WEBVTT 00:00.000 --> 00:10.000 Hello, we can start. Hello, hello. 00:10.000 --> 00:13.000 Fantastic. Well, thank you so much for coming. 00:13.000 --> 00:15.000 Look at all of you wonderful people. 00:15.000 --> 00:18.000 Actually, let me put my glasses on so I can see you. 00:18.000 --> 00:21.000 Ah, you look even better, close up. That's good. 00:21.000 --> 00:25.000 So I'm going to talk a little bit about office documents. 00:25.000 --> 00:31.000 One of the deep joys of the world of the moment is the so much excitement about digital sovereignty 00:31.000 --> 00:37.000 and moving office sweets into a sovereign infrastructure. 00:37.000 --> 00:41.000 And that's great. But then people think, well, office is easy. 00:41.000 --> 00:45.000 Let me write my own. There's all this money out there for this. 00:45.000 --> 00:49.000 Let me try and do that. The browser can do it for free for me. 00:49.000 --> 00:53.000 So it's easy. There's almost an office suite in a web browser. 00:53.000 --> 00:55.000 So let's just use that. And that would be great. 00:55.000 --> 01:01.000 And so I'm just going to walk you through a few of the unanticipated, well, the ostrich is unanticipated. 01:01.000 --> 01:05.000 And it unanticipated difficulties that you might experience when you try this. 01:05.000 --> 01:11.000 So obviously, very easy things. People know about proportional text and monospace text. 01:11.000 --> 01:16.000 And that's all great. Unicode solves everything until you hit the web dings and the wing dings. 01:16.000 --> 01:18.000 And all that sort of thing. But it's easy to solve these problems. 01:18.000 --> 01:24.000 You know, we need metric compatible fonts because the size of the glyphs is really important 01:24.000 --> 01:26.000 because that determines the layout of your document. 01:26.000 --> 01:30.000 But either way, it's easy to fix because we can embed these things. 01:30.000 --> 01:34.000 And actually PDF does this. So PDF subsets all your fonts and sticks little bits of the menu PDF. 01:34.000 --> 01:38.000 And so this is all very simple. Of course, the world is actually a little bit more complicated. 01:38.000 --> 01:42.000 There are people there that are right to left. And that's obviously not very difficult. 01:42.000 --> 01:44.000 You know, in a fine transform, we'll flip your screen. 01:44.000 --> 01:47.000 You could just get some glasses that go the other way around. 01:47.000 --> 01:52.000 Except it's not nearly as easy as that because there's mixed left to right and right to left text. 01:52.000 --> 01:56.000 And the whole thing becomes rapidly quite problematic. 01:56.000 --> 02:00.000 And then, of course, there's laying things out. So layout seems very simple. 02:00.000 --> 02:04.000 I mean, you know, when you write a document by hand, it's all just fine, isn't it? 02:04.000 --> 02:06.000 That's no problem at all. 02:06.000 --> 02:10.000 But the problem is there's all sorts of complexities with text that you don't realize. 02:10.000 --> 02:12.000 And as you start to lay it out, you discover, 02:12.000 --> 02:14.000 well, it's a bit of a nightmare, isn't it? 02:14.000 --> 02:16.000 This is a very simple layout problem. 02:16.000 --> 02:21.000 So, you know, we've got, okay, so on the right hand side is a great example of auto-hyphenation, 02:21.000 --> 02:23.000 not working in Google Docs. 02:23.000 --> 02:26.000 And what hyphenation turns out to be difficult. 02:26.000 --> 02:28.000 And that's great. 02:28.000 --> 02:32.000 But unlike, or a little bit like writing your own programming language, 02:32.000 --> 02:34.000 I don't know, lots of people write their own programming language. 02:34.000 --> 02:36.000 It's very trendy to write your own programming languages. 02:36.000 --> 02:38.000 You know, yeah, look at that. You know, excellent. 02:38.000 --> 02:40.000 Some of us do. 02:40.000 --> 02:44.000 I think there's a tourism of writing new programming languages, which is this. 02:44.000 --> 02:50.000 By the time the debugger works, the language is obsolete, right? 02:50.000 --> 02:53.000 And this is a fairly true of everything. 02:53.000 --> 02:57.000 I mean, if you try and debug Java embedded in C embedded in C++, 02:57.000 --> 03:01.000 you know, it's still, I just setting a breakpoint is, is anyway. 03:01.000 --> 03:05.000 So, yeah, I mean, like, it seems really easy, doesn't it? 03:05.000 --> 03:09.000 But actually, to do this right, you need, like, dictionary-based stemming, you know, 03:09.000 --> 03:11.000 it's quite clever. 03:11.000 --> 03:13.000 And that's before you get into smart justification, 03:13.000 --> 03:15.000 where there's combination of, you know, 03:15.000 --> 03:19.000 kerning and squeezing characters and shrinking spaces, and, you know, 03:19.000 --> 03:22.000 an anti-phenation to try and make the text look as beautiful as possible 03:22.000 --> 03:24.000 without huge jagged holes in it. 03:24.000 --> 03:26.000 And let's know, NMF, who is sitting at the back here, 03:26.000 --> 03:27.000 and waving at me. 03:27.000 --> 03:32.000 No, yeah, is a hero of creating Hun spell and doing awesome things for this. 03:32.000 --> 03:34.000 But forget even that difficulty, 03:34.000 --> 03:37.000 because this is the rocket science, just a simple stuff. 03:37.000 --> 03:40.000 You know, we're laying stuff out on the page, 03:40.000 --> 03:43.000 and we're putting things in, and it's all going super well. 03:43.000 --> 03:46.000 You know, we're playing Tetris, and it's just packing beautifully. 03:46.000 --> 03:48.000 You know, we're going to hit the high score. 03:48.000 --> 03:50.000 And then we think, well, we can get a bit more in the bottom, 03:50.000 --> 03:53.000 or we get that in, and it brings with it a footnote. 03:53.000 --> 03:54.000 Right? 03:54.000 --> 03:56.000 But the footnote then changes the page size, 03:56.000 --> 03:57.000 and now we have to shrink the page size. 03:57.000 --> 03:59.000 Go back and see if we can fit it all in again. 03:59.000 --> 04:03.000 And maybe we can, but that's actually changed the number of pages 04:03.000 --> 04:05.000 in the document, and there's a page number field, 04:05.000 --> 04:07.000 and that then forces the reflow, and now we can't fit it in, 04:07.000 --> 04:11.000 or there's an image in it that's anchored to us anyway. 04:11.000 --> 04:15.000 It all gets quite fun, quite fun. 04:15.000 --> 04:19.000 And that's before we get to actually complicated text. 04:19.000 --> 04:22.000 So I don't know, you know, who likes Ruby, you know? 04:22.000 --> 04:25.000 And of course, Ruby is a programming language, I believe. 04:25.000 --> 04:30.000 But it's also this wonderful way of somantically annotating Tokyo. 04:30.000 --> 04:31.000 This is Tokyo, I believe. 04:31.000 --> 04:34.000 I'm not very good at reading, but this is the English one there, 04:34.000 --> 04:36.000 and I can read that one, right? 04:36.000 --> 04:38.000 And actually there's whole bit of differences there, 04:38.000 --> 04:41.000 and the great news is, and then of course vertical text. 04:41.000 --> 04:45.000 So, you know, we use two web pages that you scroll downwards, 04:45.000 --> 04:47.000 you know, that you keep going down and more and more text comes, 04:47.000 --> 04:51.000 but some people scroll sideways, because their text goes, 04:51.000 --> 04:53.000 goes like this, right? 04:53.000 --> 04:55.000 And which is all fun, right? 04:55.000 --> 04:58.000 It just adds to the rich variety of life, 04:58.000 --> 05:02.000 and the good news is that if you use an HTML div, 05:02.000 --> 05:06.000 and you actually do all of the extra coding and UI stuff 05:06.000 --> 05:09.000 to make this work, then you can actually do this in the browser, 05:09.000 --> 05:10.000 and that's great. 05:10.000 --> 05:13.000 And the browser will lay it out for you, and that's beautiful. 05:13.000 --> 05:16.000 The problem is that if you want to actually get whizzy, 05:16.000 --> 05:19.000 we're getting, you can't leave the browser to its own devices. 05:19.000 --> 05:21.000 You have to switch to using the canvas, 05:21.000 --> 05:24.000 and then you have all this great control of where glyphs are, 05:24.000 --> 05:26.000 and none of this is there for you. 05:26.000 --> 05:28.000 So, that's great. 05:28.000 --> 05:30.000 You rapidly hit this wall. 05:30.000 --> 05:32.000 And then, of course, you want to print, 05:32.000 --> 05:34.000 and that's easy, because as we all know, 05:34.000 --> 05:36.000 browsers can print. 05:36.000 --> 05:38.000 And that's great, isn't it? 05:38.000 --> 05:40.000 Anyone who's printed a Google Maps or whatever, 05:40.000 --> 05:42.000 and a notice, why is it crammed in the corner? 05:42.000 --> 05:45.000 We'll know a little bit about paper sizes, 05:45.000 --> 05:47.000 and the complete lack of API there in the browser. 05:47.000 --> 05:50.000 You can't even work out what page size your printer has 05:50.000 --> 05:52.000 from JavaScript. 