| 00:00 | <hober> | yay! |
| 00:01 | hober | wonders if that'll speed up the release of the msft review |
| 00:31 | <webben> | Vaguely OT question: Does anyone have an opinion on the best way of declaring a XML syntax for XHTML5-like languages? Is Relax NG capable of declaring a syntax of that sort of complexity, or would one need schema? |
| 00:32 | <Hixie> | what do you mean by "declaring a XML syntax"? |
| 00:33 | <webben> | Hixie: What elements can be contained by what. What content is allowed in attributes. |
| 00:33 | <webben> | yeah basically nesting rules + data types. |
| 00:33 | <Hixie> | i prefer to use english prose to define that |
| 00:33 | <webben> | Hixie: I'm talking about for validation purposes. |
| 00:33 | <hdh> | validator.nu is using RN, the files are in http://svn.versiondude.net/whattf/syntax/ |
| 00:34 | <Hixie> | oh, for validatior |
| 00:34 | <Hixie> | validation, even |
| 00:34 | <webben> | hdh: Ah, interesting. Thanks! :) |
| 00:34 | <hdh> | I added them to nxml-mode, but got stuck with w: datatype lib |
| 00:34 | <Hixie> | which schema language is best probably depends on the specific rules of the language |
| 00:34 | <Hixie> | personally i'd just implement the whole thing in code, but that's just me :-) |
| 00:37 | webben | isn't sure what the code for that would look like... but that might be an option further down the road. |
| 00:40 | <Philip`> | You could invent your own schema language to cope with HTML5's requirements |
| 00:40 | <Philip`> | (Warning: may not be trivial) |
| 00:40 | <Hixie> | it could be really trivial |
| 00:41 | <Hixie> | the schema language could be designed so that the empty string represents the html5 grammar, and then the schema would just be "". |
| 00:41 | <Philip`> | Specifying that schema language is hard, though |
| 00:41 | <Hixie> | Specifying that schema language is my day job. |
| 00:42 | <inimino> | Philip`: not if you can use HTML5 as a normative reference |
| 00:44 | <Philip`> | Specifying that schema language well (i.e. built up in a formal way from primitive components) is still hard :-) |
| 00:44 | <Hixie> | implementing the schema language is hard, yes |
| 00:45 | <Hixie> | that's hsivonen's day job :-) |
| 00:45 | <inimino> | hehe true |
| 00:57 | <Hixie> | sweet kittens. if it's not my site that's down it's the w3c's. |
| 00:57 | <Hixie> | how am i supposed to get any work done. |
| 00:57 | <hdh> | CSS to emulate irrelevant="" would be {width: 0; height: 0; visibility: hidden;}, right? |
| 00:58 | <Hixie> | [irrelevant] { display: none; } |
| 00:58 | <hdh> | ok, thanks |
| 00:58 | <kingryan> | Hixie: w3.org appears to work for me |
| 00:59 | <Hixie> | lists.w3.org was down briefly |
| 00:59 | <Hixie> | it's back up now |
| 00:59 | <kingryan> | ah |
| 01:05 | gsnedders | needs some soul searching help |
| 01:06 | <Hixie> | Lachy: someone's talking about you here http://etfb.livejournal.com/83307.html |
| 01:07 | <gsnedders> | LiveJournal! yay! :P |
| 01:15 | <Hixie> | holy. war. on. kittens. |
| 01:15 | <Hixie> | now my pubrules checking web service host is down. |
| 01:19 | <Philip`> | In the future, we will be able to download applications and run them on our own computers, which will solve all these downtime problems |
| 01:20 | gsnedders | points at what Eric Smit (forgive my spelling) said a few days ago: 90% of everything can be done online |
| 01:21 | Philip` | finds it's more useful to do 100% of something, rather than getting stuck just before the end because someone else broke something outside your control |
| 01:21 | <Hixie> | Philip`: there are so many reasons that wouldn't work... |
| 01:43 | csarven | thinks <aside> is a bad name |
| 01:44 | gsnedders | moves csarven aside |
| 01:44 | <csarven> | good one :) |
| 01:44 | <Hixie> | csarven: what do you propose instead? |
| 01:45 | <csarven> | Hixie <section> would have been sufficient imo |
| 01:45 | <Hixie> | <section> implies you should read it |
| 01:46 | <Hixie> | i wanted something for, e.g., the examples and notes in the spec |
| 01:46 | <csarven> | what to read is subjective |
| 01:46 | <Hixie> | <section> implies its contents are an intergal, inline, part of its container |
| 01:47 | <csarven> | id="sidebar" -> <aside> -- im not sure if this is really better. id="sidebar" -> id="supplemental_information" would have been okay and perhaps <supplemental> |
| 01:47 | <csarven> | <aside> indicates some sort of a physical direction |
| 01:48 | <csarven> | (imo) |
| 01:48 | <Hixie> | supplemental is harder to type |
| 01:48 | <othermaciej> | <aside> sounds less presentational than <sidebar> |
| 01:48 | <hdh> | something for the footnotes |
| 01:48 | <othermaciej> | but it also sounds like you'd use it for inline parenthetical remarks |
| 01:48 | <gsnedders> | I wanted to work on HTML 5. You probably don't care, but I was away working for money instead, getting paid to drop Ogg. |
| 01:48 | <Philip`> | <blockaside> would stop it sounding like it's inline |
| 01:48 | <gsnedders> | (where the latter sentence is an aside) |
| 01:48 | <hdh> | and longer to say |
| 01:48 | <gsnedders> | My former maths teachers was always talking about asides |
| 01:48 | <gsnedders> | how about… <sidenote>? |
| 01:48 | <Hixie> | <aside> is fine |
| 01:48 | <gsnedders> | or isn't that what you're meaning? |
| 01:49 | <csarven> | Hixe <supp> or how about <wazzup> :P |
| 01:49 | <hdh> | inline aside can use small? |
| 01:49 | <csarven> | Hixie |
| 01:49 | <hdh> | oops, remember it wrong |
| 01:49 | <gsnedders> | Hixie: still livin' high the life after you pay-off from MS/Apple/Nokia? :P |
| 01:49 | <Philip`> | <div lessrelevant> |
| 01:50 | <gsnedders> | <div gsneddersisawesome> |
| 01:50 | <othermaciej> | wait, I thought Google was bribing Apple |
| 01:50 | <othermaciej> | to stop XHTML2 so that AdSense can keep working |
| 01:50 | <othermaciej> | did we renegotiate the bribe agreement? |
| 01:50 | <gsnedders> | othermaciej: I thought they needed to stop XHTML totally, which they already failed at. |
| 01:51 | <csarven> | Hixie to me <supp> (short for supplemental) or <aux> (short for auxiliary) |
| 01:51 | <othermaciej> | I think <sidebar> would not be so bad |
| 01:51 | <othermaciej> | that's what they are called in print |
| 01:51 | <othermaciej> | even when not literally on the side |
| 01:52 | <othermaciej> | but I don't care about the name that much |
| 01:52 | <csarven> | <aside> really comes across as something that exists in a physical space. to me HTML documents don't (or perhaps shouldn't) convey that information |
| 01:52 | <csarven> | <sidebar> is worse imo because again that suggests a location |
| 01:52 | <hdh> | should footnotes be inside <aside> too |
| 01:53 | Hixie | mumbles something about giving you all cans of paint so you can paint the bikeshed yourselves :-P |
| 01:53 | <csarven> | supplemental information (that we usually see on the side as a sidebar) can be located anywhere.. it could be before or below the document even |
| 01:53 | <csarven> | Hixie oh c'mon. just reasoning out. i'm willing to hear the problem with my suggestion |
