750 --> 00:19:43.083 And as something that releases with the 645 00:19:43.083 --> 00:19:44.125 same cadence, using 646 00:19:44.125 --> 00:19:45.333 tooling and things like that. 647 00:19:45.333 --> 00:19:46.541 Because a lot of these like safety 648 00:19:46.541 --> 00:19:48.333 critical areas, they don't release 649 00:19:48.333 --> 00:19:50.000 quarterly, let alone 650 00:19:50.000 --> 00:19:51.375 yearly a lot of the time. 651 00:19:51.625 --> 00:19:52.916 And the fact that they're generally 652 00:19:52.916 --> 00:19:54.500 keeping up with the open source project 653 00:19:54.500 --> 00:19:56.666 while not being a bummer to them, says 654 00:19:56.666 --> 00:19:58.416 that these kinds of tools don't have to 655 00:19:58.416 --> 00:20:01.041 be very intrusive and painful over an 656 00:20:01.041 --> 00:20:02.208 expected like quality 657 00:20:02.208 --> 00:20:03.708 open source workflow. 658 00:20:03.708 --> 00:20:05.208 Like if you're just like shoving commits 659 00:20:05.416 --> 00:20:06.916 and not writing tests or whatever, like 660 00:20:06.916 --> 00:20:08.041 this is not the first 661 00:20:08.041 --> 00:20:09.208 thing that you should be doing. 662 00:20:09.666 --> 00:20:11.041 But if you're like: I have a pretty 663 00:20:11.041 --> 00:20:13.208 stable project and I wanna make sure that 664 00:20:13.208 --> 00:20:15.166 I don't accidentally break things or 665 00:20:15.166 --> 00:20:16.250 people build a lot of 666 00:20:16.250 --> 00:20:17.291 stuff on top of this. 667 00:20:17.291 --> 00:20:18.708 So like right now I'm sitting down and 668 00:20:18.708 --> 00:20:20.541 writing a new revision of a specification 669 00:20:20.958 --> 00:20:23.166 for Postcard the serialization format, 670 00:20:23.666 --> 00:20:25.708 some internal pieces that are used in 671 00:20:25.708 --> 00:20:27.250 different parts, and then 672 00:20:27.250 --> 00:20:28.666 Postcard-RPC the protocol. 673 00:20:29.166 --> 00:20:30.500 And this is really where this itch comes 674 00:20:30.500 --> 00:20:32.416 from, is I wish I had a way to say like: 675 00:20:32.541 --> 00:20:34.291 okay, I'm writing a human text version of 676 00:20:34.291 --> 00:20:35.416 how this encoding format 677 00:20:35.416 --> 00:20:36.708 and this protocol works. 678 00:20:37.125 --> 00:20:38.750 And instead of just like, "Go read the RFC 679 00:20:39.083 --> 00:20:40.416 and maybe there's some code comments that 680 00:20:40.416 --> 00:20:43.125 tell you why this test exists," I wish I 681 00:20:43.125 --> 00:20:45.125 had an ability to say like, "Okay, there 682 00:20:45.125 --> 00:20:47.333 are 2000 lines in this protocol 683 00:20:47.333 --> 00:20:49.416 definition," and every single one of those 684 00:20:49.416 --> 00:20:51.916 lines, you can jump to a chunk of code or 685 00:20:51.916 --> 00:20:53.541 a test or the same way if you're digging 686 00:20:53.541 --> 00:20:54.958 through the code, why is 687 00:20:54.958 --> 00:20:56.333 this hashing like this? 688 00:20:56.625 --> 00:20:58.416 You can jump to the spec and it will 689 00:20:58.416 --> 00:21:00.500 explain exactly why it's like that. 690 00:21:00.500 --> 00:21:01.958 Because that's one of those things that 691 00:21:01.958 --> 00:21:03.583 like-- yeah, I don't know, it's totally 692 00:21:03.583 --> 00:21:04.958 valuable, especially when you're like 693 00:21:04.958 --> 00:21:06.500 refining things and making sure they're 694 00:21:06.500 --> 00:21:07.875 all still correct, that you 695 00:21:07.875 --> 00:21:09.291 didn't leave any loose edges. 