05:52.000 --> 05:54.000 You just have no idea. 05:55.000 --> 05:57.000 So, it's a great thing we use the canvas to render, 05:57.000 --> 05:59.000 because that's not going to print at all, really. 05:59.000 --> 06:01.000 And we can then handle all that complex text, 06:01.000 --> 06:05.000 and then generate PDFs and render it twice to make your print. 06:05.000 --> 06:06.000 So, it, yes. 06:06.000 --> 06:08.000 And of course, the browser is very frightened of, 06:08.000 --> 06:10.000 yes, fingerprinting your devices, 06:10.000 --> 06:12.000 so it loves to turn off things that 06:12.000 --> 06:14.000 can tell you about your system. 06:14.000 --> 06:17.000 So, anyway, we've done a bit of text, and that's all good, 06:17.000 --> 06:19.000 but let's think about spreadsheets. 06:19.000 --> 06:21.000 So, spreadsheets are, you know, 06:21.000 --> 06:25.000 the source of never-ending fun, and a personal passion of mine. 06:25.000 --> 06:28.000 And you see, if I can read this, what does it say? 06:28.000 --> 06:29.000 Yeah. 06:29.000 --> 06:32.000 So, we have more than two to the 32 possible cells 06:32.000 --> 06:34.000 in a single sheet in a spreadsheet, 06:34.000 --> 06:36.000 which is quite a lot. 06:36.000 --> 06:39.000 Now, of course, almost no sheets actually have that much data, 06:39.000 --> 06:41.000 because it will probably kill your machine, 06:41.000 --> 06:44.000 unless you have a very efficient way of storing the data. 06:44.000 --> 06:46.000 And so, we're doing pretty well. 06:46.000 --> 06:49.000 We used to have this quite mad way of storing cells, 06:49.000 --> 06:52.000 because it's a very sparse data structure, 06:52.000 --> 06:54.000 which was essentially a row number, 06:54.000 --> 06:56.000 and then a pointer to what was in the cell. 06:56.000 --> 06:58.000 So, we just have these great swathlets of numbers, 06:58.000 --> 07:01.000 which would typically go 1, 2, 3, 4, 5, 07:01.000 --> 07:03.000 with a whole lot of pointers next to them. 07:03.000 --> 07:05.000 Well, sometimes they wouldn't go 7. 07:05.000 --> 07:07.000 They would go 8, you know, and it's exciting. 07:07.000 --> 07:10.000 And it turns out that there's a thing called interpolation search. 07:10.000 --> 07:11.000 I don't know if you've ever heard of it. 07:11.000 --> 07:14.000 It's probably not a very popular computer science search form, 07:14.000 --> 07:17.000 but it has a computational complexity of log log, 07:17.000 --> 07:20.000 the number of things, which is the kind of complex to you want in life. 07:20.000 --> 07:24.000 If you can get a log log in, that's really good. 07:24.000 --> 07:27.000 We switched away from that, because of obvious problems with this method, 07:27.000 --> 07:29.000 particularly pointer chasing is a nightmare, 07:29.000 --> 07:31.000 and modern CPUs hate pointer chasing. 07:31.000 --> 07:35.000 I mean, just caches, yeah, it just doesn't work basically. 07:35.000 --> 07:37.000 And this is partly why Java is quite as slow as it is, 07:37.000 --> 07:40.000 because everything's a pointer, you know? 07:40.000 --> 07:41.000 But there you go. 07:41.000 --> 07:43.000 So, we then did a whole lot of work to inline, 07:44.000 --> 07:47.000 for example, great chunks of double, so you can use SSE, 07:47.000 --> 07:49.000 accelerated arithmetic and so on on things. 07:49.000 --> 07:50.000 It's just pretty cool. 07:50.000 --> 07:53.000 And so we have now a much more compact and better structure, 07:53.000 --> 07:54.000 and it's brilliant. 07:54.000 --> 07:57.000 And you can thread it, and you can do SIMD, and it's great. 07:57.000 --> 08:00.000 Or, you could use JavaScript. 08:00.000 --> 08:03.000 And the implementation's here, you sparse arrays. 08:03.000 --> 08:07.000 You know, it's just basically a hash table on the array index, 08:07.000 --> 08:09.000 often down a row or a column. 08:09.000 --> 08:11.000 And that's great, obviously. 08:12.000 --> 08:15.000 And in theory, constant time look up, but in reality, 08:15.000 --> 08:18.000 another massive pointer chasing disaster areas. 08:18.000 --> 08:20.000 And you may think that this is not relevant, 08:20.000 --> 08:23.000 but the component that first gave us the, 08:23.000 --> 08:26.000 your zips are not 64 bits safe. 08:26.000 --> 08:29.000 And we have streams inside these zips files that decompress 08:29.000 --> 08:33.000 to more than four gigabytes of data, as we pass it was calc. 08:33.000 --> 08:36.000 And this just does happen quite a lot. 08:36.000 --> 08:38.000 And so talking of computation orders, 08:38.000 --> 08:40.000 we have a fixed cost XML parser, 08:40.000 --> 08:44.000 that will pass any size file in constant time. 08:44.000 --> 08:48.000 And we want to heckle me? 08:48.000 --> 08:51.000 Well, it's sort of true. 08:51.000 --> 08:53.000 Basically, we thread it. 08:53.000 --> 08:55.000 So there's a fixed cost of starting up the thread 08:55.000 --> 08:57.000 and getting the parsing and the unzipping in one. 08:57.000 --> 09:00.000 And as long as you consume and set up your structure, 09:00.000 --> 09:02.000 slower than the parsic and pars, 09:02.000 --> 09:04.000 you effectively get your parsing for free, 09:04.000 --> 09:07.000 which is kind of nice when you have gigabytes of data. 09:07.000 --> 09:08.000 Of course, in the JavaScript, 09:08.000 --> 09:10.000 well, this is difficult to do to put it mildly, 09:10.000 --> 09:14.000 because the threading model is not really as helpful, 09:14.000 --> 09:16.000 as it could be. 09:16.000 --> 09:18.000 So either way, we've made a beautiful, 09:18.000 --> 09:20.000 entirely new office suite, 09:20.000 --> 09:22.000 and we've solved the number of these problems 09:22.000 --> 09:23.000 and made them beautiful. 09:23.000 --> 09:25.000 And we're making beautiful looking documents. 09:25.000 --> 09:26.000 They really look great, 09:26.000 --> 09:29.000 but the problem is that there are trillions 09:29.000 --> 09:31.000 of existing documents out there. 09:31.000 --> 09:34.000 And the file formats are of mixed quality. 09:34.000 --> 09:37.000 Let's say, some of them are nicer than others. 09:37.000 --> 09:42.000 And yes, some of them have been wiped on before you get them. 09:42.000 --> 09:45.000 And the problem with that is that you end up with things like this. 09:45.000 --> 09:48.000 So in LibroFish and Calabro, 09:48.000 --> 09:52.000 we have this wonderful tool that will render with Microsoft Office 09:52.000 --> 09:55.000 and render with CalabroFish compared to you. 09:55.000 --> 09:58.000 And where it's read, it was different. 09:58.000 --> 10:00.000 And that's not good. 10:00.000 --> 10:01.000 And you can see, basically, 10:01.000 --> 10:03.000 maybe it looks like you've just got double vision or something, 10:03.000 --> 10:06.000 and you have another drink, and it will look great. 10:06.000 --> 10:09.000 But we can't afford that for all of our users. 10:09.000 --> 10:11.000 You know, it's difficult. 10:11.000 --> 10:13.000 So you need to fix it, and so it wants the problem. 10:13.000 --> 10:14.000 While you look into it, 10:14.000 --> 10:18.000 and you discover that actually the fix here is it's a compatibility level. 10:18.000 --> 10:20.000 There are lots and lots of different ways 10:20.000 --> 10:24.000 to render the same trillion documents, right? 10:24.000 --> 10:26.000 And some of them are the old way, 10:26.000 --> 10:28.000 and some of the new way. 10:28.000 --> 10:31.000 And the number of ways is something like this. 10:31.000 --> 10:32.000 Well, actually there's more than that. 10:32.000 --> 10:34.000 It goes further back in time. 10:34.000 --> 10:38.000 And in the infinite healthfulness of the standardization process, 10:38.000 --> 10:39.000 you get a number that says, 10:39.000 --> 10:42.000 do it, well, actually it doesn't say do it 13, 10:42.000 --> 10:46.000 because there's some superstition in the Microsoft Office team, 10:46.000 --> 10:47.000 amazingly. 10:47.000 --> 10:50.000 But yeah, so you get a number, 10:50.000 --> 10:53.000 and then not a whole lot of hints on how to do it differently. 10:53.000 --> 10:55.000 So you think you're implementing one thing, 10:55.000 --> 10:58.000 and actually you get to implement really a lot. 10:58.000 --> 11:00.000 And that's because everybody is trying to improve their layout 11:00.000 --> 11:03.000 and make it better and nicer and more beautiful, 11:03.000 --> 11:07.000 but it's difficult to describe exactly what that means. 11:07.000 --> 11:09.000 So layout is fundamentally not standardized. 11:09.000 --> 11:11.000 It's not there in the specs. 11:11.000 --> 11:13.000 You're not going to find any of it. 11:13.000 --> 11:15.000 But actually it makes a really big difference. 11:15.000 --> 11:17.000 Well, it can make a really big difference. 11:17.000 --> 11:19.000 If you really care about the number of pages, 11:19.000 --> 11:21.000 and it's got to be exactly right. 