| 01:53 | <othermaciej> | http://en.wikipedia.org/wiki/Aside |
| 01:53 | <othermaciej> | it's not about physical space |
| 01:53 | <Hixie> | csarven: i don't see the problem with <aside> |
| 01:54 | <gsnedders> | Hixie: can I have pink and blue, so I can paint my bikeshed pink with blue spots? |
| 01:55 | gsnedders | waves g'nite before he can get his paint (just throw it at him after he leaves) |
| 01:56 | <csarven> | Hixie lets start here then: do you think <aside> is more accurate then <supplemental> ? |
| 01:56 | <Hixie> | mildly so, but not especially, no |
| 01:56 | <Hixie> | there are dozens of words that would be equally good |
| 01:57 | <Hixie> | <sidebar>, <aside>, <comment>, <note>, <supplemental>, etc |
| 01:57 | <Hixie> | some have problems (e.g. <comment> is display:none in some old versions of IE) |
| 01:57 | <Hixie> | (and supplemental is too long) |
| 01:58 | <Hixie> | but at the end of the day, they're not especially better or worse than each other |
| 01:58 | <othermaciej> | the name is not terribly important |
| 01:58 | <hdh> | aux is a nice fit for Dvorak, supp too :) |
| 02:00 | csarven | wonders if <supplemental> would be the longest tag out there |
| 02:00 | <othermaciej> | on the one hand, the time to fine-tune tag names is drawing short (as browsers start implementing); but on the other hand, it's not clear that it would be the best use of time to do so |
| 02:00 | <othermaciej> | unless there's an actual practical problem with specific element names |
| 02:00 | kingryan | is laughing at "Mr "Burn All Oggs" himself" |
| 02:01 | <Hixie> | the names were fine tuned months ago |
| 02:01 | <Hixie> | (<aside> was originally <sidebar>, <time> originally <t>, etc) |
| 02:01 | <csarven> | im glad <sidebar> changed to <aside> |
| 07:22 | <Hixie> | http://www.petitiononline.com/lortow3/petition.html <-- one of the most pointless petitions ever |
| 07:25 | <Hixie> | the text was removed because people said they'd ignore it... but sure... let's put it back... i'm sure they won't ignore it then! |
| 07:40 | <hsivonen> | <aside> WFM |
| 07:57 | <Hemebond> | ? |
| 07:57 | <Hemebond> | Why would WHATWG define a codec requirement? |
| 07:58 | <Hixie> | Hemebond: to encourage convergence on a common codec |
| 07:58 | <Hixie> | so that people don't have to encode everything two or more times |
| 07:58 | <hober> | Hemebond: it's required in order for <video> to actually be interoperable |
| 08:35 | <Hemebond> | But... |
| 08:35 | <Hemebond> | hmm |
| 08:35 | <Hemebond> | Isn't the web suppose to be dumb? |
| 08:38 | <hsivonen> | the IP layer is dumb. the Web is lazy. |
| 09:09 | <othermaciej> | any Opera folks around? |
| 09:42 | <hsivonen> | on case anyone is wondering, I am aware that Validator.nu is in a semi-broken state. |
| 09:43 | <hsivonen> | (I should have had the common sense not to install kernel updates right now...) |
| 10:06 | <Hemebond> | Hey.... question. |
| 10:07 | <Hemebond> | XHTML is modular... can be extended. Could browser makers not create their own modules to add in stuff they wanted? Kind of like extensions in OpenGL? |
| 10:08 | <Philip`> | HTML can be extended without being modular or namespaced - see e.g. Apple adding elements like <canvas> |
| 10:09 | <hsivonen> | Hemebond: the modularity is in the dtd and has nothing to do with browsers |
| 10:09 | <hsivonen> | canvas did not need modularization to happen |
| 10:09 | <Hemebond> | Except that it muddies the markup with elements/tags that don't exist in the DTD. No? |
| 10:09 | <Philip`> | Hemebond: That's easy to solve by just ignoring the DTD :-) |
| 10:12 | <Philip`> | If you add an extension to (X)HTML by any method, it's not going to be in the standard DTD, and there is not much value in creating a new non-standard DTD with the extension because that doesn't solve any useful problems except for letting DTD-based validators claim your non-standard document is valid |
| 10:15 | <Hemebond> | I thought you could... like namespaces and such. Like mixing in <svg:blah> and such. |
| 10:16 | <Hemebond> | I guess that's not a module/extension. |
| 10:16 | <Hemebond> | If browser makers created a seperate namespace for their extensions. |
| 10:16 | <Hemebond> | A bit like CSS does. |
| 10:17 | <Philip`> | If we had <apple:canvas>, and people started using it, what would happen when Mozilla started supporting the same feature? |
| 10:18 | <Hemebond> | What happens in CSS when more than one browser supports an extension? |
| 10:19 | <webben> | Philip`: What would happen? They could implement <apple:canvas> (indeed in XHTML5 one could write apple:canvas, since that's a prefix not a namespace) |
| 10:20 | <Hemebond> | And if it ever became a part of the proper "standard", it would move to the XHTML namespace. |
| 10:20 | <Philip`> | Hemebond: People write "-moz-border-radius: 5px; -webkit-border-radius: 5px; border-radius: 5px;" which isn't very pretty and favours certain browsers and will still break if the official CSS definition of border-radius is changed in the future |
| 10:20 | jgraham_ | wonders how likely MS would be to implement something in the http://apple.com/ns/ namespace |
| 10:21 | <jgraham_> | (that NS URI is made u of course) |
| 10:21 | krijn | just wrote -moz-border-radius: 10px 0 0 10px; -webkit-border-top-left-radius: 10px; -webkit-border-bottom-left-radius: 10px; |
| 10:21 | <webben> | Philip`: If they only added border-radius when finalized, it wouldn't break |
| 10:22 | <Philip`> | webben: It seems that people don't want to wait for that, because they want their current pages to work in future CSS3 UAs, e.g. in http://virtuelvis.com/archives/2004/11/imageless-rounded-corners |
| 10:22 | <webben> | jgraham: As likely as they are to implement an Apple invention like canvas in the first place. Because of exceptional market demand. (So far not much sign of that, and they've got Silverlight to push...) |
| 10:22 | <webben> | Philip`: Yes... but that's a problem with extensions that aren't "namespaced" not extensions that are. |
| 10:23 | <webben> | Or a problem with people using experimental drafts. |
| 10:23 | <webben> | it's a different problem. |
| 10:23 | <jgraham_> | webben: You think? I would have thought that they would be much less likely to implement something that required their competitors name to be associated with a useful feature |
| 10:24 | <Hemebond> | They would put it in their own namespace. |
| 10:24 | <webben> | jgraham_: Apple's name is associated with canvas. Namespaced or not. |
| 10:24 | <Hemebond> | Each browser would put it in their own namespace. |
| 10:26 | <Philip`> | webben: It's a problem with vendor-namespaced extensions that are later standardised into differently-namespaced extensions, because people will want to use the differently-namespaced extension to get forward-compatibility, even though they can only test pre-standard implementations, and so it limits one of the benefits of using vendor-namespaced extensions |
