| 02:08 | <Hixie> | http://www.w3.org/mid/D11E7D60-62A0-426A-B379-62A950A9C487⊙nc |
| 06:26 | <hsivonen> | http://blogs.adobe.com/jd/2009/04/thoughts_on_rubys_html_reunifi.html |
| 08:29 | <jgraham> | I am unconvinced that the bibliographic data use case is really strong. I mean it is very useful if there is a list of papers to be able to get bibliographic data for them but the main problem I had when using such systems was a) the bibliographic data was incorrect or b) the bibliographic data was in the wrong format and could not be trivially converted to the correct fprmat |
| 08:30 | <jgraham> | Neither of which seems like it can be solved at the HTML level |
| 08:32 | <hsivonen> | its pretty useful that ACM Portal gives you snippets you can paste into .bib |
| 08:32 | <hsivonen> | I used .bib with HTML in oder to avoid reinventing the wheel |
| 08:32 | <hsivonen> | *order |
| 08:33 | <jgraham> | My experience is mostly with NASA's ADS which can provide .bib |
| 08:33 | <jgraham> | A link to a bib file is a fine solution to the problem as long as a) you are using TeX and b) the information in the .bib is correct |
| 08:34 | <jgraham> | (my understanding is that in life sciences research, LaTeX is very uncommon and citations are usually in some other format) |
| 08:36 | <hsivonen> | jgraham: I used .bib without TeX |
| 08:36 | <hsivonen> | jgraham: off-the-shelf .bib parsers FTW |
| 08:37 | <Philip`> | Google Scholar provides BibTeX entries too |
| 08:37 | <Philip`> | (among other formats) |
| 08:37 | <jgraham> | hsivonen: What exactly did you do? Preprocess a HTML document adding bibtex-extracted information? |
| 08:38 | <hsivonen> | jgraham: preprocess an XHTML document to add a bibliography outputting both HTML and XHTML |
| 08:39 | <jgraham> | hsivonen: Code? |
| 08:39 | <jgraham> | (it's not important, I'm just curious) |
| 08:39 | <hsivonen> | jgraham: http://hsivonen.iki.fi/thesis/ has a link |
| 08:39 | <hsivonen> | http://hsivonen.iki.fi/thesis/bib4ht-0.9.tar.gz |
| 08:40 | <jgraham> | hsivonen: Thanks |
| 08:45 | Philip` | notes that the ACM people ask for the .tex or .doc source of submitted papers, and in the .tex case it must include the post-processed bibliography section (i.e. the .bbl file, not the .bib), so they can extract the metadata for the ACM digital library |
| 08:45 | <Philip`> | but I guess in the .doc case they'll have to extract it manually |
| 08:45 | <jgraham> | Philip`: The same is true on arXiv.org although I'm not sure they do anything useful with the bbl file |
| 08:48 | <jgraham> | It seems like "annotate structured data for private/small group use" and "using the data should not involve learing new vocabularies" are rather mutually contradictory requirements |
| 08:49 | <jgraham> | I'm also not sure what Amazon has to do with that use case since Amazon marking up its prices would presumably not be for small group use |
| 08:50 | <hsivonen> | jgraham: what if the small group already has a vocabulary |
| 08:51 | <hsivonen> | amazon is a group of one, I guess |
| 08:52 | <jgraham> | hsivonen: Then class values - which are specifically excluded as a solution seem fine |
| 08:52 | <jgraham> | I'm not really sure why the use case document is discussing solutions there at all to be honest |
| 09:04 | <Philip`> | http://www.charlespetzold.com/etc/CSAML.html |
| 09:07 | <Hixie> | class values are excluded as a solution? |
| 09:09 | <jgraham> | Hixie: Using the data should not involve learning a plethora of new APIs, formats or vocabularies [...] it's possible to get information from sites consistently using 'class' values in a documented way, but doing so requires learning a new vocabulary |
| 09:09 | <jgraham> | (that should have been in quotes) |
| 09:10 | <Hixie> | ah, yes. well, that's the requirement someone had. |
| 09:10 | <jgraham> | That seems to at least suggest that class values are not an acceptable solution |
| 09:10 | <Hixie> | i don't know if it makes sense particularly. |