11:21.000 --> 11:23.000 You know, if your MI6 or something, 11:23.000 --> 11:27.000 you know, James Bond has to kill the guy on page seven. 11:27.000 --> 11:29.000 You know, it's nice if that's not your contact, 11:29.000 --> 11:32.000 due to some re-raping problem earlier in the document. 11:32.000 --> 11:36.000 So then the other thing is little graphics. 11:36.000 --> 11:39.000 So I mean, we have loads of nice little graphics in documents. 11:39.000 --> 11:41.000 And you'll be pleased to know that the native vector format 11:41.000 --> 11:44.000 for Microsoft Office is this wonderful thing called, 11:44.000 --> 11:49.000 well, EMF, WMF, Metaphiles, EMF plus. 11:49.000 --> 11:53.000 And what you see in these variants is that the rendering API 11:53.000 --> 11:55.000 has changed in Windows over time. 11:55.000 --> 11:58.000 And so all we have, ultimately, is an API dump 11:58.000 --> 12:01.000 of the low-level Windows rendering API. 12:01.000 --> 12:04.000 It's pretty much at the level of individual methods being called. 12:04.000 --> 12:07.000 And so the great thing about that is if you have some random 12:07.000 --> 12:11.000 component that's written by some crazy in your early object, 12:11.000 --> 12:14.000 you can just copy what it's drawing on the screen, 12:14.000 --> 12:17.000 and then someone else can do that again without that software. 12:17.000 --> 12:21.000 Right? So it's a typical metaphor. 12:21.000 --> 12:25.000 And the great thing is that WMF is the original Windows 16 API. 12:25.000 --> 12:27.000 You know, it's just quite amazing. 12:27.000 --> 12:30.000 Exclusive all renderings, you know, one bit masks everywhere. 12:30.000 --> 12:33.000 No anti-aliasing, et cetera. 12:33.000 --> 12:36.000 Obviously that didn't survive, well, it survived a very long time. 12:36.000 --> 12:38.000 And then they wanted to extend it, 12:38.000 --> 12:40.000 luckily it had a comment feature in it. 12:40.000 --> 12:43.000 So inside your 16 bit, you know, or your terrible rendering of the past, 12:43.000 --> 12:46.000 you can have a comment. And inside the comment, 12:46.000 --> 12:48.000 is the EMF file. 12:48.000 --> 12:49.000 Right? So that's great. 12:49.000 --> 12:53.000 So there's then a duplicate rendering of a different API level. 12:53.000 --> 12:56.000 And then, you know, EMF plus, which has a whole, whole, whole, whole. 12:56.000 --> 12:58.000 It's just quite amazing. 12:58.000 --> 13:01.000 And early objects inside these have a preview, 13:01.000 --> 13:03.000 so you can rapidly render them. 13:03.000 --> 13:05.000 So you can see stuff, but actually just common objects. 13:05.000 --> 13:07.000 This, this, this thing here on the right. 13:07.000 --> 13:09.000 These two things, this chart. 13:09.000 --> 13:11.000 My chart doesn't render. 13:11.000 --> 13:14.000 Why doesn't my chart render? You suck at charting. 13:14.000 --> 13:15.000 Well, I just not a chart. 13:15.000 --> 13:20.000 That's a massive legacy stream of vector graphic rendering, 13:20.000 --> 13:22.000 madness that is being rendered properly, 13:22.000 --> 13:24.000 and actually you just see a nice chart. 13:24.000 --> 13:26.000 And the user doesn't have any understanding. 13:26.000 --> 13:30.000 You have a life blissfully unconcerned about the horror that lies beneath, 13:30.000 --> 13:33.000 that we have handled for them over so many years, 13:33.000 --> 13:36.000 with so many people. So yes, thank you. 13:36.000 --> 13:39.000 And of course, it's been extended over time this metaphyl. 13:39.000 --> 13:42.000 So we have this, these nice little, little things like this, 13:42.000 --> 13:45.000 you know, past through arbitrary data to the printer driver. 13:45.000 --> 13:47.000 You know, so this is also used for printer drivers, 13:47.000 --> 13:51.000 the back end, and so you can get post script inside your EMF, 13:51.000 --> 13:54.000 anyway, just to give you sort of a flavor of the scale of the problem 13:54.000 --> 13:56.000 that people love. 13:56.000 --> 13:58.000 I talked about, I keep this being much loved. 13:58.000 --> 14:00.000 It's not just the Windows metaphyl that's, you know, 14:00.000 --> 14:03.000 the legacy of the 16 bit error. 14:03.000 --> 14:06.000 You may wonder why when you look at the spreadsheet, 14:06.000 --> 14:08.000 the document, the main document module, 14:08.000 --> 14:11.000 is actually 10 files instead of 1. 14:11.000 --> 14:14.000 It's the kind of thing that you think, well, that's a bit odd, isn't it? 14:14.000 --> 14:19.000 And the answer is, the OS2, the brand new 32 bit operating system of, 14:19.000 --> 14:21.000 well, you know, 30 years ago, 14:22.000 --> 14:25.000 actually had a 16 bit token limit on the number of tokens, 14:25.000 --> 14:27.000 it could pass in its C++ compiler. 14:27.000 --> 14:30.000 So it was necessary at that time to split the modules, 14:30.000 --> 14:33.000 each into less than that many tokens, 14:33.000 --> 14:35.000 and 30 years later, it's still like that. 14:35.000 --> 14:37.000 I mean, actually that's quite good for revision controls, 14:37.000 --> 14:39.000 so you can see what happened. 14:39.000 --> 14:41.000 Now, you may think that these are extreme examples, 14:41.000 --> 14:42.000 but we could go on all day, 14:42.000 --> 14:44.000 luckily I'm not going to, I don't have the time, 14:44.000 --> 14:45.000 but let me tell you about smartart. 14:45.000 --> 14:47.000 This is absolutely beautiful feature, 14:47.000 --> 14:49.000 where you can generate from these bullets, 14:49.000 --> 14:52.000 schema-driven, amazing graphical loveliness, 14:52.000 --> 14:54.000 and, you know, this is great. 14:54.000 --> 14:56.000 There's an even an XML spec for it, 14:56.000 --> 14:59.000 with, you know, hundreds of abbreviated terms in it, you know? 14:59.000 --> 15:01.000 And there's a constraint solver, 15:01.000 --> 15:04.000 and a sheet layout language that goes on and on like this, 15:04.000 --> 15:06.000 for hundreds upon hundreds of lines 15:06.000 --> 15:08.000 to describe each shape. 15:08.000 --> 15:10.000 And you have to write a rendering engine 15:10.000 --> 15:11.000 that will follow these constraints, 15:11.000 --> 15:13.000 understand them reverse engineer what they do, 15:13.000 --> 15:15.000 and then draw that, 15:15.000 --> 15:17.000 which looks very smiley and friendly, doesn't it? 15:17.000 --> 15:21.000 It's just underneath it's all dry bones and horror, you know? 15:21.000 --> 15:22.000 So there you go. 15:22.000 --> 15:24.000 As they say often, you know, 15:24.000 --> 15:27.000 people have this wonderful view of everything being nice and clean 15:27.000 --> 15:29.000 and well done and an open and so on, 15:29.000 --> 15:32.000 and actually they have absolutely no ideas 15:32.000 --> 15:34.000 quite a famous saying, isn't it? 15:34.000 --> 15:36.000 One of the things that I fear 15:36.000 --> 15:41.000 is a plague of new duplicate lame office sweets. 15:41.000 --> 15:44.000 So simplified that they don't really do much, 15:44.000 --> 15:47.000 and assume a huge amount of money before they hit 15:47.000 --> 15:49.000 the obvious roadblocks. 15:49.000 --> 15:50.000 And with that money, 15:50.000 --> 15:53.000 we can actually just improve the state of the art so much, you know? 15:53.000 --> 15:55.000 So there you go. 15:55.000 --> 15:58.000 Jeng is a win ski who is sort of an old, old, 15:58.000 --> 16:01.000 and yes, well interesting, interesting chap. 16:01.000 --> 16:04.000 Mozilla Hacker, one of the earlier Mozilla Hacker, 16:04.000 --> 16:06.000 he talked about this CADT development model, 16:06.000 --> 16:09.000 and this is a genome bug that he suffered. 16:09.000 --> 16:11.000 And he was complaining that his bug, 16:11.000 --> 16:13.000 as always, was closed with, I'm sorry, 16:13.000 --> 16:16.000 and this entire subsystem, please try again, you know? 16:16.000 --> 16:19.000 And yes, he, of course, ran, you know, 16:19.000 --> 16:21.000 hardly seems worth even having a bug system. 16:21.000 --> 16:23.000 If the pace of from scratch rewrites, 16:23.000 --> 16:25.000 outstrips the pace of bug fixing, 16:25.000 --> 16:28.000 now this is particularly problematic in a stack of software, 16:28.000 --> 16:31.000 the people at the bottom of the stack need to be even more responsible 16:31.000 --> 16:33.000 and people at the top, you know? 16:33.000 --> 16:35.000 I just encourage that, and we'll see in a minute, 16:35.000 --> 16:37.000 why that might be useful. 16:37.000 --> 16:40.000 But when this cute naivety of everything is easy, 16:40.000 --> 16:41.000 and we can rewrite everything, 16:41.000 --> 16:43.000 and these other guys are idiots, 16:43.000 --> 16:45.000 meets the quest for personal glory. 16:45.000 --> 16:48.000 You know, I want to be the next Linus Torvals. 16:48.000 --> 16:50.000 I'm going to call it Meeks Office, or whatever. 16:50.000 --> 16:52.000 You know, it's just a bit lunatic. 