| 10:26 | <webben> | Hemebond: have nested <apple:canvas><microsoft:canvas><firefox:canvas><camino:canvas> could make CSS selection and DOM manipulation triky. |
| 10:26 | <Philip`> | webben: That would make Opera users unhappy too :-) |
| 10:26 | <jgraham_> | webben: Really? By what fraction of the web authoring population? How often is it drummed in? Do you think the association between netscape and <script> is stronger or weaker than it would have been if it had been called <netscape:script>. What about Microsoft and CSS? |
| 10:27 | <webben> | Philip`: Yes... but removing vendor-namespaced extensions makes the problem worse, because /everyone/ is forced to use differently-namespaced extensions that break. |
| 10:27 | <Hemebond> | webben: That's what happens when you implement non-standard extensions. |
| 10:29 | <Hemebond> | http://www.opengl.org/resources/features/OGLextensions/ |
| 10:30 | <Philip`> | I guess vendor-namespaced extensions work well when nobody uses that extension, so the trick is to make sure it doesn't become popular until after it's been standardised |
| 10:30 | <webben> | jgraham_: Unknown about current knowledge. I agree the association would be stronger. Not sure about the CSS reference. I still think that if MS is rabidly avoiding other companies' tech it will only implement out of market demand: and that includes canvas. Having Silverlight makes an implementation much less likely. |
| 10:31 | <webben> | jgraham_: In other words, yes it would make a difference but probably not enough of a difference to be a deciding factor. |
| 10:31 | <Philip`> | I have a vague memory that Vista's GL implementation supports some APPLE_* extensions, but I might be totally wrong about that and I'm too lazy to check right now |
| 10:32 | <Philip`> | (NVIDIA definitely has some e.g. ATI_* extensions, though) |
| 10:32 | <Hemebond> | Windows has no OpenGL implementation. Implementations are provided by ICD drivers. |
| 10:33 | <Philip`> | Windowses up to XP had a software renderer for OpenGL 1.1, and Vista has GL 1.4 implemented on top of DirectX, if I remember correctly |
| 10:34 | <webben> | http://msdn2.microsoft.com/en-us/library/ms537767.aspx : "The Microsoft implementation of OpenGL for the Windows operating system is industry-standard graphics software with which programmers can create high-quality still and animated three-dimensional color images." |
| 10:34 | <Hemebond> | Ah, you're right. |
| 10:36 | <Philip`> | http://www.delphi3d.net/hardware/viewreport.php?report=1566 - hmm, maybe that's the Vista one, and there's no Apple :-( |
| 10:37 | Philip` | blames his memory |
| 10:44 | <webben> | It's notable that Microsoft's aversion to other brands didn't stop them calling IE "Mozilla". |
| 10:44 | <webben> | (in the UA string) |
| 10:45 | <Hemebond> | Haha yeah. |
| 10:46 | <Hemebond> | But most users would never see that so they're fairly safe. |
| 10:47 | <webben> | I doubt the most users view source. |
| 10:47 | <webben> | *that |
| 10:53 | <Philip`> | Most people who saw the IE UA string wouldn't know what "Mozilla" meant |
| 10:53 | <Philip`> | (back when it was adopted, when it was just an internal codename used by Netscape) |
| 10:54 | <Philip`> | or at least I had no idea what it meant for years after I first saw it :-) |
| 10:54 | <webben> | they do now |
| 10:54 | <Philip`> | IE can't change it now because it'd break compatibility |
| 10:54 | <webben> | yes, so it's still a demonstration of the power of the market |
| 10:54 | <Philip`> | regardless of how much they like/dislike advertising their competitors hundreds of billions of times a day |
| 10:55 | <webben> | Oh I'm sure they like nothing better ;) |
| 10:56 | <othermaciej> | Hixie likes to tell people they should make their user agent way simpler |
| 10:56 | <Hemebond> | I thought they had removed it. |
| 10:57 | <othermaciej> | but it would break a lot of content |
| 10:57 | <othermaciej> | this is Safari's latest UA string: |
| 10:57 | <othermaciej> | Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10_5; en-us) AppleWebKit/525.1+ (KHTML, like Gecko) Version/3.0.4 Safari/525.1 |
| 10:58 | <othermaciej> | it claims to be three different browsers that it actually isn't |
| 10:58 | <Philip`> | HTTP_USER_AGENT = Opera/9.24 (X11; Linux i686; U; en) |
| 10:58 | <othermaciej> | some other WebKit-based browsers include both Safari and their own name, so they get up to claiming 4 browsers that they are not |
| 10:59 | <othermaciej> | at the time we developed it, a lot of the werid details really mattered for compatibility |
| 10:59 | <othermaciej> | Opera does indeed have a nice clean one |
| 11:00 | <Philip`> | You should solve all your compatibility concerns by having a DOCTYPE switch, and only sending the old ugly UA string to pages written before the switch |
| 11:00 | <othermaciej> | has it always had that, or is that a recent change? |
| 11:00 | <othermaciej> | but the UA string goes to the web server |
| 11:00 | <othermaciej> | doctype switch is too late to know which to send in the http request |
| 11:00 | <Hemebond> | Are there any web browsers that actually do everything properly? |
| 11:00 | <Hemebond> | No hacks whatsoever, |
| 11:01 | <Hemebond> | Supports the specs correctly. |
| 11:01 | <Philip`> | I think it was Opera 9 that stopped sending the Mozilla string to everyone |
| 11:01 | <Hemebond> | No "error correction" and such? |
| 11:01 | <Philip`> | Hemebond: And no bugs? |
| 11:01 | <krijn> | http://www.user-agents.org/index.shtml?n_s |
| 11:01 | <Hemebond> | One that has a proper SGML parser. |
| 11:02 | <othermaciej> | Hemebond: I'm pretty sure the answer is no |
| 11:02 | <Hemebond> | lame |
| 11:02 | <Philip`> | I imagine a project to make a 'correct' browser would die out quite quickly due to a lack of users |
| 11:02 | <webben> | Hemebond: I don't think any maintained browser interprets <br /> in text/html as <br>> |
| 11:02 | <othermaciej> | there's a lot of specs, supporting them 100% correctly and with no error handling or extensions would lead to a browser that can't handle many important web sites |
| 11:03 | <webben> | Hemebond: (which is what SGML would require) .... so no. |
| 11:03 | <webben> | Hemebond: Lynx's strict SGML mode might be close though. But Lynx doesn't support all of HTML 4.01 (no one does) |
| 11:03 | Philip` | wonders what would happen if browsers downloaded DTDs, and a billion users started accessing www.w3.org every day |
| 11:03 | <othermaciej> | it's not clear what it would even mean to support all of HTML 4.01 |
| 11:04 | <webben> | Philip`: Why would they download a DTD rather than using a local copy. |
| 11:04 | <Philip`> | webben: Are DTDs allowed to change? |
| 11:04 | krijn | wonders if he should use <!doctype html> on the to-be-launched-today Fronteers website :/ |
| 11:04 | <webben> | Philip`: They probably shouldn't be. |
| 11:04 | <krijn> | People will hate me if I do, and others will hate me if I don't |
| 11:05 | <Philip`> | krijn: Some people will love you regardless of what you do, so it doesn't matter that much :-) |