696 00:21:09.666 --> 00:21:11.708 So that's where that itch comes from now, 697 00:21:11.708 --> 00:21:14.125 is I'm writing a lot of documentation and 698 00:21:14.125 --> 00:21:15.916 I don't have a better way than just 699 00:21:15.916 --> 00:21:17.208 putting on line numbers on my 700 00:21:17.208 --> 00:21:19.541 documentation and in my code saying, "This 701 00:21:19.541 --> 00:21:24.083 is version 1.1 line 247, this paragraph," 702 00:21:24.416 --> 00:21:25.125 you know what I mean? 703 00:21:25.375 --> 00:21:26.375 And that stuff goes stale 704 00:21:26.375 --> 00:21:28.500 immediately and is painful. 705 00:21:28.916 --> 00:21:30.125 So it requires a lot of care 706 00:21:30.125 --> 00:21:31.750 and tools can help with that. 707 00:21:31.875 --> 00:21:33.750 How is that we don't already have some 708 00:21:33.750 --> 00:21:35.125 open source tooling for that? 709 00:21:35.375 --> 00:21:36.833 We have open source tooling for 710 00:21:36.833 --> 00:21:39.208 everything, even if it's crap, we usually 711 00:21:39.208 --> 00:21:40.541 have some version of it. 712 00:21:40.708 --> 00:21:42.166 Like, yeah, someone came from the 713 00:21:42.166 --> 00:21:43.416 industry and they were like, "Oh, I wish I 714 00:21:43.416 --> 00:21:45.500 could do the same for my..." why, like... are 715 00:21:45.500 --> 00:21:46.083 you really the first 716 00:21:46.083 --> 00:21:47.666 person to ask for it in 2025? 717 00:21:48.291 --> 00:21:49.625 I don't know! I wrote 718 00:21:49.625 --> 00:21:51.166 up... I'd have to go find it. 719 00:21:51.166 --> 00:21:53.500 I wrote up a pitch of if I were to do it, 720 00:21:53.500 --> 00:21:55.708 this is how I would do it in the Rust 721 00:21:55.708 --> 00:21:57.083 ecosystem a couple of years back. 722 00:21:57.083 --> 00:21:58.291 And a couple of people were like: yeah. I 723 00:21:58.291 --> 00:22:00.375 think it's a mix of one, it's seen as 724 00:22:00.375 --> 00:22:02.833 like a vegetable of the process. 725 00:22:03.125 --> 00:22:04.250 Like it's something you do because it's 726 00:22:04.250 --> 00:22:05.958 good for you, not because you enjoy it. 727 00:22:06.208 --> 00:22:06.833 It's like eating your 728 00:22:06.833 --> 00:22:08.041 vegetables is how I should say it. 729 00:22:08.041 --> 00:22:09.000 Yeah, that's a great expression. 730 00:22:09.250 --> 00:22:10.916 And so like a lot of people, especially 731 00:22:10.916 --> 00:22:12.458 people... I don't know, this is true with 732 00:22:12.458 --> 00:22:14.541 any like process where people go with the 733 00:22:14.541 --> 00:22:16.250 motions without understanding the why. 734 00:22:16.541 --> 00:22:18.166 I think there's a lot of people who hate 735 00:22:18.166 --> 00:22:19.875 doing traceability because it's just a 736 00:22:19.875 --> 00:22:21.166 box they check and they 737 00:22:21.166 --> 00:22:22.416 don't get any value from it. 738 00:22:22.416 --> 00:22:24.208 But if you do it well, you 739 00:22:24.208 --> 00:22:25.500 do it because it helps you. 740 00:22:25.500 --> 00:22:27.000 And it actually, I miss not 741 00:22:27.000 --> 00:22:28.458 having the ability to do it. 742 00:22:28.541 --> 00:22:29.833 And I would think if any language 743 00:22:29.833 --> 00:22:31.916 community is receptive to this, it would 744 00:22:31.916 --> 00:22:34.083 be Rust who's already like subjecting 745 00:22:34.083 --> 00:22:35.083 themselves to lifetime 746 00:22:35.083 --> 00:22:36.041 annotations and whatnot. 