16:52.000 --> 16:56.000 And so this huge bent to try to reinvent the wheel, 16:56.000 --> 16:58.000 and software is really hard. 16:58.000 --> 17:01.000 So I think the, the take home thing is that there are just 17:01.000 --> 17:03.000 so many axes, here are a few of them, 17:03.000 --> 17:05.000 to getting just office software right. 17:05.000 --> 17:09.000 And, you know, yeah, it's very easy to think of one of these. 17:10.000 --> 17:12.000 So I tend to say this to a lot of people, 17:12.000 --> 17:15.000 and they say, yeah, but collaboration, you know? 17:15.000 --> 17:17.000 Collaboration is the killer feature, and it is. 17:17.000 --> 17:19.000 I mean, collaboration made us use wikis, 17:19.000 --> 17:23.000 and learn new markup languages with asterisk in them. 17:23.000 --> 17:24.000 You know? 17:24.000 --> 17:26.000 I'm going back to word five days. 17:26.000 --> 17:27.000 It made us throw out wizi wik. 17:27.000 --> 17:31.000 So, so all we need to do is first rewrite it all in language of choice. 17:31.000 --> 17:34.000 Or, and then we'll rewrite everything to be, 17:34.000 --> 17:37.000 operational, transform, CRDT, insert, 17:37.000 --> 17:39.000 test, trendy technology. 17:39.000 --> 17:42.000 And it's not just idiots saying you should rewrite it completely. 17:42.000 --> 17:45.000 You know, there are major tech luminaries out there 17:45.000 --> 17:47.000 saying that we should, we should really be doing this. 17:47.000 --> 17:51.000 And 2009, thanks to the register for preserving that. 17:51.000 --> 17:53.000 But let me just give you an example of a little rewrite. 17:53.000 --> 17:54.000 This is a very simple one. 17:54.000 --> 17:56.000 It has no file format impact. 17:56.000 --> 17:58.000 It's a very simple refactor. 17:58.000 --> 17:59.000 Ha ha. 17:59.000 --> 18:01.000 Except we have quite a lot of user interface. 18:01.000 --> 18:05.000 At 1,260, Glade style XML UI dialogs. 18:05.000 --> 18:07.000 Quail on, spent multiple person years, 18:07.000 --> 18:09.000 and was helped by many other people. 18:09.000 --> 18:12.000 And if there is a 10x developer out there, 18:12.000 --> 18:14.000 it's probably Quail on. 18:14.000 --> 18:17.000 So, yes, a very simple thing, just rewriting the UI. 18:17.000 --> 18:21.000 So, but instead, we should rewrite the whole core engine of our office sweep. 18:21.000 --> 18:26.000 To use the amazingly popular Google Wave operational transforms model. 18:26.000 --> 18:27.000 Yeah. 18:27.000 --> 18:30.000 And the great news about this is you get undue for free. 18:30.000 --> 18:34.000 And here was I thinking, so just to give you a sort of background, 18:34.000 --> 18:36.000 there's a mathematical elegance to this. 18:36.000 --> 18:38.000 You have operations and inverse operations, 18:38.000 --> 18:41.000 and you can move things over each other. 18:41.000 --> 18:43.000 And it's just wonderful with these transforms. 18:43.000 --> 18:46.000 And everything looks very nice to a computer scientist. 18:46.000 --> 18:47.000 It all sounds great. 18:47.000 --> 18:50.000 And if you've only ever written small programs from scratch, 18:50.000 --> 18:51.000 it sounds beautiful. 18:51.000 --> 18:53.000 I totally agree. 18:53.000 --> 18:57.000 Where we just start from scratch, maybe that would be a good idea, maybe. 18:57.000 --> 19:00.000 But actually, undue is already quite easy. 19:00.000 --> 19:01.000 You've got copy and paste, right? 19:01.000 --> 19:04.000 You just copy it and stash it all somewhere and then shove it back. 19:04.000 --> 19:08.000 You don't really even need to understand it, just put it back like it was, right? 19:08.000 --> 19:10.000 So, yeah. 19:10.000 --> 19:12.000 So, here's a great example of this spreadsheet. 19:12.000 --> 19:14.000 You know, one guy does a sort of the cells, 19:14.000 --> 19:16.000 the other guy enters 42. 19:16.000 --> 19:18.000 And so, they get them in different orders. 19:18.000 --> 19:20.000 And we have to work out which order is the right one. 19:20.000 --> 19:21.000 But this guy is going to be the wrong one. 19:21.000 --> 19:23.000 So, now we're going to swap this round. 19:23.000 --> 19:26.000 We've already sorted, you know, 100,000 cells or something like that. 19:26.000 --> 19:30.000 And now we need to work out where to put the 42 so that we change the right cell. 19:30.000 --> 19:31.000 Yeah. 19:31.000 --> 19:33.000 Otherwise, these models are going to get out of sync. 19:33.000 --> 19:35.000 And life is not going to be a happy place, okay? 19:35.000 --> 19:37.000 And that seems reasonably easy, right? 19:37.000 --> 19:40.000 We just track where all the rows went and then we can, you know, find, 19:40.000 --> 19:43.000 but there's a whole lot of complexity that you might not be thinking about here. 19:43.000 --> 19:48.000 So, for example, you can actually, you can actually use random numbers in, 19:48.000 --> 19:49.000 informially. 19:49.000 --> 19:52.000 So, you know, you need to synchronize your random number generators 19:52.000 --> 19:54.000 and make sure that they are in the same state. 19:54.000 --> 19:57.000 But, you know, we can probably fit that in somewhere. 19:57.000 --> 20:00.000 But you can also make random dependency chains. 20:00.000 --> 20:04.000 So, cells can be dependent on or depend on things around them, right? 20:04.000 --> 20:07.000 Or, you know, so it's not very clear who does what. 20:07.000 --> 20:10.000 And of course, we need to calculate everything to the end. 20:10.000 --> 20:12.000 So, we have the real numbers having changed something. 20:12.000 --> 20:14.000 Otherwise, the source is going to be wrong. 20:14.000 --> 20:15.000 And you're out of sync. 20:15.000 --> 20:19.000 And, you know, what about those external data sources that we were downloading data from 20:19.000 --> 20:20.000 somewhere in between? 20:20.000 --> 20:21.000 Maybe that. 20:21.000 --> 20:25.000 But, you know, what about that beautiful feature of not calculating everything all at once, 20:25.000 --> 20:29.000 every time you type something, so that, you know, it's efficient and stuff. 20:29.000 --> 20:31.000 You know, like, say the background calculations running. 20:31.000 --> 20:33.000 So, all of these problems can be solved. 20:33.000 --> 20:36.000 Nothing I've said is, in solve, solve, obviously. 20:36.000 --> 20:39.000 But we're just talking about moving a editing a number, 20:39.000 --> 20:43.000 which incidentally doesn't have to be in the range of cells that's being edited. 20:43.000 --> 20:45.000 It could be anywhere in the sheet. 20:45.000 --> 20:46.000 You can't tell. 20:47.000 --> 20:51.000 So, you have to do all that work for this one pair of this one transform, right? 20:51.000 --> 20:54.000 So, hooray, after we've done that, we've done this and that's brilliant. 20:54.000 --> 20:57.000 But the problem is, this is only two operations. 20:57.000 --> 21:01.000 As you increase the number of operations that you need to transform over each other, 21:01.000 --> 21:06.000 the CPU complexity goes up, but that's nothing to the programmer complexity. 21:06.000 --> 21:11.000 The programming complexity of sorting these goes with the square of the number of operations. 21:11.000 --> 21:12.000 Okay? 21:12.000 --> 21:15.000 We talked about log log search complexity. 21:15.000 --> 21:19.000 How about squared programmer time complexity? 21:19.000 --> 21:20.000 Yeah? 21:20.000 --> 21:21.000 So, I mean, this is a great way. 21:21.000 --> 21:22.000 Operational transformations are great. 21:22.000 --> 21:25.000 And there's lots of hand waving about, but some of the operations are quite similar, 21:25.000 --> 21:27.000 and, you know, we can bucket them and so on. 21:27.000 --> 21:29.000 And obviously, Google Docs sort of works, 21:29.000 --> 21:33.000 but this ends up, meaning you don't have many features. 21:33.000 --> 21:37.000 Or you have a vast engineering team, or both. 21:37.000 --> 21:40.000 So, CRDTs are then the latest fashion in this, 21:40.000 --> 21:44.000 and, you know, they're absolutely ideal for very simple flat text. 21:44.000 --> 21:48.000 But our real documents are not just complex in their file, 21:48.000 --> 21:51.000 you know, they're standards and layout and memory representation, 21:51.000 --> 21:53.000 but they're also complex on disk. 21:53.000 --> 21:59.000 So, you know, we have these zipped file systems with embedded XML trees and millions of nodes in them, right? 21:59.000 --> 22:02.000 And we mutate them really quite a lot. 22:02.000 --> 22:05.000 And there's just a ton of invariants, you know, 22:05.000 --> 22:08.000 if this is that, that there also has to be that. 22:08.000 --> 22:11.000 There are implicit, there's no documentation for those. 22:11.000 --> 22:14.000 And so, of course, probably something like this can be done. 22:14.000 --> 22:19.000 There are people that have JSON trees that exist in CRDTs, 22:19.000 --> 22:23.000 and often they use a separate table of GUIDs, you know, 22:23.000 --> 22:27.000 these huge random numbers, to associate linkage and pointers and manage these, 22:27.000 --> 22:30.000 and it's all very clever, and they try and hide that from you inside. 