| 09:11 | <jgraham> | The whole requirement makes no sense to me because I don't think it is possible to get at novel information without learning the associated vocabulary |
| 09:11 | <Hixie> | well this is why we right down the requirements without thinking about whether they are possible or not |
| 09:11 | <Hixie> | if it turns out we can't find a solution, then we see which requirement we'd be first willing to drop |
| 09:12 | <jgraham> | Oh, I didn't realise that reasonableness and logical consistency weren't requirements on requirements |
| 09:12 | <Hixie> | well it's hard to tell what's reasonable until you try to look for solutions |
| 09:13 | <jgraham> | This seems to fall more under logical consistency |
| 09:15 | jgraham | gets to a requirement with the words "generate custom UI" in, wishes the alarm bells and sirens would stop |
| 09:16 | <Hixie> | jgraham: it would be possible to come up with an API that solved all problems without the author having to learn any new languages |
| 09:17 | <Hixie> | jgraham: window.doTheRightThing(freeformTextInAnyNaturalLanguage); |
| 09:17 | <Hixie> | it might require a lot of work from implementations |
| 09:18 | <jgraham> | Hixie: Still doesn't work because the user would have to learn the right natural-language terms to use |
| 09:18 | <Hixie> | i could spec it so the method does the right thing regardless of which words are used :-) |
| 09:19 | <jgraham> | well yes, I guess window.readMyMind() would work for some definition of "work" which doesn't mind the fact that it may be impossible |
| 09:20 | <Hixie> | there's already research into mind-computer interaction |
| 09:20 | <Hixie> | it might not be as impossible as you think |
| 09:20 | <Hixie> | admittedly, unlikely to be suitable for a spec today |
| 09:20 | <jgraham> | Hixie: It's not the neural interface that I think is the hard part |
| 09:21 | <jgraham> | It's providing the answer to an ill-defined question provided by the neural interface |
| 09:21 | <Hixie> | fair enough |
| 09:21 | <Hixie> | anyway |
| 09:21 | <jgraham> | :) |
| 09:21 | Philip` | thought mind-computer interaction was currently limited to distinguishing two states and therefore playing Pong |
| 09:21 | <Hixie> | i don't want to exclude requirements or use cases until i consider them seriously |
| 09:22 | <Hixie> | and i don't want to do that until i have a complete list |
| 09:44 | <Philip`> | Someone should write a tool to generate names to use in use cases |
| 09:44 | <Philip`> | (Paul, David, Fred, etc) |
| 09:45 | <Philip`> | Ideally it would have a checkbox to make the names follow the pattern A*, B*, C*, etc, and sliders for gender diversity and racial diversity |
| 09:51 | <annevk2> | Hixie, the sending steps specify WebSocket-Protocol but the receiving steps specify websocket-protocol |
| 09:52 | <Hixie> | correct |
| 09:52 | <Hixie> | the receiving steps lowercase the header. |
| 09:54 | <annevk2> | hmm ok |
| 09:56 | <Hixie> | i was reading something recently about someone saying they liked minimal specs and comparing SVG and SMIL to canvas and other specs |
| 09:56 | <Hixie> | showing how the simpler ones were more effectively deployed and implemented |
| 09:56 | <Hixie> | but i can't work out where that is anymore |
| 09:56 | <Hixie> | anyone recognise what i'm talking about? |
| 09:57 | <Hixie> | might have been in the context of o3d... |
| 09:57 | <annevk2> | yes, it was in response to O3D by some Mozilla guy |
| 09:57 | <annevk2> | right |
| 09:57 | <annevk2> | hmm |
| 09:57 | <Hixie> | aha |
| 09:57 | <Hixie> | http://www.0xdeadbeef.com/weblog/?p=1223 |
| 09:57 | <Hixie> | google rocks |
| 09:58 | <Hixie> | first hit for [o3d svg smil canvas] |
| 09:59 | <Philip`> | He seems to be overstating how high-level O3D is |
| 10:01 | <Philip`> | Something like X3D, where you can declare a sphere and it will be drawn, is much higher level than O3D, where you have to create a render graph that says to clear the buffer and traverse the object tree and execute a draw-list which is a lot of draw-elements, etc |