747 00:22:36.416 --> 00:22:37.500 Yeah, I think it's one of those things 748 00:22:37.500 --> 00:22:39.375 it's just the population of people who 749 00:22:39.375 --> 00:22:41.750 have seen it done and seen it done well, 750 00:22:42.208 --> 00:22:44.083 and are interested in taking the time to 751 00:22:44.083 --> 00:22:46.416 build the annotation and tooling in a way 752 00:22:46.416 --> 00:22:48.458 that is pleasing-- or really not even 753 00:22:48.458 --> 00:22:50.375 necessarily pleasing, but not disruptive 754 00:22:50.375 --> 00:22:52.458 to the development process where it 755 00:22:52.458 --> 00:22:54.625 doesn't feel like eating your vegetables. 756 00:22:55.291 --> 00:22:57.708 That's a very narrow set, which is I 757 00:22:57.708 --> 00:22:58.666 think a lot of that 758 00:22:58.666 --> 00:22:59.958 stuff becomes paid dev tools. 759 00:22:59.958 --> 00:23:01.458 All that stuff where like you either have 760 00:23:01.458 --> 00:23:03.750 to have it or only the people that really 761 00:23:03.750 --> 00:23:05.166 have seen enough shit that 762 00:23:05.166 --> 00:23:06.291 they know that they need that. 763 00:23:06.625 --> 00:23:08.166 A lot of that stuff becomes paid dev 764 00:23:08.166 --> 00:23:08.833 tools because the 765 00:23:08.833 --> 00:23:10.250 upkeep of it takes effort. 766 00:23:10.541 --> 00:23:12.708 And especially if you weren't in Rust, a 767 00:23:12.708 --> 00:23:15.666 community that has docs and tests and 768 00:23:15.666 --> 00:23:18.833 code in the same repo and has a care for 769 00:23:18.833 --> 00:23:21.791 validation and correctness... like you're 770 00:23:21.791 --> 00:23:24.208 saying, I think Rust is the first maybe 771 00:23:24.208 --> 00:23:26.416 large scale community of 772 00:23:26.416 --> 00:23:27.833 people who are likely adopt that. 773 00:23:28.208 --> 00:23:29.625 And I think there's just not enough 774 00:23:29.625 --> 00:23:31.416 overlap with the safety critical domain. 775 00:23:31.833 --> 00:23:33.250 And even in that population, there's a 776 00:23:33.250 --> 00:23:35.708 lot of people who like don't get why you 777 00:23:35.708 --> 00:23:37.166 would want to do it or haven't had a good 778 00:23:37.166 --> 00:23:38.750 experience doing it because they were 779 00:23:38.750 --> 00:23:39.833 just forced to do it. 780 00:23:40.083 --> 00:23:41.250 That I think they're just, the Venn 781 00:23:41.250 --> 00:23:44.291 diagram has a very narrow sliver in it. 782 00:23:44.291 --> 00:23:45.958 Yeah, now that I think about it, even 783 00:23:45.958 --> 00:23:48.250 though the Rust people might be receptive 784 00:23:48.250 --> 00:23:50.708 to it, the others needed so much more, I 785 00:23:50.708 --> 00:23:52.791 think, because they don't have those like 786 00:23:52.791 --> 00:23:53.625 the safety provided 787 00:23:53.625 --> 00:23:54.541 by the language itself. 