22:30.000 --> 22:35.000 And that's all brilliant, until you try and save it in a file format. 22:35.000 --> 22:37.000 Because the far form, I doesn't have any of that. 22:37.000 --> 22:39.000 It doesn't have a massive table of GUIDs, 22:39.000 --> 22:43.000 associating attributes with other attributes to try and make this work, right? 22:43.000 --> 22:47.000 Yeah, and storing all of the state anyway, this never mind that. 22:47.000 --> 22:49.000 There's a whole live conflicting constraint. 22:49.000 --> 22:53.000 So, it's quite common that we get conflicts in constraints. 22:53.000 --> 22:56.000 So, we get files that say, hey, your minimum cell height, 22:56.000 --> 22:58.000 it's got to be at least this big, right? 22:58.000 --> 23:01.000 But actually, the page size is only this big. 23:01.000 --> 23:06.000 So, what do you do if you can't fit the minimum table cell height onto the page? 23:06.000 --> 23:08.000 And it's not obvious, right? 23:08.000 --> 23:10.000 Like there's a lot of ways you could do it. 23:10.000 --> 23:12.000 And we try and we try and get this right, 23:12.000 --> 23:16.000 but the world of office is full of invalid states, 23:16.000 --> 23:18.000 and transitioning between one and another. 23:18.000 --> 23:23.000 And so, you know, if you have your million monkeys typing your office productivity suite, 23:23.000 --> 23:26.000 it's not going to produce anything valid anytime soon. 23:26.000 --> 23:29.000 If that's going to produce a lot of very invalid things. 23:29.000 --> 23:31.000 So, there you go. 23:31.000 --> 23:33.000 So, here's another little UX win. 23:33.000 --> 23:36.000 So, another thing, just changing the widget toolkit, 23:36.000 --> 23:39.000 to use a native widget toolkit instead of the internal one. 23:39.000 --> 23:42.000 Again, touching a whole native UI staff, again, 23:42.000 --> 23:44.000 cray-lawn with help. 23:44.000 --> 23:46.000 More multiple person years. 23:46.000 --> 23:48.000 Again, no standards, problems, no complicated, 23:48.000 --> 23:52.000 some staff, simple refactors, and just reusing things. 23:52.000 --> 23:57.000 And we just got there in time to get native G2K3 dialogues 23:57.000 --> 24:01.000 by the time G2K4 was released and totally broke our threading model. 24:01.000 --> 24:03.000 So, there you are, you know, what can you do? 24:03.000 --> 24:07.000 In order to save a few, you know, days of work 24:07.000 --> 24:09.000 and improve windows, bindings or something, 24:09.000 --> 24:12.000 they just broke it really badly. 24:12.000 --> 24:15.000 Good news is it's the foundation for JavaScript dialogues, 24:15.000 --> 24:16.000 which you're doing good. 24:16.000 --> 24:18.000 So, anyhow, two better strands. 24:18.000 --> 24:19.000 Better ways of doing this. 24:19.000 --> 24:22.000 So, collaborative online, of course, uses online collaboration. 24:22.000 --> 24:26.000 And that exploits people's misunderstanding of how fast the internet is. 24:26.000 --> 24:29.000 So, do you know that you can get from Frankfurt to London, 24:29.000 --> 24:31.000 quicker than your Bluetooth keyboard, 24:31.000 --> 24:35.000 can get from your Bluetooth keyboard to the PC next to it. 24:35.000 --> 24:36.000 Right? 24:36.000 --> 24:40.000 Now, that's a pretty interesting result, isn't it? 24:40.000 --> 24:46.000 And, you know, Frankfurt Milan is getting faster than your 60-hurt's frame time. 24:46.000 --> 24:47.000 Right? 24:47.000 --> 24:51.000 So, the internet is an unbelievable wonder of performance. 24:51.000 --> 24:53.000 It's just extraordinary. 24:54.000 --> 24:57.000 And, yeah, this is, I think, a look, no, this is a linear plot, 24:57.000 --> 25:01.000 but you see the, you see the thing on our typing speed and our blinking speed. 25:01.000 --> 25:03.000 You know, a human, a human eye blink. 25:03.000 --> 25:05.000 That's a very fast human eye blink. 25:05.000 --> 25:08.000 People say it's, you know, 200 milliseconds, please. 25:08.000 --> 25:10.000 See, say by bringing everyone into one model, 25:10.000 --> 25:12.000 we can do really well on having done that, 25:12.000 --> 25:15.000 for the people who use their laptop and want on a train. 25:15.000 --> 25:17.000 We can make their life much better. 25:17.000 --> 25:20.000 We hackers understand a dipping, right? 25:20.000 --> 25:22.000 We have dipping tools, and that's great. 25:22.000 --> 25:24.000 We don't want to put ODF in get. 25:24.000 --> 25:26.000 We just want to show people the differences, right? 25:26.000 --> 25:28.000 What changed on one side and the other, 25:28.000 --> 25:32.000 and make it really easy for the whole world to get changed tracking right? 25:32.000 --> 25:34.000 Because it's really not currently easy. 25:34.000 --> 25:37.000 So, another thing that we've been investing in 25:37.000 --> 25:39.000 is improving user experience, 25:39.000 --> 25:42.000 because actually the user experience is what the user's experience. 25:42.000 --> 25:46.000 I know, I know, it took me weeks to work this out. 25:46.000 --> 25:49.000 And so, we, we have lots of really, really cool data, 25:50.000 --> 25:52.000 and actually the Zendis has done some great work 25:52.000 --> 25:55.000 to actually build the analytics here to help us work out. 25:55.000 --> 25:57.000 You know, which commands are used in what order? 25:57.000 --> 25:58.000 Who undoes what? 25:58.000 --> 26:00.000 What are they unhappy with? 26:00.000 --> 26:03.000 What are the most popular things that are going on there? 26:03.000 --> 26:05.000 How can we make it easier? 26:05.000 --> 26:08.000 How can we make the contextual toolbar near where your mouse is? 26:08.000 --> 26:10.000 So, you don't need that power assisted mouse anymore. 26:10.000 --> 26:13.000 You know, to track all the way across your full case screen 26:13.000 --> 26:15.000 up to the toolbar and back again, right? 26:15.000 --> 26:16.000 It's just convenient. 26:16.000 --> 26:18.000 And the tools in there are the tools that people use. 26:18.000 --> 26:21.000 By frequency in order, close to where you are. 26:21.000 --> 26:25.000 I'm context tooltabs so that you can, you know, find things that you want quickly 26:25.000 --> 26:27.000 and get to the tools that you want. 26:27.000 --> 26:30.000 Lots of new tabs to make things a prettier 26:30.000 --> 26:33.000 and more easier to use, more functional. 26:33.000 --> 26:35.000 And well, it goes on and on. 26:35.000 --> 26:36.000 Look at all of these great things. 26:36.000 --> 26:38.000 All of these features that we had already. 26:38.000 --> 26:39.000 We didn't actually have to do much to write these. 26:39.000 --> 26:41.000 We did a bit of JavaScript, a bit of cleanup, 26:41.000 --> 26:44.000 a bit of clarification, some CSS and all of that. 26:44.000 --> 26:45.000 Good stuff. 26:45.000 --> 26:49.000 A responsive toolbar so that they fit nicely on your screen. 26:49.000 --> 26:51.000 And, you know, I'm a big user of bullets 26:51.000 --> 26:53.000 as it turns out, our other people. 26:53.000 --> 26:55.000 So, when the paragraph thing collapses, it just collapses 26:55.000 --> 26:58.000 to a single bullet button, which is basically the button you want. 26:58.000 --> 27:00.000 You know, it's just nice to see it and use it and go, 27:00.000 --> 27:02.000 oh, it did what I expected. 27:02.000 --> 27:03.000 That's really cool. 27:03.000 --> 27:05.000 I'm better search results in the sidebar. 27:05.000 --> 27:08.000 Hiding latency and showing you, allowing you to navigate that 27:08.000 --> 27:10.000 and see your document tree and so on. 27:10.000 --> 27:13.000 So, all of that has been shipping nicely. 27:13.000 --> 27:15.000 But we got some new things coming soon. 27:15.000 --> 27:17.000 So, there's a two-page view coming up. 27:17.000 --> 27:19.000 So, side by side, and I think you'll always 27:19.000 --> 27:21.000 nice book binding things. 27:21.000 --> 27:22.000 A nice new inter-op feature here. 27:22.000 --> 27:25.000 Marcus Mohard doing fantastic work and Balash Varga. 27:25.000 --> 27:27.000 Looks particularly ugly in the screen shop, 27:27.000 --> 27:30.000 but it's getting better and a nice inter-op feature 27:30.000 --> 27:32.000 that people have been missing. 27:32.000 --> 27:34.000 Multi-sheet views. 27:34.000 --> 27:36.000 So, my competition have cheated. 27:36.000 --> 27:38.000 They totally cheated. 27:38.000 --> 27:41.000 So, they replicate the model on each of their machines. 27:41.000 --> 27:44.000 So, the first thing that they do is download the whole document 27:44.000 --> 27:49.000 every user and then they try and calculate a new on each machine. 27:49.000 --> 27:51.000 We do it once, essentially, which is nice. 27:51.000 --> 27:54.000 But these guys had a really easy feature to implement, 27:54.000 --> 27:56.000 which is this temporary views. 