| 10:02 | <Hixie> | luckily for me, 3d is far outside my area of expertise |
| 10:04 | <Philip`> | (http://coderhump.com/archives/427 says some possibly interesting things) |
| 10:21 | <hsivonen> | I guess 3D solutions depend a lot on use cases |
| 10:21 | <hsivonen> | declarative scene graphs probably aren't important to have in browsers if you want to address gaming |
| 10:22 | <hsivonen> | but if you want to address rotatable e-commerce product samples, declarative scene graphs would be nice |
| 10:24 | <Philip`> | I don't think they help much with rotatable e-commerce product samples - you'd get a mesh from a 3D modelling program, use a standard library that renders it and rotates it using whatever underlying API there is, and that's that |
| 10:24 | <Philip`> | Declarative scene graphs (like in X3D) seem much more useful when people are creating complex 3D scenes, e.g. virtual worlds |
| 10:25 | <hsivonen> | Philip`: I guess that depends on one's philosophical outlook on the Rule of Least Power |
| 10:25 | <Philip`> | (either by hand or using a special editing program) |
| 10:26 | <Philip`> | (People who I've heard using X3D seem to generally claim they like it because they don't have to be programmers to use it) |
| 10:26 | <hsivonen> | Philip`: there were an open multivendor declarative format with quality renderers in all browsers, wouldn't you want to use that on an ecommerce site? |
| 10:27 | <Philip`> | hsivonen: I don't see any reason why I'd prefer it to an open multivendor imperative format with quality renderers in all browsers |
| 10:27 | <Philip`> | s/format/API/ |
| 10:27 | <jgraham> | Are there any real-life examples of non-games where 3D is adding significant value? |
| 10:27 | <jgraham> | (and I guess non-specialised application i.e. not medical imaging) |
| 10:27 | <hsivonen> | (I very well do see the point that the game-oriented API is needed and has faster time-to-market, so it may be that the scene graph approach never takes off) |
| 10:27 | <Philip`> | jgraham: Modern desktop UIs |
| 10:28 | <Philip`> | with their fancy zooms and rotating windows and stuff |
| 10:28 | <jgraham> | Philip`: That's not really 3D is it? |
| 10:28 | <jgraham> | I mean it's using the 3D card but most of the useful affects are really 2D |
| 10:28 | <jgraham> | /a/e/ |
| 10:28 | <hsivonen> | jgraham: pseudo-3D (QuickTime VR) is used for showing shiny Apple products after Stevenotes |
| 10:29 | <MikeSmith> | jgraham: is 2nd life a non-game? does it use real 3D? |
| 10:29 | hsivonen | doesn't understand the use cases of 2nd life |
| 10:29 | <hsivonen> | (I've never used it) |
| 10:29 | <jgraham> | MikeSmith: I don't know if 2nd life is a non-game. I always saw it as a game |
| 10:29 | <jgraham> | (but I have also never used it) |
| 10:30 | <Philip`> | jgraham: They're generally mostly-2D-ish effects, but sometimes they're really 3D, like the window-switcher in Vista |
| 10:30 | <Philip`> | and other effects like the ripples when dropping widgets in OS X are only really possible with 3D technology like pixel shaders |
| 10:30 | <hsivonen> | Philip`: those are for CSS 3D transitions--not the 3D scene rendering stuff whatever that one ends up being |
| 10:31 | <Philip`> | Also there's stuff like the Cover Flow UI |
| 10:31 | <jgraham> | Philip`: Fair enough. That sounds like it needs a low-level API though |
| 10:31 | <hsivonen> | how does O3D compare to Opera's canvas 3D from 2006? |
| 10:32 | <Philip`> | O3D differs in that it has features |
| 10:32 | <Philip`> | and a scene graph |
| 10:32 | <MikeSmith> | I would hope that some of the use cases of 2nd life are the same as real life (and not just the same as games) |
| 10:32 | <hsivonen> | oh, I thought Opera had a scene graph |
| 10:32 | <Philip`> | Opera's thing just had some functions for transforming the viewpoint, and global state parameters for colours and textures, and then functions for drawing triangles and for drawing models (which are collections of triangles), and that was pretty much it |