788 00:23:54.541 --> 00:23:56.000 So they need the whole test of 789 00:23:56.000 --> 00:23:58.666 specification code trial much more. 790 00:23:59.416 --> 00:24:00.791 Because there's only so much, so many 791 00:24:00.791 --> 00:24:02.541 invariants you can put into code. 792 00:24:02.791 --> 00:24:04.083 Like that is a thing that Rust tries to 793 00:24:04.083 --> 00:24:05.541 do is you try and make your invariants 794 00:24:05.541 --> 00:24:07.083 encode so the compiler 795 00:24:07.083 --> 00:24:08.208 can help you enforce them. 796 00:24:08.208 --> 00:24:09.541 But there's some stuff that's just not 797 00:24:09.541 --> 00:24:11.750 reasonable to encode or 798 00:24:11.750 --> 00:24:13.666 not efficient to encode or-- 799 00:24:13.750 --> 00:24:14.166 I mean, you have 800 00:24:14.166 --> 00:24:15.500 unsafe interfaces as well. 801 00:24:15.500 --> 00:24:16.791 You have to deal with the real 802 00:24:16.791 --> 00:24:19.083 physicality of the world at some point. 803 00:24:19.125 --> 00:24:20.708 Or some like generics and type states 804 00:24:20.833 --> 00:24:21.958 sometimes just make the code 805 00:24:21.958 --> 00:24:24.166 so egregious to actually use. 806 00:24:24.541 --> 00:24:25.916 So like the actual enforcing 807 00:24:25.916 --> 00:24:28.375 makes it so unpleasant to use. 808 00:24:28.750 --> 00:24:31.041 It becomes awful, you know what I mean? 809 00:24:31.041 --> 00:24:32.208 So there's a balance for sure. 810 00:24:32.708 --> 00:24:34.333 I keep thinking of my question like, "Why 811 00:24:34.333 --> 00:24:36.208 isn't this already done and great and 812 00:24:36.208 --> 00:24:36.958 available for everyone?" 813 00:24:37.166 --> 00:24:39.708 It's because it's the docs of the docs. 814 00:24:39.708 --> 00:24:42.791 Like there's already-- the docs tooling is 815 00:24:42.791 --> 00:24:44.416 already overlooked 816 00:24:44.416 --> 00:24:45.791 compared to everything else. 817 00:24:45.791 --> 00:24:47.000 Like if you look at the number of 818 00:24:47.000 --> 00:24:49.583 contributors to Rust core projects, I 819 00:24:49.583 --> 00:24:51.041 think it goes first rustc, 820 00:24:51.416 --> 00:24:53.333 then Cargo, then rustdoc. 821 00:24:53.708 --> 00:24:54.833 So if we were to make something like 822 00:24:54.833 --> 00:24:56.708 that, there would be like 0.5 of a person 823 00:24:56.708 --> 00:24:58.208 working on it because-- 824 00:24:58.208 --> 00:24:59.958 It's the vegetables of the vegetables really. 825 00:25:00.291 --> 00:25:00.833 You know what I mean? 826 00:25:01.083 --> 00:25:03.166 Like it's not just making sure that 827 00:25:03.166 --> 00:25:05.000 you're testing and documenting, it's 828 00:25:05.000 --> 00:25:06.333 making sure that you're testing and 829 00:25:06.333 --> 00:25:07.958 documenting correctly. 830 00:25:08.708 --> 00:25:10.625 So I mean, yeah, like I said, it's one of 831 00:25:10.625 --> 00:25:11.708 those things that I think if there was a 832 00:25:11.708 --> 00:25:13.458 good tool for it where people could just 833 00:25:13.458 --> 00:25:15.666 add it as a step and it's something that 834 00:25:15.666 --> 00:25:17.583 you could do incrementally... You do have 835 00:25:17.583 --> 00:25:18.458 to hit some threshold. 