27:56.000 --> 27:59.000 I want to view of this sheet, other people don't see. 27:59.000 --> 28:02.000 And that's kind of easy when you duplicate the model everywhere. 28:02.000 --> 28:03.000 Like everyone has the model anyway. 28:03.000 --> 28:05.000 Like, what's the problem? 28:05.000 --> 28:08.000 For us, it's more tricky because we have a single model at the center. 28:08.000 --> 28:11.000 And so, yeah, so we did a whole lot of work here. 28:11.000 --> 28:13.000 And now, you know, it's very easy to then filter 28:13.000 --> 28:16.000 and not affect other people's views and see different views of that data. 28:16.000 --> 28:19.000 And we hope to return the favor by having multiple views, 28:19.000 --> 28:23.000 you know, and easily to use it in the sheet tab and switch between. 28:23.000 --> 28:25.000 So, that's fun. And why is that hard? 28:25.000 --> 28:28.000 Well, it seems a super easy thing to do as everything seems. 28:28.000 --> 28:32.000 Until you realize the calculation engine can use the visibility state 28:32.000 --> 28:34.000 to calculate different numbers. 28:34.000 --> 28:37.000 So, you have to, you know, do some fun stuff. 28:37.000 --> 28:40.000 We're improving mobile tool bars and all sorts of things. 28:40.000 --> 28:41.000 Follow me presentation. 28:41.000 --> 28:45.000 This is great. So, NLNet, NGI, zero, have funded this thing. 28:45.000 --> 28:46.000 And it's really cool. 28:46.000 --> 28:49.000 So, apart from having beavers and lots of beavers. 28:49.000 --> 28:52.000 Actually, I'm going to need help giving them out later. 28:52.000 --> 28:55.000 This then allows you to project your slides to other people effectively. 28:55.000 --> 28:57.000 Much as we're doing now. 28:57.000 --> 28:59.000 And why would you want to do that? 28:59.000 --> 29:01.000 Because, of course, what you can do is you can render that 29:01.000 --> 29:04.000 on your high resolution screen and then you can take screenshots of that. 29:04.000 --> 29:07.000 And you can compress that into H264. 29:07.000 --> 29:08.000 And you can stream that to everyone else. 29:08.000 --> 29:10.000 You can then decompress it again. 29:10.000 --> 29:13.000 And you can do that at 60 frames a second as your webGL transform, 29:13.000 --> 29:14.000 you know, happens. 29:14.000 --> 29:17.000 And it might, someone might even keep up if they're lucky. 29:17.000 --> 29:21.000 Up all, you can just send some JavaScript to them to run the presentation 29:21.000 --> 29:23.000 and the assets they need. 29:23.000 --> 29:26.000 And then they can run this presentation, see it all beautifully, 29:26.000 --> 29:29.000 beautifully happening on each of these people's screens. 29:29.000 --> 29:33.000 And they have actual content instead of pixel. 29:33.000 --> 29:36.000 So if they're visually impaired, they can go at their own speed. 29:36.000 --> 29:37.000 They can go back. 29:37.000 --> 29:39.000 They can read that bit on the Braille display better. 29:39.000 --> 29:41.000 They can get, it be included more. 29:41.000 --> 29:46.000 And if the presenter speaks too quickly, as can happen. 29:46.000 --> 29:49.000 Then they can, you know, keep up. 29:49.000 --> 29:52.000 You know, they can go back and check again and make sure they understood that. 29:52.000 --> 29:54.000 So, yeah, some nice stuff. 29:54.000 --> 29:57.000 And pretty as settings you I, just built in. 29:57.000 --> 30:00.000 And one of the things we're really focusing on is interoperability. 30:00.000 --> 30:02.000 We've talked to you about some of the problems. 30:02.000 --> 30:04.000 We've, you know, we solve huge numbers of them. 30:04.000 --> 30:06.000 But we want to get the numbers down to zeroes everywhere. 30:06.000 --> 30:08.000 So we've been using a tool. 30:08.000 --> 30:10.000 The game was helpful funded by Zendes. 30:10.000 --> 30:15.000 And just exposes the very worst bug documents out of our very worst bugs 30:15.000 --> 30:17.000 that are causing problems around tripping. 30:17.000 --> 30:21.000 And so there's a great team there of people just chasing that down to zero. 30:21.000 --> 30:23.000 And once we get it to zero, it stays there. 30:23.000 --> 30:25.000 You know, like this is, this is what we want to see. 30:25.000 --> 30:27.000 Lots of, lots of, lots of zeros. 30:28.000 --> 30:31.000 So we have all this beautiful new user experience in collaboration line. 30:31.000 --> 30:35.000 And we've been doing some great work to bring that to the desktop. 30:35.000 --> 30:37.000 But actually, that's something we've done for years. 30:37.000 --> 30:42.000 So on Chrome OS, you've always had something that runs on your Android phone. 30:42.000 --> 30:45.000 And if you plug it into a big screen and put a keyboard in, suddenly it's a full 30:45.000 --> 30:47.000 full office suite running on that device. 30:47.000 --> 30:51.000 Same thing on iPhone for Mac if you compile it right. 30:51.000 --> 30:55.000 And so, you know, native code brings a whole load of advantages. 30:55.000 --> 30:57.000 Um, offline support is one of them. 30:57.000 --> 31:01.000 Obviously, as you travel the same UI, but better, you know, 31:01.000 --> 31:05.000 high performance layer latency editing there, obviously, you know, in a slow train. 31:05.000 --> 31:07.000 But better than that, you have platform API. 31:07.000 --> 31:09.000 So the browser is a wonderful tool. 31:09.000 --> 31:12.000 And it provides an amazing infrastructure everywhere. 31:12.000 --> 31:14.000 Quite amazing in some ways. 31:14.000 --> 31:17.000 Um, two amazing and JavaScript, let's say, but anyway, 31:17.000 --> 31:20.000 um, but it also really is a bit security paranoid. 31:20.000 --> 31:23.000 So getting access to your clipboard is really difficult. 31:23.000 --> 31:26.000 You can't get rich content through it nicely, and that sucks. 31:26.000 --> 31:29.000 If you want to see whether you've got multiple screens or what the 31:29.000 --> 31:31.000 resolutions are, get a projector and project on it. 31:31.000 --> 31:32.000 None of that works. 31:32.000 --> 31:35.000 Um, if you want proper printing APIs, we talk about that. 31:35.000 --> 31:36.000 What about image formats? 31:36.000 --> 31:38.000 You know, there's a whole lot of things there. 31:38.000 --> 31:42.000 And we tried initially to do this, and we're still doing that work on 31:42.000 --> 31:43.000 WebAssembly. 31:43.000 --> 31:47.000 So Stefan Bergman, somewhere is a hero of, uh, 31:47.000 --> 31:50.000 I'm making this work, but the problem with that is that it's quite disappointing. 31:50.000 --> 31:53.000 It's not really share those webAssembly that big webAssembly 31:53.000 --> 31:56.000 blog between multiple tabs, which makes a real problem. 31:56.000 --> 31:58.000 There's a couple of bugs there in the browsers that, uh, 31:58.000 --> 32:00.000 Yeah, a problem for us. 32:00.000 --> 32:02.000 So anyway, native code makes sense. 32:02.000 --> 32:05.000 And so we're just bringing the same UI to the desktop with a 32:05.000 --> 32:08.000 little bit of wrapper around it to make it look, you know, nice and 32:08.000 --> 32:11.000 pretty, uh, and all of the features you know and love, uh, 32:11.000 --> 32:15.000 the look, uh, integration with your Mac goodness and so on, 32:15.000 --> 32:17.000 or badness, it's proprietary. 32:17.000 --> 32:18.000 You shouldn't use it. 32:18.000 --> 32:20.000 It's just free desktop. 32:20.000 --> 32:24.000 Um, and this is really complimentary to the classic product that we're all aware of. 32:24.000 --> 32:28.000 Um, so we continue into work on classic, and there's a huge tech overlap 32:28.000 --> 32:31.000 there, obviously in this giant community backing that code base. 32:31.000 --> 32:34.000 And the new collaborative is then really complimentary, and that really slim 32:34.000 --> 32:35.000 down thing. 32:35.000 --> 32:36.000 No Java. 32:36.000 --> 32:37.000 No database. 32:37.000 --> 32:39.000 No mail merge, master documents. 32:39.000 --> 32:41.000 We're taking lots of stuff out of that. 32:41.000 --> 32:45.000 Um, to make it cleaner and smaller and, you know, 32:45.000 --> 32:47.000 easier to handle and wrestle. 32:47.000 --> 32:49.000 We're doing a lot of collaboration. 32:49.000 --> 32:50.000 I'll skip that now. 32:50.000 --> 32:53.000 I think, um, for, uh, we have a talk later in the collaborative room. 32:53.000 --> 32:55.000 If you want to come and see how we're doing, um, 32:55.000 --> 32:58.000 collaboration and better, a hackfest at the bed for the hotel. 32:58.000 --> 33:01.000 So just walk upstairs and visit us Monday Tuesday. 33:01.000 --> 33:03.000 If you're interested, but a lot of beavers 33:03.000 --> 33:06.000 tooling away are doing doing good things. 33:06.000 --> 33:08.000 But if you can't make that because it's a bit late as notice, 33:08.000 --> 33:11.000 we'd love to see you in Hamburg, come and join the community, 33:11.000 --> 33:14.000 meet people, do team building, uh, give lots of input. 