| 10:34 | <Philip`> | Google Earth and Google Street View are obvious uses of 3D |
| 10:34 | <jgraham> | Ah, Google street view is quite a nice example I guess |
| 10:34 | <Philip`> | And if you have Google Earth in a browser, you might also want an editing program like SketchUp in there too |
| 10:34 | <hsivonen> | would one want to hand a declarative model of the Earch including building to a browser and let it figure out LOD and stuff? |
| 10:35 | <hsivonen> | *Earth |
| 10:35 | <hsivonen> | I do want browsers to obsolete the dedicated Google Earth desktop app |
| 10:36 | <Philip`> | hsivonen: You could do something like http://www.web3d.org/x3d/specifications/ISO-IEC-19775-1.2-X3D-AbstractSpecification/Part01/components/geodata.html#GeoLOD |
| 10:36 | <hsivonen> | anyway, moving Google Earth into a browser seems like a use case for a low-level rendering API--not for a standard scene graph |
| 10:36 | <Philip`> | where your declarative model of the Earth is a low-LOD version, and it declares the URLs of recursively higher-LOD versions |
| 10:37 | <hsivonen> | Philip`: ok. makes sense |
| 10:37 | hsivonen | wonders how many polygons a URL is worth in terms of bytes |
| 10:37 | <Philip`> | Not much fun if you're the first person in the world to want to do a 3D model of the Earth and the relevant features haven't been added to the declarative spec and all the implementations yet |
| 10:38 | <Philip`> | unless it provides enough low-level primitives that you can emulate the right effect with decent performance |
| 10:39 | <Philip`> | hsivonen: Use TinyURL! |
| 10:41 | <hsivonen> | It would be cool to have a .pov scene renderer on top of canvas 3D |
| 10:41 | <MikeSmith> | so Google Street View seems potentially related to the Social Web to me -- at least more so than widgets (which are mentioned specifically in the Social Web WG charter are) |
| 10:41 | <hsivonen> | procedural textures would be hard to convert, though |
| 10:42 | <hsivonen> | in 2002, I wrote a simple generator that took OpenGL calls and emitted .pov |
| 10:42 | <hsivonen> | when our team found that we wanted a scene graph for pre-rendering textures in POV-Ray when we already had the data entangled in the source of a C program |
| 10:44 | <Philip`> | hsivonen: Convert procedural textures into GLSL :-) |
| 10:47 | <Philip`> | Oh, another use case for 3D is to take the Web 2.0 idea of logos-on-a-reflective-surface and extend it to logos-on-a-reflective-surface-that-is-animated-like-rippling-water |
| 10:48 | <Philip`> | Also: adverts |
| 10:48 | <Philip`> | which need to be as visually stimulating as possible, while not taking so long to download that the user has left the page before it's shown |
| 10:51 | hsivonen | notes that Google became successful in the ad business by having ads that aren't visually stimulating |
| 10:55 | <Philip`> | But obnoxious adverts still exist, so they're a valid use case |
| 10:55 | <Hixie> | hsivonen: wrong lesson. :-) |
| 10:55 | <Hixie> | hsivonen: we became successful by having them be useful |
| 10:58 | <Philip`> | "rel=next failed because UAs didn't expose it because authors didn't use it because UAs didn't expose it" - Opera's approach seems like a solution to that, since the 'forward' button will go to the rel=next page (if there is one) else it'll use some heuristics to guess where to go next |
| 10:59 | <Philip`> | and the heuristics are the way to bootstrap the process, so users will use the feature, and then authors will find it useful to improve the feature by using rel=next |
| 10:59 | <Hixie> | ok, rel=up |
| 10:59 | jgraham | disagrees with Philip` |
| 10:59 | <Hixie> | but in general yes, if the browsers did something, it would be more useful |
| 10:59 | <Hixie> | rel=chapter might be an even better example |
| 10:59 | <jgraham> | Because I guess users don't use features that have unpredicatable behaviour |
| 11:00 | <Philip`> | jgraham: It might not actually work in practice, but I still think it seems like a solution :-) |