| 11:05 | <webben> | Philip`: Although they do occasionally have bugs, which might that problematic. |
| 11:05 | <webben> | Philip`: It hardly makes sense to check on a daily basis, anyhow. |
| 11:06 | <krijn> | Philip`: they don't care about which sites I build though ;p |
| 11:06 | <webben> | Philip`: I guess they could also set some long-term expiration headers. |
| 11:07 | <webben> | firefox addons doesn't seem to go down from Fx constantly checking for updated extensions. |
| 11:08 | <Hemebond> | Don't extensions check their own sites for updates? |
| 11:08 | <webben> | Hemebond: A lot of extensions are hosted on addons. |
| 11:08 | <webben> | AFAIK |
| 11:08 | <Philip`> | Firefox addon update messages are smaller than DTDs, and the addon system was designed for lots of users, e.g. by having its own domain name rather than sharing with the rest of www.mozilla.org |
| 11:09 | <webben> | it was designed a lot later than 1999 ;) |
| 11:09 | <webben> | Philip`: You should only have to do a HEAD request to check for DTD updates. |
| 11:10 | <Philip`> | I remember Netscape complaining recently about the cost of sending RSS DTDs from their domain to feed readers |
| 11:10 | <webben> | Well, they actually pulled the page for a while. |
| 11:10 | <Philip`> | Do web browsers use HEAD much? (I've seen a small number of servers which handle it very buggily) |
| 11:10 | <webben> | Philip`: Dunno. They /should/ do, but there we go. |
| 11:11 | <webben> | Seems to work pretty nicely with curl whenever I try it. |
| 11:11 | <Philip`> | (One server sent four random bytes (uninitialised memory?) in the Content-Type for HEADs) |
| 11:12 | <Philip`> | http://www.louvre.fr/llv/commun/home.jsphas |
| 11:12 | <webben> | Awesome. That's because people who sit down to code web servers don't start by thinking "I should really read the spec for this HTTP nonsense." |
| 11:12 | <Philip`> | Content-Type: p#�/html;charset=UTF-8 |
| 11:13 | <webben> | (that's always the scary thing about this stuff: not that authors don't read specs, but that so many implementers don't either) |
| 11:13 | <Philip`> | Oops |
| 11:13 | <Philip`> | s/has// |
| 11:13 | <Philip`> | (in the URL) |
| 11:14 | <annevk> | and that people writing specs make mistakes too, etc. |
| 11:15 | <webben> | annevk: If implementers read the specs, they could point out the mistakes and thump the people writing the specs to fix em ;) |
| 11:16 | <webben> | All I know is that when someone starts talking about coding a web server, I shouldn't be having to point them in the direction of the HTTP spec. :) |
| 11:16 | <annevk> | this happens, but given that so far no complicated specs that made REC required two complete interoperable impl people don't review enough |
| 11:17 | <annevk> | HTTP is a good example of not being good enough |
| 11:17 | <webben> | HTTP wasn't on the same standards pipeline. |
| 11:20 | <takkaria> | where's the schedule for fpwd? |
| 11:54 | <Hemebond> | 1am. Night all. |
| 12:21 | <hsivonen> | Philip`: the *nix readme for Netscape said, IIRC, "And remember, it is spelled N-e-t-s-c-a-p-e but pronounced Mozilla." but the Mac and Windows masses didn't see that |
| 12:26 | <hsivonen> | Re: DTD discussion: http://hsivonen.iki.fi/no-dtd/ |
| 12:36 | <webben> | hsivonen: Your FAQ similarly seems to assume that it would need a GET request not a HEAD request. HEAD could still cause DOS, but it might be less likely to. |
| 12:36 | <webben> | "actually retrieving the referenced DTDs" |
| 12:37 | <webben> | rather than "actually checking if the referenced DTDs have been modified" |
| 12:38 | <webben> | And also that the HTML spec does seem to recommend not connecting to the web to retrieve DTDs (at least, not every time): http://www.w3.org/TR/html4/sgml/intro.html#h-19.2 |
| 12:38 | <webben> | (admittedly, your discussion is mainly about XML ... maybe the XML spec says something else) |
| 12:39 | <kig> | argh, implementing svg on the 2d canvas is rather challenging. no gradient/pattern transforms, no dashed strokes, no markers, lack of filters and blendmodes, no masks, no text. and no getTransform either |
| 12:39 | <webben> | kig: Why are you trying to implement SVG on canvas? |
| 12:39 | <hsivonen> | webben: if you implement your idea in the default Xerces resolver and manage to get it deployed so that the deployers don't get autobanned from w3.org, I would gladly revise my statements in face of new evidence |
| 12:40 | <webben> | hsivonen: Xerces doesn't use a local override? |
| 12:41 | <hsivonen> | webben: xerces is widely used and a piece of software where one might try out stuff like this more realistically than in browsers |
| 12:42 | <hsivonen> | webben: If I'm not mistaken, the defaults in Java XML parsers are *very* naive |
| 12:44 | <kig> | webben: to use svg images on a canvas |
| 12:44 | <webben> | kig: Oh I see. |
| 12:49 | <hsivonen> | would it good or bad if Validator.nu had a preset for XHTML 1.0 *Transitional* + Ruby + RDF + SVG 1.1 + MathML 2.0? |
| 12:49 | <hsivonen> | it seems to me that adhering to Strict XHTML content models and using embedded SVG or orthogonal issues in the real world |
| 12:50 | <webben> | hsivonen: So a bit like modular XHTML with the legacy module? |
| 12:51 | <hsivonen> | webben: yeah |
| 12:51 | <webben> | hsivonen: I should think they are more related issues in the real world than the ideal world, actually ;) |
| 12:53 | <webben> | hsivonen: I'm not really sure what the presets are for. |
| 12:54 | <webben> | What if you wanted to validate (say) XHTML 1.0 Transitional + Ruby + MathML but not RDF and SVG. |
| 12:54 | <webben> | maybe it would be better to have some tickboxes |
| 12:56 | <hsivonen> | webben: if you want to do that, the current approach is that you download the schema driver and put an edited copy on your on HTTP server |
| 12:56 | <hsivonen> | webben: probably too hard for most users |
| 12:57 | <hsivonen> | webben: but it seems to me that people want either a single language or full compound doc support |
| 12:58 | <hsivonen> | webben: although some people might actually want more restricted compound document combos, putting it all in the UI is problematic |
| 12:58 | <webben> | hsivonen: That's why I think tickboxes would be better. |
| 12:58 | <webben> | So I want Strict (tick), and Ruby (tick), or maybe just go ahead and tick All. |
| 12:59 | <hsivonen> | webben: positive checkboxes are problematic from the point of view of validator URI design :-( |
| 13:00 | <webben> | hsivonen: &components=xhtmlstrict+ruby+rdf ? |
| 13:00 | <hsivonen> | webben: in order to make the URIs simple in the common case, the checkboxes should exclude stuff |
| 13:00 | <hsivonen> | and that would be bad UI |
| 13:01 | <webben> | hmm |
| 13:02 | <webben> | hsivonen: what does the URI bit for presets look like atm? |
| 13:05 | <kig> | webben: this is how far i've gotten: http://glimr.rubyforge.org/cake/canvas.html#SVGParser |
| 13:05 | <hsivonen> | webben: horrible if you actually choose a preset but nice if you don't choose a preset and let it pick one from the root namespace |