836 00:25:18.458 --> 00:25:19.833 Like you said, the value of it comes from 837 00:25:19.833 --> 00:25:21.000 how good your annotations are. 838 00:25:21.416 --> 00:25:23.166 So you do have to get some care to it. 839 00:25:23.375 --> 00:25:25.208 And I think there's some design to be had 840 00:25:25.291 --> 00:25:27.541 of how do you litter a document with all 841 00:25:27.541 --> 00:25:28.416 of these annotations 842 00:25:28.625 --> 00:25:30.458 without making the document worse. 843 00:25:31.250 --> 00:25:32.708 And that's how these tools like DOORS and 844 00:25:32.708 --> 00:25:34.208 stuff, like it's built into the editor 845 00:25:34.500 --> 00:25:35.541 where you have one column with your 846 00:25:35.541 --> 00:25:37.041 requirement numbers and then with your 847 00:25:37.041 --> 00:25:37.750 actual requirements. 848 00:25:37.750 --> 00:25:39.625 So like it's intuitive, but how do you do 849 00:25:39.625 --> 00:25:42.208 that in a given markdown file or in a 850 00:25:42.208 --> 00:25:43.750 given Typst file or 851 00:25:43.750 --> 00:25:45.291 restructured text or whatever? 852 00:25:45.541 --> 00:25:47.041 And I think that's what like Sphinx is 853 00:25:47.041 --> 00:25:49.500 doing, but that's a population that is 854 00:25:49.500 --> 00:25:51.750 motivated to deal with those annotations 855 00:25:51.875 --> 00:25:53.125 because it's an entry 856 00:25:53.125 --> 00:25:54.208 criteria for existing. 857 00:25:54.541 --> 00:25:55.750 I think it could happen though. 858 00:25:55.750 --> 00:25:57.916 Cause the whole memory safety thing was 859 00:25:57.916 --> 00:25:59.791 like a fringe thing. 860 00:26:00.000 --> 00:26:02.750 I think a decade ago, like a bunch of 861 00:26:02.750 --> 00:26:05.708 people did it cause they saw the value 862 00:26:05.708 --> 00:26:07.083 before everyone else. 863 00:26:07.083 --> 00:26:08.708 And now the government says, 864 00:26:08.708 --> 00:26:10.125 "You have to do it that way." 865 00:26:10.125 --> 00:26:11.625 So there's been a hell of 866 00:26:11.625 --> 00:26:13.708 an adoption curve for that. 867 00:26:13.708 --> 00:26:15.416 So I could totally see something else 868 00:26:15.416 --> 00:26:16.708 like that taking off. 869 00:26:16.708 --> 00:26:17.583 We need it. 870 00:26:17.583 --> 00:26:19.333 Cause the documentation situation for 871 00:26:19.333 --> 00:26:20.458 most projects is dire. 872 00:26:20.458 --> 00:26:22.208 And I'm the first guilty party here. 873 00:26:22.416 --> 00:26:23.208 Don't look at my stuff. 874 00:26:23.291 --> 00:26:25.416 I think it just becomes once there's a 875 00:26:25.416 --> 00:26:27.500 good tool and it's desirable. 876 00:26:27.833 --> 00:26:28.708 Well, first people need 877 00:26:28.708 --> 00:26:29.875 to know to ask for it. 878 00:26:29.875 --> 00:26:31.291 So maybe that's what this is, is me 879 00:26:31.291 --> 00:26:32.416 telling people that like: 880 00:26:32.416 --> 00:26:33.166 Thank you James. 881 00:26:33.458 --> 00:26:35.541 Hey, there's established art on how to 882 00:26:35.541 --> 00:26:37.000 make sure your docs and your code and 883 00:26:37.000 --> 00:26:38.625 your tests are all coherent with each 884 00:26:38.625 --> 00:26:40.916 other and you can use tools to help you 885 00:26:40.916 --> 00:26:41.583 make sure it's right 886 00:26:41.583 --> 00:26:42.750 instead of just 'get good.' 887 00:26:43.000 --> 00:26:44.833 So like that's sort of step one of 888 00:26:44.833 --> 00:26:46.708 knowing that there is like established 889 00:26:46.708 --> 00:26:48.833 art and we could go