| 11:00 | <Philip`> | It'd be okay if the heuristics were generally reliable |
| 11:00 | <jgraham> | That's the problem with things that sound like they ought to work |
| 11:00 | <Philip`> | but the next-page heuristics in Opera only seem to work about half the time I try to use it |
| 11:01 | <hsivonen> | Fast Forward is the #1 Opera feature I miss in Firefox |
| 11:01 | Philip` | likes how it works at navigating Apache directory listings, automatically knowing the next file in the sequence |
| 11:04 | hsivonen | notes that Opera Mobile is pretty good at locating the start of real content after boilerplate cruft even without HTML5 structural elements and without ARIA |
| 13:05 | <Philip`> | Hmm, did Sourceforge recently change its bug tracker to be even less usable? |
| 13:05 | <Philip`> | In particular it seems to hide all the comments by default |
| 13:06 | <Philip`> | (And the comments are in reverse chronological order, so you have to read up the page to understand the conversation) |
| 13:51 | <Lachy> | Geocities has finally collapsed. http://tech.slashdot.org/article.pl?sid=09/04/23/2339224 |
| 13:51 | <Lachy> | "So Long, GeoCities: We Forgot You Still Existed" - LOL :-) http://www.pcworld.com/article/163765/so_long_geocities_we_forgot_you_still_existed.html |
| 13:52 | Philip` | hopes the content will be archived somewhere |
| 13:54 | <Philip`> | I remember having fun searching for a nice-looking four-digit number in TimesSquare to create my site |
| 13:56 | <Lachy> | I guess the archive team might have a go at saving as much as they can. http://archiveteam.org/ |
| 13:57 | <Lachy> | it's already on their Deathwatch list http://archiveteam.org/index.php?title=Deathwatch |
| 14:41 | <beowulf> | if i had a gecko based html5 rendering bug, where'd be the best place to get help with defining it? |
| 14:43 | <jgraham> | beowulf: That might depend on exatly what you mean by a "gecko based HTML5 rendering bug" |
| 14:45 | <beowulf> | jgraham: yeah, i was pretty sure that I'd get the description of what i want wrong too :) |
| 14:46 | <Philip`> | This is a good place to get help with defining the description of what your problem is :-) |
| 14:46 | <beowulf> | i have nested lists which contain <a href=""><h2>foo</h2><p>bar</p></a> |
| 14:48 | <Philip`> | That sounds potentially problematic |
| 14:48 | <beowulf> | when this list gets repeated a few times the dom on one of these is different for one of the li's near the end on ff & camino |
| 14:48 | <beowulf> | i've tried reducing it here http://qa.getexceptional.com/test.html |
| 14:48 | jgraham | guesses that camino is using an older version of gecko or something |
| 14:49 | <beowulf> | what happens is the <a> closes and reopens beofre and after each of the elements it wraps |
| 14:49 | <beowulf> | but only on one, and only when i make the list over a certain size |
| 14:50 | <beowulf> | in that test you'll see one set of elements bordered, this is the problem |
| 14:50 | <Philip`> | Hmm, they all look the same to me |
| 14:50 | <beowulf> | i hope you see it anyway |
| 14:50 | <Philip`> | (in Firefox 2) |
| 14:51 | <beowulf> | we're using ff3.02 |
| 14:51 | <beowulf> | and camino 1.6.7 |
| 14:51 | Philip` | hypothesises that it's a packet-boundary issue |
| 14:52 | <beowulf> | Philip`: explain more, as it does seem limited to an ec2 instance |
| 14:52 | <Philip`> | Ah, I see the bordering in FF3 |
| 14:54 | Philip` | checks in Wireshark |
| 14:54 | <Philip`> | Definitely looks like a packet-boundary issue :-) |
| 14:55 | <Philip`> | The HTTP response is split into two packets, and the boundary happens to be inside one of the <a>s |
| 14:56 | <Philip`> | and Firefox's parser does something weird which means it ends up closing various elements, before receiving the rest of the HTML in the next packet |
| 14:56 | <Philip`> | If you reload from cache, or if the packet boundary is outside the <a>, then you won't experience that bug |