33:14.000 --> 33:17.000 Now, there's a whole lot of really cool people who have have done great stuff. 33:17.000 --> 33:20.000 And it would be remiss of me to stand here as one person. 33:20.000 --> 33:22.000 And not mention their names. 33:22.000 --> 33:24.000 Loads of people on, uh, can have a line. 33:24.000 --> 33:27.000 Loads of great translators, uh, doing, doing great work. 33:27.000 --> 33:30.000 And, and, and, and look at the complex text goodness in the middle of 33:30.000 --> 33:32.000 there to, uh, to get some of these people written right. 33:32.000 --> 33:34.000 Now, we also have a mascot. 33:34.000 --> 33:37.000 And, um, uh, toast and don't, can I grab you to come and grab the, 33:37.000 --> 33:39.000 the thing there because we'll need this surely. 33:39.000 --> 33:41.000 There's a whole case of them over here. 33:41.000 --> 33:43.000 So, so one of the things that I have as a problem, 33:43.000 --> 33:45.000 and maybe you, you do too. 33:45.000 --> 33:47.000 So, I work quite hard. 33:47.000 --> 33:48.000 You know, I love programming. 33:48.000 --> 33:50.000 Don't get enough chance to do enough of it, you know. 33:50.000 --> 33:53.000 And so, so there's a great temptation to ignore the real world 33:53.000 --> 33:55.000 and, and to go away and, uh, program. 33:55.000 --> 33:58.000 And, and say, we have a mascot that is the sort of, uh, 33:58.000 --> 34:01.000 uh, epitome of hard work, you know. 34:01.000 --> 34:03.000 It's always out there doing, doing hard things. 34:03.000 --> 34:05.000 And, and John Hall is modeling on here at the, at the front. 34:05.000 --> 34:07.000 Thank you, John. You know, excellent. Just like that. 34:07.000 --> 34:09.000 And, and we have a lot more here and say, where people will, um, 34:09.000 --> 34:12.000 people will get these if they ask questions or come and grab me 34:12.000 --> 34:16.000 afterwards. And, and this is actually a character from a comic about open source. 34:16.000 --> 34:19.000 So, if you, if you go, it's called the Open Road to Freedom. 34:19.000 --> 34:22.000 And it's a comic about communities, an open source, and companies, 34:22.000 --> 34:24.000 and it might even be funny if you read it. 34:24.000 --> 34:26.000 So, uh, yes, have a look at that. 34:26.000 --> 34:29.000 So, here's my conclusions. My conclusions are, what are they? 34:29.000 --> 34:32.000 Ah, office is a little bit harder than you think. 34:32.000 --> 34:33.000 It looks easy. 34:33.000 --> 34:37.000 Ah, lives of people are suckered by the ease of writing their own office suite 34:37.000 --> 34:40.000 and the glory and the fame of creating something new. 34:40.000 --> 34:43.000 But, actually, it's a bit of a nightmare. So, just be a little bit careful, 34:43.000 --> 34:47.000 uh, you know, before you jump in and come and talk to me and maybe you will learn something. 34:47.000 --> 34:50.000 Ah, the browser really does get you part way there. 34:50.000 --> 34:54.000 Um, but it would be really helpful if you could just join us and we build upwards. 34:54.000 --> 34:57.000 Instead of doing the Jamie Zowinsky, CADT, you know, 34:57.000 --> 35:03.000 attention deficit teen version of point eight being succeeded by completely rewritten point eight. 35:03.000 --> 35:05.000 And then another totally rewritten point eight. 35:05.000 --> 35:09.000 And when we could have been a, you know, uh, well, I can't multiply very well. 35:09.000 --> 35:14.000 Uh, 2.4. Um, so yes, look at this. This is, we should have a gun that kind of, you know, 35:14.000 --> 35:17.000 sprays people with, with hardworking beavers, but anyway. 35:17.000 --> 35:22.000 Anyway, I think I've said thank you, but, but the people I'd really like to highlight here are, 35:22.000 --> 35:25.000 uh, the volunteer is a fantastic, um, but we have loads of staff, 35:25.000 --> 35:28.000 some of them here, Torsten is, is, is, is, is, is one. 35:28.000 --> 35:34.000 Uh, we have lots of partners and customers that help fund it and actually we can't do anything without. 35:34.000 --> 35:36.000 There's a special insurance for this, it turns out. 35:37.000 --> 35:41.000 Um, and we, you really rock. So thank you for everyone that's contributed, you're absolutely amazing. 35:41.000 --> 35:45.000 Get involved, see me. We are hiring. We are trying to grow quite rapidly. 35:45.000 --> 35:50.000 Uh, there's obviously a big opportunity here and we want to make free software and documents better. 35:50.000 --> 35:54.000 So please come and join us. Thank you for your patience. You've been very good. 35:55.000 --> 35:57.000 Thank you. 36:03.000 --> 36:06.000 Now an upside of talking too quickly is that we have time for questions. 36:06.000 --> 36:10.000 Um, and I'm trying to get us back on, on schedule. So hopefully we, uh, 36:10.000 --> 36:14.000 we have a bit of, uh, bit of time. So if you have some questions of the form, you know, 36:14.000 --> 36:18.000 can I have a beaver? Um, we've got another hundred of them in the hotel that will come tomorrow. 36:18.000 --> 36:21.000 So just come and grab me somewhere around the Libra office next cloud, uh, 36:22.000 --> 36:25.000 uh, but in the meantime, if you have a question or a comments or a hickling or, uh, 36:25.000 --> 36:28.000 I love CRDTs, they're easy or anything like that, you know? 36:28.000 --> 36:31.000 John, what question do you have? What second? I can give you a microphone. 36:34.000 --> 36:38.000 First of all, I love my beaver. Okay. 36:38.000 --> 36:41.000 Second of all, uh, my pleasure. 36:41.000 --> 36:44.000 You mentioned spreadsheets and spreadsheets has always been a problem. 36:44.000 --> 36:48.000 How is your, how are you at importing Excel spreadsheets? 36:48.000 --> 36:49.000 Fantastic. 36:49.000 --> 36:57.000 So Excel support is, I think, outstanding that you're not going to find anything else outside of Excel that does it as well as we do. 36:57.000 --> 37:02.000 Um, the performance is great. We're threading stuff. It works really well. 37:02.000 --> 37:05.000 And of course, because we're running out on the server, we can have quite a big server. 37:05.000 --> 37:08.000 So you can load really a big spreadsheet on your phone, 37:08.000 --> 37:11.000 and you're not downloading all of this stuff. You're just getting a view into that. 37:11.000 --> 37:15.000 So, you know, it's, it's, it's pretty good. I think the ergonomics and the UI, 37:15.000 --> 37:19.000 we're continually improving, but that's, you know, that's what I'd be. 37:19.000 --> 37:21.000 Good question. Yeah, I can. 37:21.000 --> 37:26.000 On one, you skipped quickly over the collaboration thingy. 37:26.000 --> 37:32.000 Can you give short impression of what the state of the artists of this protocol? 37:32.000 --> 37:35.000 The communication protocol? 37:35.000 --> 37:37.000 Two slides back. 37:37.000 --> 37:43.000 You said there's a presentation later, but for people who cannot make it, 37:44.000 --> 37:49.000 there's quite a lot of slides on them. 37:49.000 --> 37:50.000 This guy, shoot. 37:50.000 --> 37:52.000 Yeah. 37:52.000 --> 37:55.000 Just for you, Hiker. 37:55.000 --> 37:58.000 You know, the, the collaboration. 37:58.000 --> 38:06.000 It's a slide. You skip very quickly over collaboration between applications. 38:06.000 --> 38:09.000 I still can't hear what you're saying, which is one of the things. 38:09.000 --> 38:11.000 We're a fine Hiker. 38:11.000 --> 38:13.000 Collaboration. Oh, yeah, sure. 38:13.000 --> 38:17.000 I mean, we do collaboration, but we do it by bringing everyone into the same document model. 38:17.000 --> 38:19.000 So, with collaborative line, you can collaborate. 38:19.000 --> 38:23.000 All these document formats, and that works nicely, but there's a single model. 38:23.000 --> 38:26.000 So, you don't have this model synchronization problem. 38:26.000 --> 38:29.000 Right? The issue with this is that you have two models, 38:29.000 --> 38:33.000 and you have to somehow make sure they're consistent and kept in stay. 38:33.000 --> 38:35.000 And that's really not easy. 38:35.000 --> 38:44.000 Like, like, caching is one thing, but model synchron distributed model replication is a tricky problem to get right. 38:44.000 --> 38:49.000 And people do it in databases, for example, and also, so, you know, places, 38:49.000 --> 38:54.000 but it's, it's not easy with extremely complicated data structure. 38:54.000 --> 38:56.000 Stuff. 38:56.000 --> 38:57.000 Good question. 38:57.000 --> 38:58.000 Anyone else? 38:58.000 --> 38:59.000 Yeah. 38:59.000 --> 39:00.000 Yeah. 39:00.000 --> 39:05.000 You're starting desktop editors based on, like, the existing collaboration line stuff, right? 39:05.000 --> 39:06.000 Correct. 39:06.000 --> 39:07.000 Yes. 39:07.000 --> 39:11.000 Are there plans to do integration with something like, let's say, next cloud? 39:11.000 --> 39:16.000 So, if I think a bunch of documents to my computer, I can just, like, double click a ODT file. 39:16.000 --> 39:18.000 It opens and is collaborative. 