| 13:06 | <webben> | kig: Can't you just overlay svg objects using the browser's svg renderer rather than trying to parse SVG? |
| 13:06 | <kig> | not if i want to draw them on the canvas.. |
| 13:07 | <hsivonen> | webben: the basic idea of the generic facet is that the interface is uniform regardless of what schemas you use. some schemas just are faster than others because they are precompiled in RAM and some are faster than completely remote schemas because the source bytes are stored locally |
| 13:08 | <hsivonen> | webben: the basic idea of the HTML5 facet is that there are no tweakable settings |
| 13:08 | <webben> | hsivonen: You could have tickboxes and still not have any in the uri by default. |
| 13:08 | <webben> | a missing components key could just mean "no preset" |
| 13:10 | <hsivonen> | webben: yeah, I think I have to sit down and design a checkbox UI at some point |
| 13:10 | <hsivonen> | webben: I don't know how to do it such that |
| 13:10 | <hsivonen> | 1) It's good UI |
| 13:10 | <hsivonen> | 2) The URIs are nice |
| 13:10 | <hsivonen> | 3) Users can bring their own schemas |
| 13:11 | <hsivonen> | 4) I don't need to build an insane number of schemas |
| 13:11 | <webben> | hsivonen: Schemas can be declared in the submitted documents. |
| 13:11 | <hsivonen> | 5) I don't to kill perf by generating schemas on the fly |
| 13:11 | <hsivonen> | 6) The default do the right thing for most people |
| 13:11 | <webben> | hsivonen: Well, if you modularize it, you can create all the schemas from those combinations /once/. |
| 13:12 | <hsivonen> | webben: putting schema-related syntax in the document itself is wrong and putting validation service-specific syntax there is even worse |
| 13:13 | <webben> | hsivonen: I'm just talking about schema declarations (if declarations is the right term). Nothing specific to using your validator. |
| 13:13 | <hsivonen> | s/I don't to kill perf/I don't hove to kill perf/ |
| 13:13 | <hsivonen> | s/hove/have/ |
| 13:13 | <hsivonen> | typo++ |
| 13:14 | <hsivonen> | webben: doctype and xsi:schemaLocation are bugs |
| 13:14 | <webben> | xsi:schemaLocation="http://www.w3.org/MarkUp/SCHEMA/xhtml11.xsd"; (is that called a "declaration") |
| 13:14 | <hsivonen> | webben: that's Considered Harmful |
| 13:14 | <webben> | hsivonen: Well, yes, but here you're hitting a side-effect of that principle. |
| 13:15 | <hsivonen> | yes |
| 13:15 | <webben> | the alternative is for people to throw a schema uri into their validator query: but that's just as prone to single-point failure. |
| 13:15 | <hsivonen> | I'm assuming that most people are OK with validating against the compound Web language |
| 13:16 | <hsivonen> | that is, the schema allowing SVG is not a real problem for people who want to check if they mistyped their XHTML or MathML |
| 13:17 | <webben> | No. The people who must want to do subsetting would go rather beyond what you could easily represent in a UI anyway. |
| 13:17 | <webben> | e.g. subsetting for comments on blogs |
| 13:18 | <webben> | (or indeed blog posts for that matter) |
| 13:18 | <webben> | where you might want to allow SVG but disallow SCRIPT for instance |
| 13:20 | <hsivonen> | webben: indeed. which is why the XHTML Modularization failed. and which is why a UI that doesn't involve user-written driver schemas can become too complex and still lack the options the user wants |
| 13:21 | <webben> | hsivonen: I think XHTML modularization's problems have more to do with it being utterly unclear what exactly it's for, the spec never having been finished, and schemas being a bit of a nightmare as a tool. |
| 13:22 | <webben> | (compared to the HTML DTD, navigating XHTML schemas is horrible) |
| 13:23 | <hsivonen> | webben: I think XHTML Modularization failed when their flagship use case, XHTML MP, chose a subset that doesn't follow the prescribed module boundaries |
| 13:23 | <webben> | Doesn't that suggest they got the module boundaries very wrong? |
| 13:24 | <hsivonen> | webben: yes. hence, fail |
| 13:24 | <webben> | ideally a module should be a set of related things that you would never not want together. if you have dependencies that you might or might not want, those should be in different modules. |
| 13:25 | <webben> | (i.e. modules should be about relatedness, the relationships between modules should be about dependencies) |
| 13:37 | <hsivonen> | Philip`: does X3D have a namespace URI? |
| 13:37 | <annevk> | seems it was not mentioned here yet: http://lists.w3.org/Archives/Public/www-archive/2007Dec/0072.html |
| 13:37 | <annevk> | kig, Opera 9.5 adds some SVG integration for <canvas> |
| 13:37 | <annevk> | kig, so hopefully in due course you won't need your own library anymore and you can just paint SVG graphics directly |
| 13:42 | <webben> | annevk: Does it preserve text alternatives for SVG graphics when it does so? |
| 13:43 | <webben> | or does it effectively unsemanticize the SVG much like using it as a CSS background image might? |
| 13:51 | <Philip`> | hsivonen: No, it isn't in a namespace |
| 13:51 | <Philip`> | (That has been complained about years ago, but the Web3D group works in secret, and I've got no idea if they're doing anything about it) |
| 13:52 | <hsivonen> | Philip`: ok. |
| 13:52 | <hsivonen> | Philip`: Google shows that different people are using different bogus namespaces, so I was wondering if there's a correct one... |
| 13:53 | <Philip`> | hsivonen: http://www.xj3d.org/javadoc2/overview-summary.html under "XML Integration" is the one I've seen, so I copied that into my implementation |
| 13:53 | <Philip`> | (with a compatibility transform so you can load 'legacy' non-namespaced X3D documents) |
| 13:55 | <Philip`> | http://www.web3d.org/specifications/x3d-schema-changelog.txt has some comments mentioning namespaces and suggesting xmlns:x3d="http://www.web3d.org/specifications/x3d-3.0.xsd";, which seems a really bad idea because it'll be different for every version |
| 13:56 | <hsivonen> | Philip`: should I add a UI string for http://www.web3d.org/specifications/x3d-namespace ? |
| 13:57 | <Philip`> | hsivonen: Doesn't seem worthwhile at the moment, since probably nobody uses it |
| 13:57 | <hsivonen> | Philip`: OK |
| 13:58 | <hsivonen> | (I have UI strings for various stuff that might occur on purpose or by mistake in document types roughly advertised in the presets) |
| 13:59 | <hsivonen> | (but I don't have UI strings for Atom extensions or RDF ontologies) |
| 14:00 | <Philip`> | I can only see three places on the web which mention http://www.web3d.org/specifications/x3d-namespace and one of those places is the #whatwg logs |
| 14:01 | <hsivonen> | ok |
| 14:03 | Philip` | will post something to the X3D mailing list, and will complain about the namespace and see if anyone has helpful suggestions |
| 14:03 | <Philip`> | (because I'm assuming the lack of a namespace is a bad thing) |
| 14:08 | <hsivonen> | well, if the namespace is lacking, it should be "" and people shouldn't use bogonamespaces |