| 14:57 | <Philip`> | (By "inside the <a>", I mean somewhere inside the content of the element, not inside the actual tag itself) |
| 14:57 | <beowulf> | which is fine if you're not putting a background on the <a> |
| 14:58 | <Philip`> | Putting <a> around block elements is probably a bad idea in general, given how Firefox parses such things |
| 14:58 | <Philip`> | but I think(?) that doing <a><span>...stuff with block elements...</span></a> will trick it into working sensibly |
| 14:59 | <beowulf> | Philip`: cool stuff, thanks for that |
| 15:00 | <Philip`> | hsivonen: Has someone already suggested that the validator should warn when you put block elements in <a> because it confuses things like Firefox? |
| 15:00 | <Philip`> | beowulf: Oh, and this is probably the best place to ask such questions :-) |
| 15:01 | <beowulf> | http://meyerweb.com/eric/thoughts/2009/01/02/an-event-apart-and-html-5/ # tell meyer too :) |
| 15:01 | <Rik|work> | Philip`: why the validator should warn ? If it's allowed by the spec then it shouldn't warn |
| 15:01 | <beowulf> | Philip`: noted :) |
| 15:02 | <Philip`> | beowulf: Hmm, I read that when it was posted but didn't realise it was probably this packet-boundary issue |
| 15:02 | <hsivonen> | Philip`: I don't remember. However, there's this new parser thing in the works for Firefox. :-) |
| 15:02 | jgraham | maintains that the <a>-around-elements thing is a silly spec hack |
| 15:02 | <Philip`> | hsivonen: Doesn't help people like beowulf or meyer who are trying to write web pages that work today, and encounter crazy impossible-to-debug bugs |
| 15:03 | <jgraham> | s/elements/block-level-elements |
| 15:03 | <beowulf> | Philip`: what meyer is doing wouldn't neccessarily expose it, i don't think |
| 15:03 | <beowulf> | it was obvious to me because the <a>'s were styled as blocks with height and a background image |
| 15:03 | <beowulf> | (in my work) |
| 15:04 | <Philip`> | Rik|work: Because it helps inform people who try using the new HTML5 feature/hack of putting <a> around block elements and discover that it doesn't quite work as expected |
| 15:04 | <Rik|work> | Philip`: then you'll have to warn for all the features that are not yet implemented |
| 15:04 | <Rik|work> | or badly implemented |
| 15:05 | <Philip`> | Rik|work: They're usually pretty obvious and easy to understand, when a feature is just missing or really broken; whereas the <a> parsing thing is highly non-obvious to anybody who hasn't read Hixie's blog post from several years ago discussing the issue :-) |
| 15:05 | <Philip`> | beowulf: He says "What I didn’t want, though, was the randomized layout weirdness that resulted once I started styling the descendants of the link. Sometimes everything would lay out properly, and other times the bits and pieces were all over the place. I could (randomly) flip back and forth between the two just by repeatedly hitting reload." so it sounds like he exposed that bug |
| 15:06 | <beowulf> | Philip`: i should start reading the prose as well as the html :) |
| 15:06 | <jgraham> | We should really avoid new features that require validator warnings |
| 15:14 | <Philip`> | Oh, the <a><span>block</span></a> trick is invalid HTML5 :-( |
| 15:17 | <Philip`> | jgraham: But the warnings could be removed in maybe five years from now, and then we'd have a useful widely-deployed feature that people could use, whereas five years from now we won't want to still be changing the definition of valid HTML5 |
| 15:18 | <Philip`> | (and we'll all have got bored with HTML and moved on to other areas, so there won't be an HTML6 that we could add the feature to) |
| 16:32 | <Philip`> | Hmm, http://www.dotnetdotcom.org/ is now providing 600,000 pages via Amazon S3 |
| 16:33 | <Philip`> | and transferring data between S3 and EC2 is free, so it could be sensible to run analyses on EC2 |
| 16:34 | <Philip`> | but if it's only 3.2GB I could just download it onto my own computer, which has much more CPU power than sensibly-priced EC2 instances |