39:18.000 --> 39:22.000 Because currently, we have the problem of, if I want to collaborate, I need to open it in the browser, 39:22.000 --> 39:26.000 but if I double click a dog eggs, then it locks it for everyone else, and it's just a mess. 39:26.000 --> 39:28.000 That's a brilliant question. 39:28.000 --> 39:29.000 Yes. 39:29.000 --> 39:30.000 Of course. 39:30.000 --> 39:34.000 And the answer is that, come to first of all, I'll talk in the collaboration dev room later. 39:34.000 --> 39:37.000 So, we'll be talking about that, and Stefan's been doing some great work, if I can. 39:37.000 --> 39:39.000 Ah, Stefan Bergman over here. 39:39.000 --> 39:40.000 If you wave, Stefan. 39:40.000 --> 39:41.000 Don't. 39:41.000 --> 39:44.000 So, if you're particularly interested, talk to Stefan. 39:44.000 --> 39:49.000 And the beautiful thing is there, because you have the same UI, you can switch, then, 39:49.000 --> 39:53.000 transparently in your tab, and ideally, keep your cursor position or whatever, and suddenly, you're 39:53.000 --> 39:55.000 collaborating with other people on a remote server. 39:55.000 --> 39:59.000 Oh, until your connection doesn't work, and then you can bring it back to your machine 39:59.000 --> 40:00.000 and edit there. 40:00.000 --> 40:04.000 So, we can take a load of CPU load off the server when we need to. 40:04.000 --> 40:09.000 If you're just one person in the document, which turns out to be 90% of the time in collaborative 40:09.000 --> 40:12.000 systems, actually, is just one person collaborating with themselves. 40:12.000 --> 40:14.000 Which is a little bit sad. 40:14.000 --> 40:20.000 But actually, if you consider what you do, typically, you finish your document, and then you ask 40:20.000 --> 40:26.000 for review, rather than mashing it up horribly without thinking and, you know, collaborating the same 40:26.000 --> 40:27.000 time. 40:27.000 --> 40:29.000 So, yes, perfect. 40:29.000 --> 40:31.000 We're, of course, engaged in that. 40:31.000 --> 40:36.000 And there's some fun stuff there, detecting this file you've opened locally, is actually being 40:36.000 --> 40:40.000 synced by a, you know, a sync client, and now I need to go and talk, you know, there's a lot of plumbing, 40:40.000 --> 40:43.000 but not a lot of complexity in doing that. 40:43.000 --> 40:46.000 So, yeah, we're looking forward to getting that working. 40:47.000 --> 40:53.000 Oh, so, no, no much of a technical more on the marketing side. 40:53.000 --> 41:01.000 So, currently, obviously, the world is going from like desktop office suits to online ones. 41:01.000 --> 41:09.000 Like this gives you an opportunity because, like, then people stop expecting pixel perfect rendering. 41:09.000 --> 41:15.000 And, over the, I guess, you have, like, two main competitors, the Microsoft and the Google, 41:15.000 --> 41:21.000 the Google suits are apart from digital sovereignty. 41:21.000 --> 41:30.000 What other killer features would you recommend as a, like, what little bit of office does. 41:30.000 --> 41:33.000 Our collaborative story. 41:33.000 --> 41:37.000 So, killer features in caliber office that aren't there in the other online office 41:37.000 --> 41:38.000 sweets. 41:38.000 --> 41:42.000 Yeah, I mean, we have some just amazing cutting-edge features that you can use in 41:42.000 --> 41:46.000 you can have a online writer, for example, that aren't there in Microsoft's writer. 41:46.000 --> 41:47.000 Good work, Benoit. 41:47.000 --> 41:48.000 This guy's a hero. 41:48.000 --> 41:49.000 He's one of the customers. 41:49.000 --> 41:50.000 Yeah. 41:50.000 --> 41:51.000 Awesome. 41:51.000 --> 41:54.000 Million plus seats in France, French education, that man. 41:54.000 --> 41:55.000 What a hero. 41:55.000 --> 41:56.000 So, yes. 41:56.000 --> 42:03.000 So, in Microsoft Office online, you can't insert a chart in a word document. 42:03.000 --> 42:04.000 I'll just say that again. 42:04.000 --> 42:06.000 You can't insert a chart in a word document. 42:06.000 --> 42:10.000 Now, you're not going to believe me because that seems totally lunatic, right? 42:10.000 --> 42:15.000 The leading office suite provider has suddenly decided we don't need charts in word processing 42:15.000 --> 42:16.000 sweets. 42:16.000 --> 42:20.000 Now, as you probably can copy it from Excel, but you're probably going to get a bit map image 42:20.000 --> 42:21.000 with false color around it. 42:21.000 --> 42:24.000 So, when you size it, you get like blue red green, you know? 42:24.000 --> 42:27.000 I mean, just startlingly bad. 42:27.000 --> 42:32.000 And the thing itself is riddled with adverts for using the desktop version 42:32.000 --> 42:33.000 instead. 42:33.000 --> 42:36.000 So, you know, it's constantly saying, are you sure you want to do this? 42:36.000 --> 42:37.000 This seems a bit difficult. 42:37.000 --> 42:39.000 Please continue to use Windows. 42:39.000 --> 42:42.000 And store your data in one cloud while you're there, right? 42:42.000 --> 42:44.000 So, yeah. 42:44.000 --> 42:48.000 I mean, I think from a feature function perspective web versus web, 42:48.000 --> 42:52.000 we kick Microsoft's backside already quite hard in many ways. 42:52.000 --> 42:55.000 And Google, you know, on a par or, you know, 42:55.000 --> 42:56.000 we win and lose. 42:56.000 --> 42:57.000 Interrupt. 42:57.000 --> 42:58.000 We kick Google's backside. 42:58.000 --> 43:01.000 We're just much better interrupt than they are. 43:01.000 --> 43:04.000 So, yeah, I think there's everything to play for in that market. 43:04.000 --> 43:05.000 And we don't need to be ashamed. 43:05.000 --> 43:06.000 We should say you should use this. 43:06.000 --> 43:08.000 It's really good. 43:08.000 --> 43:09.000 Thank you. 43:09.000 --> 43:10.000 Oh, different. 43:10.000 --> 43:11.000 So, wait. 43:11.000 --> 43:12.000 And thank you to you. 43:12.000 --> 43:13.000 Yeah. 43:13.000 --> 43:17.000 So, my previous job and very clear talk, 43:17.000 --> 43:22.000 would you mind going back to the number of bags along the time? 43:22.000 --> 43:23.000 Oh, yeah, this one. 43:23.000 --> 43:25.000 This is a very specific kind of bug. 43:25.000 --> 43:27.000 It's an interoperability bug. 43:27.000 --> 43:32.000 Of a specific targeted interoperability type. 43:32.000 --> 43:33.000 Yeah. 43:33.000 --> 43:38.000 So, is there any reason why ODD is still? 43:38.000 --> 43:42.000 Oh, yes, just bad prioritization by me. 43:42.000 --> 43:46.000 So, I was particularly annoyed with the presentation fidelity. 43:46.000 --> 43:48.000 And so, I was like, right, got to fix and press. 43:48.000 --> 43:50.000 So, we did that first. 43:50.000 --> 43:51.000 And now we're doing these. 43:51.000 --> 43:53.000 But there's nothing wrong with ODT. 43:53.000 --> 43:57.000 This is just, this is just exporting it back to DocX. 43:57.000 --> 44:00.000 Because there's quite a large difference. 44:00.000 --> 44:05.000 If you get a very complicated ODT, then sometimes when you import that in Microsoft, 44:05.000 --> 44:08.000 it will complain about the document structure in some way. 44:08.000 --> 44:09.000 Okay. 44:09.000 --> 44:10.000 Oh, it seems bad. 44:10.000 --> 44:11.000 Usually you get the document. 44:11.000 --> 44:13.000 It just likes to complain. 44:13.000 --> 44:16.000 So, you know, if we can stop the complaints, they scare users, you know. 44:16.000 --> 44:19.000 But all your contents there and we can reload it. 44:19.000 --> 44:20.000 But yeah. 44:20.000 --> 44:25.000 So, and just to give you a perspective, you know, we have like a test corpus of 44:25.000 --> 44:29.000 quarter of a million, third of a million of the worst documents we can find in the world. 44:29.000 --> 44:30.000 Right? 44:30.000 --> 44:34.000 And so, this is some fraction of a percent of the real usage. 44:34.000 --> 44:39.000 And, you know, this is the sort of numbers we're seeing for actual problems doing that 44:39.000 --> 44:40.000 roundship from Microsoft. 44:40.000 --> 44:41.000 So, yeah. 44:41.000 --> 44:43.000 So, we're making real progress. 44:43.000 --> 44:44.000 I'm hopefully next falls down. 44:44.000 --> 44:45.000 This number will be zero. 44:45.000 --> 44:48.000 And then it will stay zero because we have automation. 44:48.000 --> 44:52.000 And, you know, as soon as it goes up, we'll stop it happening. 44:52.000 --> 44:53.000 So, thank you so much. 44:53.000 --> 44:55.000 You've been super good audience. 44:55.000 --> 44:56.000 So, bless you. 44:56.000 --> 44:57.000 And come and see me. 44:57.000 --> 44:58.000 Grab a beaver. 44:58.000 --> 44:59.000 Thank you so much. 44:59.000 --> 45:06.000 Thank you so much. 45:06.000 --> 45:07.000 Thank you very much. 45:07.000 --> 45:12.000 And I hope you didn't scare away all the developers, the potential ones. 45:12.000 --> 45:13.000 And yeah. 45:13.000 --> 45:14.000 Thanks again.