| 14:09 | <hsivonen> | bogonamespaces are definitely a bad thing for interop |
| 14:09 | <Philip`> | Bogonamespaces? |
| 14:09 | <hsivonen> | Philip`: someone pulls a URI from their sleeve in order to have *some* namespace |
| 14:09 | <Philip`> | Ah |
| 14:10 | <Philip`> | But if everyone picks the same bogonamespace, that's good for interoperability, so that's alright :-) |
| 14:10 | <hsivonen> | yeah |
| 14:12 | Philip` | wonders if there are standard arguments against non-namespaced XML languages |
| 14:12 | <webben> | Philip`: Wouldn't that boil down to people using the same names for different things? |
| 14:13 | <webben> | Philip`: Also note the argument that in order to change functionality for the same names, you must use a new namespace (e.g. WHATWG critique of XHTML 2 WG's decision to reuse XHTML1's namespace for different functionality) |
| 14:13 | <webben> | in other words, used right, namespaces preserve interoperability and compatibility |
| 14:14 | <Dashiva> | Isn't the latter rather circular? |
| 14:14 | <Dashiva> | Namespaces must be separate to avoid collisions; collisions are being made, therefore we must split namespace |
| 14:14 | <webben> | Dashiva: No. Because it's ultimately grounded in giving clients a way to distinguish two names with similar but different functionality. |
| 14:14 | <hsivonen> | case studies RSS 2.0 (which you probably don't want to use in any argument about anything) and DocBook |
| 14:14 | <webben> | Dashiva: Namespaces must be separate to avoid collisions between names. |
| 14:15 | <webben> | obviously if you have collisions /within/ namespaces, that's bad |
| 14:15 | <Dashiva> | But that's what you're trying to prove, not your premise |
| 14:15 | <Philip`> | I suppose a problem is <html><X3D><script></X3D></html>, since the XHTML layer has to either ignore the entire subtree rooted at the unknown-to-XHTML <X3D> element, which is bad for extensibility of XHTML; or it has to ignore that element and process the child elements, in which case it'll incorrectly execute the script |
| 14:15 | <webben> | Dashiva: Well, in a world without namespacing of any sort, functionality for names couldn't change without breaking backwards compat. |
| 14:15 | <Philip`> | (assuming that's meant to be an X3D script element, not an XHTML script element) |
| 14:16 | <Dashiva> | webben: That is true |
| 14:16 | <webben> | And a name, once chosen, would be universal throughout documents even if it's wildly inappropriate. |
| 14:16 | <Philip`> | (and assuming the non-XHTML languages always have a root element which is not in XHTML) |
| 14:16 | <webben> | over time, you'd have ever more complicated and non-intuitive names to embrace new functionality |
| 14:16 | <Dashiva> | So instead you make all names complicated and non-intuitive |
| 14:16 | <webben> | and you'd need a central registry which could actually manage all those names (even names used in private companies for private purposes) |
| 14:17 | <webben> | to ensure that when a public standard uses a new name it doesn't conflict with internal XML. |
| 14:18 | <webben> | I think namespacing is sound and used in most languages (e.g. CPAN). Whether the particular way XML does namespacing (the whole weird URIs, and prefixes, and null namespaces + schemas + doctypes craziness) makes any sense is another matter |
| 14:18 | <Dashiva> | Also true, but it's quite a significant matter |
| 14:18 | <webben> | Very significant. Yes. |
| 14:18 | <webben> | possibly grounds for revision to XML at some point |
| 14:19 | <Dashiva> | If we could do namespaces in an easy, transparent way I don't see many arguments against it at all |
| 14:23 | <hsivonen> | if anyone knows of legitimate annotation-xml content that is not MathML, SVG, XHTML or OpenMath, please let me know |
| 14:24 | <kig> | (re: svg on canvas, another bonus of it is that as long as you have javascript and implement the canvas api, you can use svg in your drawing (say, on the iPhone and on the OpenGL canvas, etc.)) |
| 14:25 | <webben> | hsivonen: What is "legitimate annotation-xml"? |
| 14:27 | <hsivonen> | webben: stuff that is deployed in annotation-xml and is appropriate for sending over the wire in the Web context |
| 14:27 | <Philip`> | kig: About getTransform: why not do all the transform matrix calculations in JS, then just use setTransform (or decompose into scale/rotate/translate) once before rendering? |
| 14:28 | <kig> | i am doing that now, and i'd rather not |
| 14:28 | <Philip`> | (That's what I've done for X3D in canvas-3d, and it seems to work alright) |
| 14:28 | <kig> | (both) |
| 14:29 | <kig> | i mean, the underlying canvas implementation must have knowledge of the ctm to be able to draw anything, so duplicating the math in JS is a waste and a bother |
| 14:32 | <Philip`> | The underlying implementation might be an external library which the browser uses, and the library doesn't necessarily have to expose the CTM to the browser |
| 14:32 | <Philip`> | but at least Cairo has a cairo_get_matrix, and it'd be a silly library to not expose that kind of information, so that's probably not a practical problem |
| 14:34 | Philip` | guesses the next version of canvas will have dashed lines, since that seems the most demanded feature after text rendering |
| 14:34 | <kig> | are skew matrices invertible :? |
| 14:34 | <hsivonen> | but that CTM is the CTM of the back end relative to the world. it isn't the "CTM" relative of the canvas relative to page |
| 14:35 | <Philip`> | kig: Yes, since they're not singular |
| 14:35 | <Philip`> | (Well, assuming it's not skewed infinitely into a straight line, because that'd be silly) |
| 14:37 | <kig> | i have a "decompose matrix into translate+rotate+scale" to do setTransform, transform and isPointInPath in JS, but i don't know what to do with skews |
| 14:38 | <Philip`> | Oh, do you mean decomposing skews into translate+rotate+scale? |
| 14:39 | <kig> | and my crappy matrix inverter uses the decomposition thing too :< |
| 14:41 | <kig> | Philip`: yeah |
| 14:41 | <Philip`> | kig: Seems easier to use something like http://www.j3d.org/matrix_faq/matrfaq_latest.html#Q23 for inverting |
| 14:42 | <kig> | nice, thanks! |
| 14:43 | <Philip`> | I'm not aware of a totally obvious way to decompose skews - it looks like they should be solvable as rotate+scale+rotate, but I don't know any details |
| 14:43 | <kig> | http://pastie.caboo.se/130487 <- what i have now |
| 14:49 | <Philip`> | Ah, okay, that seems quicker than the general 3x3 case |
| 14:49 | <Philip`> | Is it actually correct, though? |
| 14:50 | <Philip`> | (I get a quite different answer if I ask Maxima how to invert a 3x2 matrix) |
| 14:51 | <kig> | it works here, at least. first inverts scale, then inverts rotation, then inverts translation |
| 14:51 | <kig> | but might also be that my whole matrix math lib is broken, but in the right way |
| 14:52 | <Philip`> | It says invert(invert([1,2,3,4,5,6])) = 3.6764705882352944,2.1551724137931028,2.2058823529411766,0.8620689655172411,15.941176470588237,8.672413793103447 |