| 16:36 | <Philip`> | Not much point using the torrent since there's only one seed... |
| 16:37 | <Philip`> | (and no other peers) |
| 16:40 | <Philip`> | Looks like it's restricted to text/html files |
| 16:46 | <beowulf> | Philip`: any light reading you could recommend that might help me understand packet boundary issues? |
| 16:49 | <Philip`> | beowulf: Not quite sure what you're looking for - packets in general, or their effect on Firefox parsing of HTML, or something? |
| 16:51 | <beowulf> | Philip`: yeah, maybe I mean a "how a browser works" kind of read |
| 16:54 | <Philip`> | beowulf: Hmm, still not quite sure what you're looking for :-) |
| 16:55 | <Philip`> | The weirdness in Firefox isn't documented anywhere as far as I'm aware, but it's mentioned in http://ln.hixie.ch/?start=1138169545&count=1 |
| 16:55 | <Philip`> | Packets in general are just like http://en.wikipedia.org/wiki/Packet_(information_technology) (there are probably much better references but I don't know of any in particular) |
| 17:00 | <beowulf> | Philip`: thanks for the hixie.ch link (packets in general is fine) |
| 17:04 | <Philip`> | gzip is slow :-( |
| 17:04 | <Philip`> | particularly when it's a 15GB file that's been gzipped |
| 17:04 | Philip` | supposes he should split it into smaller chunks for future processing |
| 17:05 | <Philip`> | (to benefit from multi-core goodness) |
| 17:16 | <gsnedders> | w00t |
| 17:17 | <gsnedders> | just had Flash crash my browser for the second time in as many days |
| 17:23 | Philip` | tries grepping the dotnetdotcom data |
| 17:23 | <Philip`> | Hmm, quite a lot of rel="canonical" |
| 17:56 | <gmiernicki> | fuck flash... <audio> <video> is teh future! |
| 17:56 | <gmiernicki> | uh, <canvas> too :) |
| 17:57 | <gmiernicki> | actually kinda wondering lately if video and canvas tags can be combined |
| 17:57 | <gmiernicki> | perhaps embedding a <video> in a <canvas> |
| 17:57 | <Rik|work> | gmiernicki: you can in firefox |
| 17:57 | <Rik|work> | paul rouget did some nice demos |
| 17:58 | <gmiernicki> | interesting, ill see if i can find em |
| 18:04 | <gmiernicki> | amazing |
| 18:04 | <gmiernicki> | http://people.mozilla.com/~prouget/demos/DynamicContentInjection/play.xhtml |
| 18:04 | <gmiernicki> | just what i was wondering ;D |
| 18:04 | <gmiernicki> | drawImage() is more powerful than i thought |
| 18:31 | <Philip`> | Based on data from presumably quite recently, rel="canonical" is approximately the 33rd most common rel/rev value |
| 18:33 | <Philip`> | hsivonen: Hmm, I guess it's bad if I get java.nio.BufferUnderflowException when calling nu.validator.htmlparser.sax.HtmlParser.parse? |
| 18:35 | Philip` | hopes he's not just doing something stupid himself |
| 18:52 | <Philip`> | hsivonen: http://philip.html5.org/misc/gumpu-rundiary.txt |
| 18:52 | <Philip`> | using htmlparser-1.2.0 |
| 18:52 | <Philip`> | and new HtmlParser(XmlViolationPolicy.ALLOW); |
| 18:53 | <Philip`> | I get: |
| 18:53 | <Philip`> | Exception in thread "main" java.nio.BufferUnderflowException at java.nio.Buffer.nextGetIndex(Buffer.java:474) at java.nio.HeapByteBuffer.get(HeapByteBuffer.java:117) at nu.validator.htmlparser.io.HtmlInputStreamReader.read(HtmlInputStreamReader.java:297) |
| 18:53 | <Philip`> | etc |
| 18:53 | <Philip`> | (If I make the file any smaller then the problem seems to go away) |
| 19:09 | <Philip`> | (That error came from gumpu.rundiary.co.kr/content.asp?menu=550&eid=6136&title= ) |
| 19:09 | <Philip`> | (I get the same on www.aladdin.co.kr/shop/wproduct.aspx?ISBN=8956743339 ) |
| 19:09 | <Philip`> | (I blame Korea) |
| 19:26 | <aboodman> | slightlyoff: interesting thread about EventTarget |
| 19:26 | <aboodman> | I'm not sure I get the value of doing this natively though |
| 19:26 | <aboodman> | poo |
| 19:26 | <slightlyoff_> | aboodman: eh, the proposal that arv made doesn't go as far as I'd lke |
| 19:26 | <slightlyoff_> | still here = ) |
| 19:27 | <slightlyoff_> | basically I'd like all methods on objects to be listenable by default |