and look at those 890 00:26:48.833 --> 00:26:51.083 areas for ideas, but also because we're 891 00:26:51.083 --> 00:26:53.125 not tied to safety critical tooling, you 892 00:26:53.125 --> 00:26:55.708 can really focus on the UX or like, hey, 893 00:26:55.708 --> 00:26:57.083 we only support things where your docs 894 00:26:57.083 --> 00:26:58.458 and your code and your tests are in 895 00:26:58.458 --> 00:27:01.208 GitHub repos and you're willing to 896 00:27:01.208 --> 00:27:03.208 install a cargo extension or some binary 897 00:27:03.208 --> 00:27:04.666 tool that will check it for you. 898 00:27:04.666 --> 00:27:06.666 And you probably already have CI on your 899 00:27:06.666 --> 00:27:09.416 repo so you can run trace check and make 900 00:27:09.416 --> 00:27:11.250 sure that it's still good and you know, 901 00:27:11.541 --> 00:27:12.708 make sure you at least never move 902 00:27:12.708 --> 00:27:14.083 backwards and stuff like that. 903 00:27:14.083 --> 00:27:15.583 So this is my call to action. 904 00:27:15.583 --> 00:27:16.416 If that sounds appealing 905 00:27:16.416 --> 00:27:17.875 to anyone, let me know. 906 00:27:17.875 --> 00:27:18.666 Cause I have some ideas 907 00:27:18.666 --> 00:27:19.625 of how I'd want to do it. 908 00:27:19.625 --> 00:27:20.541 I might end up building it 909 00:27:20.541 --> 00:27:22.083 for some of my Postcard stuff. 910 00:27:22.375 --> 00:27:22.750 Please do. 911 00:27:22.791 --> 00:27:24.416 Just because I want it to be right. 912 00:27:24.458 --> 00:27:25.291 I have one note. 913 00:27:25.791 --> 00:27:28.375 I know it's called traceability in... for 914 00:27:28.375 --> 00:27:30.458 real, but don't call it that. Because the 915 00:27:30.458 --> 00:27:32.333 word is already so overloaded. 916 00:27:32.333 --> 00:27:32.791 It's true. 917 00:27:32.791 --> 00:27:33.166 All right. 918 00:27:33.166 --> 00:27:34.458 I need some pitches for names. 919 00:27:34.583 --> 00:27:35.458 Because people think 920 00:27:35.458 --> 00:27:37.125 tracing is logging when it's not. 921 00:27:37.333 --> 00:27:38.541 So people are going to think traceability 922 00:27:38.541 --> 00:27:39.958 is tracing, even though it's not. 923 00:27:40.333 --> 00:27:42.208 So you need to make up a new word. 924 00:27:42.208 --> 00:27:42.708 I don't know. 925 00:27:42.750 --> 00:27:42.958 Yeah. 926 00:27:42.958 --> 00:27:43.250 All right. 927 00:27:43.458 --> 00:27:44.625 I guess then email me 928 00:27:44.625 --> 00:27:45.916 if you have a-- 929 00:27:46.291 --> 00:27:48.333 Go to sdr-podcast.com/ 930 00:27:48.333 --> 00:27:50.000 ...contact? Is contact-- 931 00:27:50.000 --> 00:27:51.041 We always say the episodes page. 932 00:27:51.250 --> 00:27:53.000 I don't know sdr-podcast.com. 933 00:27:53.000 --> 00:27:54.291 There's a contact button on there. 934 00:27:54.291 --> 00:27:54.958 Send us an email. 935 00:27:54.958 --> 00:27:56.125 We have an about page...?! 936 00:27:56.125 --> 00:27:57.125 Maybe it's on the about page. 937 00:27:57.416 --> 00:27:57.625 Yeah. 938 00:27:57.750 --> 00:28:00.625 alt contact goes to the clip from the 939 00:28:00.625 --> 00:28:01.750 guy with the red hair. 940 00:28:02.458 --> 00:28:04.583 My brain is completely drawing a blank. 