| 14:54 | <Philip`> | function invert(m) { var d = 1/(m[0]*m[3]-m[1]*m[2]); return [ m[3]*d, -m[1]*d, -m[2]*d, m[0]*d, m[2]*m[5]-m[3]*m[4], m[1]*m[4]-m[0]*m[5] ]; |
| 14:54 | <Philip`> | } |
| 14:54 | <Philip`> | seems to work better |
| 14:55 | <Philip`> | (Maybe that's just a problem with inverting skewed matrices) |
| 14:55 | <kig> | 1,2,3,4,5,6 is skewed, yeah |
| 14:55 | <Philip`> | Argh |
| 14:56 | <Philip`> | return [ m[3]*d, -m[1]*d, -m[2]*d, m[0]*d, d*(m[2]*m[5]-m[3]*m[4]), d*(m[1]*m[4]-m[0]*m[5]) ]; |
| 14:56 | <Philip`> | works betterer |
| 14:56 | <Philip`> | and is actually correct |
| 14:59 | <kig> | yeah, that works for skews too 8) |
| 14:59 | <Philip`> | Doesn't help with decomposing skews for canvases without setTransform, though... |
| 15:03 | <hsivonen> | error 503 mod_jk sucks |
| 15:04 | <kig> | Philip`: it needs to be var d = (m[0]*m[3]-m[1]*m[2]); if (Math.abs(d) < 0.005 /* or such */) return [1,0,0,1,0,0]; var id = 1 / d; ... |
| 15:05 | <kig> | otherwise zero division on e.g. [1,0,0,0,0,0] |
| 15:05 | <Philip`> | Zero division is non-fatal, so returning an incorrect result ([1,0,0,1,0,0]) instead of another incorrect result ([Inf,Inf,...]) doesn't seem that much of an improvement :-) |
| 15:20 | <Philip`> | Aha - apparently A = Q R S R^T for any A and for suitable orthonormal Q and R and diagonal S |
| 15:21 | <Philip`> | (so you can decompose anything into rotations and scales) |
| 15:24 | <Philip`> | (The trick is in finding Q and R, I suppose) |
| 15:29 | <Philip`> | http://www.me.rochester.edu/courses/ME444/MatrixPolarDecomp.pdf - that looks non-trivial |
| 15:31 | <kig> | i found this for doing a skew with rotation and scale: http://newsgroups.cryer.info/comp/graphics.algorithms/200602/07/0602072134.html |
| 15:33 | <Philip`> | Ah, that looks like a slightly easier approach |
| 15:34 | <kig> | and apparently graphics gems vol 2 ch 7 describes how to extract the skew from a matrix http://tog.acm.org/GraphicsGems/AllGems.TOC |
| 15:34 | <Philip`> | though you still need to do a bit of work if you want to decompose an arbitrary matrix, rather than just a [[1,a],[0,1]] one |
| 15:34 | <kig> | http://tog.acm.org/GraphicsGems/gemsii/unmatrix.c |
| 15:35 | <Philip`> | I assume you don't care about 4x4 matrices, though |
| 15:36 | <kig> | might be useful for unbreaking the opera-3d api.. |
| 15:36 | <Philip`> | Ah, true |
| 15:38 | <Philip`> | Looks like much nicer code than http://www.soe.ucsc.edu/~pang/160/f98/Gems/GemsIV/polar_decomp/Decompose.c though I don't know if it's differently featured in a significant way |
| 15:41 | <Philip`> | Oh, that polar decomposition says it's designed to produce matrices which can be animated nicely, whereas the Graphics Gems II one tries "to synthesize an arbitrary matrix from a limited set of primitives, without regard for meaningfulness of the decomposition" |
| 15:41 | <Philip`> | (says http://www.cs.wisc.edu/graphics/Courses/838-s2002/Papers/polar-decomp.pdf) |
| 16:18 | <webben> | Is there an easy way to check that a relax ng schema is a valid schema? |
| 16:30 | <hsivonen> | webben: yes. put the schema URI in the Validator.nu schema field and put any document URI in the document field |
| 16:30 | <webben> | hsivonen: Cool, thanks. |
| 17:05 | <webben> | hsivonen: I donwloaded the validator locally and got it running. But when I point it at documents and schema at http://localhost:1024 it says "IO Error: Attempted to connect to localhost." ... yet if I take the same URIs and paste them into a firefox tab address bar, they resolve and work fine |
| 17:05 | <webben> | That doesn't happen to ring any bells about what might not be working, does it? |
| 17:05 | <hsivonen> | webben: It's a deliberate tin-foil hat. I guess it is a bad one |
| 17:06 | <webben> | tin-foil hat against what? |
| 17:06 | <Philip`> | It's a tin-foil hat with a large hole in the top of it |
| 17:07 | <Philip`> | (since you can e.g. use DNS to make the validator connect to 127.0.0.1) |
| 17:07 | <webben> | hsivonen: Is there an easy way to take off the tin-foil hat? |
| 17:08 | <hsivonen> | webben: if you run python build/build.py checkout now, it the hat will come off |
| 17:08 | <Philip`> | webben: util/src/nu/validator/xml/PrudentHttpEntityResolver.java |
| 17:08 | <Philip`> | and comment out the bit with "127.0.0.1".equals(host) etc |
| 17:08 | <hsivonen> | Philip`: now commented out in svn |
| 17:09 | <Philip`> | validator.w3.org seems a bit cleverer since it resolves the IP address and fails if it's not a public range |
| 17:10 | <webben> | hmm the hat seems not to have come off |
| 17:10 | <webben> | I'll try checkout a second time. |
| 17:12 | <hsivonen> | XMLFilterImpl is messing up my careful ErrorHandler setup :-( |
| 17:13 | <webben> | yeah, behavior is unchanged |
| 17:15 | <Philip`> | webben: I assume you built and ran the updated version too? :-) |
| 17:16 | <webben> | Philip`: checkout then run |
| 17:16 | <webben> | oh hang on |
| 17:16 | <webben> | did i need a build in between there? |
| 17:17 | <hsivonen> | webben: python build/build.py checkout build run |
| 17:18 | <webben> | ah :) sorry. Thanks. |
| 17:18 | <webben> | that seems to work |
| 17:42 | <hsivonen> | Do I read Dean Edrigdes message correctly that he is basically talking about using the spec to force IE to implement application/xhtml+xml? |
| 17:45 | <gsnedders> | hsivonen: that's how I read it too |
| 17:48 | <Philip`> | Don't forget Konqueror! |
| 17:58 | <webben> | Philip`: Konqueror has a sort-of XHTML mode already, I think. |
| 17:59 | <Philip`> | It uses its HTML parser for application/xhtml+xml in the currently released versions, and when I last looked they weren't actively working on fixing that |
| 18:02 | <webben> | Philip`: The impression I got from talking to the devs about a year ago was that it's XHTML-handling is a bit more complicated than just pretending its broken HTML. |
| 18:02 | <webben> | *its |
| 18:27 | <gsnedders> | ergh. |
| 18:27 | gsnedders | has to answer a question on why the people writing the software are not the best people to test it |
| 18:38 | <kig> | ouch, all browsers handle canvas image pattern transforms differently |
| 18:39 | <kig> | firefox's being the most useful permutation of the bunch |
| 18:42 | <kig> | opera conforms to the spec's "transformations screw everything up", firefox does "transformations affect patterns and path points created after the transformation", safari does what firefox does, but if there's a translation and rotation, it gives up and draws nothing. |
| 19:00 | <kig> | http://dark.fhtr.org/imageTest.html |
| 19:07 | <kig> | ah, safari only fails on no-repeat. go figure. |
| 19:25 | <Philip`> | webben: Ah, you're probably right - I think it did weird things when I last tried |
| 19:25 | <Philip`> | but with e.g. http://canvex.lazyilluminati.com/83/play.xhtml it definitely looks like it's using an HTML parser instead of an XML one, and messing up the doctype |
| 19:26 | <webben> | Philip`: Serving it a custom DTD is probably not playing fair ;) |