| 19:27 | <aboodman> | is it just another example of something like bind() that should have been there from the beginning,or is there mroe? |
| 19:27 | <aboodman> | basically, is there something here that isn't possible in js |
| 19:27 | <slightlyoff_> | instead of needing to create/implement EventTarget objects/interfaces |
| 19:27 | <slightlyoff_> | aboodman: no, but I think that's the wrong test these days |
| 19:27 | <aboodman> | sure, just trying to understand the proposal :) |
| 19:28 | <aboodman> | not arguing against |
| 19:28 | <slightlyoff_> | aboodman: we've gone a decade on "the minimum you could theoretically get away with" |
| 19:28 | <slightlyoff_> | I'd like us to start working up the food chain to "something you might actually want to use" |
| 19:28 | <aboodman> | it helps for the proposal to clarify whether this is a new fundamental capability though, think. |
| 19:28 | <slightlyoff_> | it's a unification |
| 19:28 | <aboodman> | one new fundamental capability i think ... someobdy .. should think about |
| 19:28 | <aboodman> | weak references |
| 19:28 | <slightlyoff_> | yes |
| 19:28 | <aboodman> | an event system w/o weak references kinda sucks |
| 19:28 | <slightlyoff_> | agreed |
| 19:28 | <slightlyoff_> | I'd like that a lot |
| 19:29 | <slightlyoff_> | and JS needs weakrefs anyway |
| 19:29 | <slightlyoff_> | else you leak badly |
| 19:29 | <aboodman> | yes, that's what i mean |
| 19:29 | <slightlyoff_> | just by dint of doing stuff that allocates |
| 19:29 | <aboodman> | separately from this proposal, a weak ref proposal would be useful |
| 19:29 | <aboodman> | we can already implement this very easily in v8 |
| 19:29 | <slightlyoff_> | I was thinking that = ) |
| 19:29 | <slightlyoff_> | just a different kind of pointer wrapper = ) |
| 19:29 | <aboodman> | i mean v8 already has native weak refernces |
| 19:29 | <aboodman> | i don't know about the other engines |
| 19:30 | <aboodman> | i assume they must have something similar too |
| 19:30 | <Philip`> | http://philip.html5.org/data/rel-rev-200904.txt - now with some rel=canonical |
| 19:31 | <jgraham> | Google code is getting mercurial! Awesome |
| 19:31 | <takkaria> | orly? |
| 19:31 | <takkaria> | that's interesting |
| 19:31 | <Philip`> | Not Git? |
| 19:31 | Philip` | vaguely remembers that Sourceforge added both |
| 19:32 | <Philip`> | (which is strange because it took them five years to add Subversion support) |
| 19:32 | <Philip`> | (Warning: "five" is a totally made up number and may not reflect reality) |
| 19:33 | <slightlyoff_> | Philip`: didn't Google Code *start* on SVN? |
| 19:33 | <aboodman> | yes. |
| 19:33 | <slightlyoff_> | oh, right, misread |
| 19:33 | <aboodman> | the backend wasn't svn though |
| 19:34 | <Philip`> | slightlyoff_: I meant Sourceforge which was stuck with CVS until long after it had gone out of fashion :-) |
| 19:34 | <slightlyoff_> | Philip`: yeah, my bad |
| 19:35 | <Philip`> | http://google-code-updates.blogspot.com/2009/04/mercurial-support-for-project-hosting.html - "Our implementation of Mercurial is built on top of Bigtable, making it extremely scalable and reliable just like our Subversion on Bigtable implementation." - they wrote their own Mercurial and Subversion servers? |
| 19:36 | <Philip`> | (or new storage backends for the existing server code or something?) |
| 19:36 | <aboodman> | Philip`: basically; they implement the protocols |
| 19:36 | <slightlyoff_> | Philip`: you didn't think they were going to get to high availability on the back of fsfs, did you? |
| 19:38 | Philip` | has only had FSFS get horribly corrupted once in several thousand commits |
| 19:57 | <gsnedders> | http://www.flickr.com/photos/gsnedders/3470720119/ — oh noes! it's mes! |
| 20:04 | <gsnedders> | Is it bad that just having annevk link to my site caused a notable spike in traffic? |
| 20:31 | <Philip`> | gsnedders: I get notable spikes in traffic just from being linked from comments in blogs |