941 00:28:04.625 --> 00:28:05.166 Carrot Top? 942 00:28:05.333 --> 00:28:05.750 Rick roll. 943 00:28:06.125 --> 00:28:06.541 Rick roll. 944 00:28:06.875 --> 00:28:07.500 Oh, okay. 945 00:28:07.500 --> 00:28:08.083 Rick Astley. 946 00:28:08.333 --> 00:28:09.041 Rick Astley. 947 00:28:09.500 --> 00:28:11.000 Please cut all of the-- please cut the 948 00:28:11.000 --> 00:28:11.708 entire-- please cut me 949 00:28:11.708 --> 00:28:12.916 entirely out of this episode. 950 00:28:13.208 --> 00:28:15.125 I'm good Lord. 951 00:28:15.666 --> 00:28:15.875 Okay. 952 00:28:16.125 --> 00:28:17.041 So yeah, let me know. 953 00:28:17.291 --> 00:28:18.458 Cause I want it to be a thing. 954 00:28:18.458 --> 00:28:19.958 If I write it, I at least love for some 955 00:28:19.958 --> 00:28:21.041 other people to help or 956 00:28:21.041 --> 00:28:22.208 some other people to use it. 957 00:28:22.458 --> 00:28:24.000 Cause otherwise people are just going to 958 00:28:24.000 --> 00:28:25.958 look at my repos weird, but hit me up! 959 00:28:25.958 --> 00:28:26.750 I'm excited for it. 960 00:28:26.750 --> 00:28:27.666 And I wish this was a thing. 961 00:28:28.000 --> 00:28:28.708 I would try it. 962 00:28:28.708 --> 00:28:29.333 Hell yeah. 963 00:28:29.333 --> 00:28:30.041 I think it's interesting. 964 00:28:39.416 --> 00:28:40.916 This episode is sponsored by Depot: 965 00:28:40.916 --> 00:28:42.208 the build acceleration platform 966 00:28:42.208 --> 00:28:43.083 that's on a mission to 967 00:28:43.083 --> 00:28:44.666 make all builds near instant. 968 00:28:44.916 --> 00:28:46.125 If you're tired of watching your builds 969 00:28:46.125 --> 00:28:47.041 and GitHub actions crawl 970 00:28:47.041 --> 00:28:47.750 like the modern day 971 00:28:47.750 --> 00:28:48.916 equivalent of paint drying, 972 00:28:49.250 --> 00:28:50.125 give Depot's GitHub 973 00:28:50.125 --> 00:28:51.291 Actions runners a try. 974 00:28:51.666 --> 00:28:52.833 They're up to 10X faster 975 00:28:52.833 --> 00:28:54.166 with unlimited concurrency, 976 00:28:54.166 --> 00:28:55.583 faster caching, support for 977 00:28:55.583 --> 00:28:56.958 Linux, macOS, and Windows, 978 00:28:56.958 --> 00:28:57.833 and they plug right into 979 00:28:57.833 --> 00:28:58.750 other Depot optimizations 980 00:28:59.041 --> 00:29:00.625 like accelerated container image builds 981 00:29:00.625 --> 00:29:02.041 and remote caching for Bazel, 982 00:29:02.333 --> 00:29:04.125 Turborepo, Gradle, and more. 983 00:29:04.500 --> 00:29:05.458 Depot was built by developers 984 00:29:05.458 --> 00:29:06.625 who were tired of wasting time 985 00:29:06.625 --> 00:29:07.916 waiting on builds instead of shipping. 986 00:29:08.208 --> 00:29:09.041 It's made for teams 987 00:29:09.041 --> 00:29:09.958 that wanna move faster 988 00:29:09.958 --> 00:29:10.958 and stay focused on 989 00:29:10.958 --> 00:29:12.000 what actually matters. 990 00:29:12.291 --> 00:29:13.625 That's why companies like PostHog 991 00:29:13.625 --> 00:29:14.625 use Depot to cut build 992 00:29:14.625 --> 00:29:15.958 times from over three hours 993 00:29:15.958 --> 00:29:17.166 to just three minutes, 994 00:29:17.166 --> 00:29:18.125 saving tens of thousands 995 00:29:18.125 --> 00:29:19.375 of build hours every week. 996 00:29:19.666 --> 00:29:20.625 Start your free 7 997 00:29:20.625 --> 00:29:21.750 day trial at depot.dev 998 00:29:21.750 --> 00:29:22.833 and let them know we sent you.