| 19:27 | <Philip`> | kig: http://philip.html5.org/tests/canvas/suite/tests/index.2d.pattern.html covers the areas where I noticed differences between browsers |
| 19:27 | <Philip`> | webben: Any XML parser should be able to handle that with no problem :-) |
| 19:27 | <webben> | Philip`: Should, no doubt. :) |
| 19:28 | <Philip`> | If it doesn't, it's not an XML parser, it's just a tag-soup-which-looks-a-bit-like-XML parser :-) |
| 19:28 | <webben> | Philip`: I did call it a "sorta-XHTML mode" |
| 19:29 | <Philip`> | I don't remember ever seeing XML parse errors from Konqueror either |
| 19:30 | <Philip`> | I would have assumed an XML parser would be the first thing someone would add to a sorta-XHTML browser |
| 19:30 | <Philip`> | (but I would have quite possibly assumed wrong) |
| 19:31 | <webben> | XML parsers? Who needs em? ;) |
| 19:34 | <Philip`> | It's lucky that Konqueror isn't quite as popular as IE, else it would be destroying XHTML |
| 19:34 | <Philip`> | (or at least destroying application/xhtml+xml) |
| 19:35 | <kig> | Philip`: btw, did you try this patch? https://bugzilla.mozilla.org/show_bug.cgi?id=401790 |
| 19:35 | <Philip`> | kig: I haven't tried it |
| 19:37 | Philip` | should get around to working on the Mozilla canvas regression tests again, to add all the currently-failing tests (though keep them disabled) and remind the developers how many they still fail |
| 19:38 | <gsnedders> | Philip`: 42 plx kthxbai |
| 19:40 | <gsnedders> | on the subject of tests, I need to answer, "Explain why the people responsible for 'writing' the software are not the best people to test it." |
| 19:40 | <kig> | gsnedders: what do you think? |
| 19:41 | <gsnedders> | kig: I think anyone who can't test their own code because they think it's perfect needs to get out of idealism. |
| 19:43 | <kig> | and this applies to other fields as well? |
| 19:45 | <kig> | e.g. teachers are the best judges of the quality of their classes? |
| 19:46 | <gsnedders> | if they can escape from an idealistic view, they can be very good judges. |
| 19:54 | <Philip`> | gsnedders: The people writing the software have a negative incentive to find bugs, because a stupid manager will think it reflects badly on the programmer's abilities, and they will have to fix the bugs now instead of letting them accumulate for a couple of releases and then quitting the job and leaving the mess to someone else |
| 19:54 | gsnedders | wonders whether to copy that verbatim :P |
| 19:55 | <Philip`> | Also if your testing is trying to determine whether the code follows some specification, the programmer might misinterpret the specification, so you want a different person to be checking it |
| 19:57 | <Philip`> | Also, maybe a programmer's time is more expensive than QA monkey's time |
| 19:57 | <Philip`> | s/than/than a/ |
| 19:57 | <gsnedders> | Philip`: oh, this is all waterfall model, so there has to been a specification :P |
| 19:57 | <Philip`> | There's plenty of reasons, of varying levels of dubiousness :-) |
| 19:58 | <gsnedders> | Philip`: I'm trying to think of any that aren't dubious. Not easy :P |
| 19:59 | Philip` | sees video discussion on the X3D mailing list, and a link to http://lists.xiph.org/pipermail/theora-dev/2003-February/000435.html which seems to be saying 'Theora is relatively safe because nobody cares enough to sue us' |
| 20:00 | <jwalden> | I think the common reason is "I can get more done and provide more benefit to users by spending time fixing bugs instead of writing tests" |
| 20:00 | <gsnedders> | I think all the booklet says is bias and pride in their own work. |
| 20:01 | <gsnedders> | Anyone who knows me knows that I'm always slagging off my code :P |
| 20:01 | <Philip`> | jwalden: Is that the actual reason, or just the stated reason which covers up the fact that testing is boring and programmers want to do fun things instead? :-) |
| 20:02 | <jwalden> | Philip`: I think it's the real reason, and I definitely know a person or two where it's probably accurate (at least one because when he writes the tests he spends so much time making them rigorous) |
| 20:03 | <jwalden> | and he usually writes them anyway, but I can see utility in having someone else doing at least some of the in-depth testing for him |
| 20:03 | <gsnedders> | How widely used is the waterfall model anyway (i.e., how irrelevant is what we're being taught)? |
| 20:06 | <krijn> | Is it okay to do Location: http://example.com/#fragment in HTTP? |
| 20:07 | <gsnedders> | krijn: Location must be an absolute URI, so yeah |
| 20:07 | <krijn> | Including the fragment identifier, I mean |
| 20:08 | <gsnedders> | AFAIK yeah |
| 20:08 | <krijn> | I thought that had issues, some years ago |
| 20:08 | <krijn> | But that could've been PHP screwing up |
| 20:09 | <hsivonen> | gsnedders: the person writing the code might not have come up with all the possible ways of abusing the code. It helps to have someone else figure out ways to put the code into unexpected states. |
| 20:09 | <gsnedders> | krijn: yeah, it is definitely allowed |
| 20:09 | <krijn> | gsnedders: Okay, thanks :) |
| 20:09 | <gsnedders> | Philip`: Location headers in the data set wold be nice :P |
| 20:09 | gsnedders | nudges Philip` |
| 20:11 | krijn | slaps Mathias ;p |
| 20:14 | <Mathias> | Well I'm guessing it was an old version of PHP messing up with the # character. |
| 20:14 | Mathias | slaps Krijn right back |
| 20:15 | <gsnedders> | PHP-- |
| 20:15 | <krijn> | Yeah, who uses PHP anyway? ;P |
| 20:16 | <gsnedders> | Me, as little as possible :P |
| 20:16 | <Mathias> | <?php echo 'My PHP is all I have.'; ?> |
| 20:19 | <csarven-> | anyone know a good frontend dev person looking for full-time work? |
| 20:21 | <csarven-> | (from Montreal) |
| 20:22 | <gsnedders> | I meet the good, person, looking, work parts, but not the rest. |
| 20:23 | <csarven-> | trying to find someone for http://faq.css-standards.org/Integrator :) |
| 20:31 | gsnedders | wonders if it is possible to call pride "naïve" |
| 21:13 | <Lachy> | apparently IE8 will pass acid 2 http://blogs.msdn.com/ie/archive/2007/12/19/internet-explorer-8-and-acid2-a-milestone.aspx |
| 21:20 | <jruderman> | does that mean it has support for data: URLs? |
| 21:20 | <Lachy> | yes, if data URLs are tested in acid2. I can't recall. |
| 21:20 | <gsnedders> | jruderman: almost certainly yes |
| 21:20 | <gsnedders> | Lachy: yeah, they are |
| 21:20 | <jruderman> | any other features we can say the same about? |
| 21:20 | <gsnedders> | jruderman: hopefully we'll get a list of new features/bug fixes in IE8 |
| 21:21 | <jruderman> | based on the acid 2 news, i mean |
| 21:21 | <gsnedders> | jruderman: display:table* |
| 21:21 | <gsnedders> | jruderman: position: fixed |
| 21:21 | <gsnedders> | jruderman: whatever bugs were blocking it |
| 21:21 | <Hixie> | we made a no-data-url version of the test for them, so not necessarily |
| 21:21 | <gsnedders> | Hixie: I'm aware, that's why I'm slightly hesitant about it :) |
| 21:21 | <gsnedders> | jruderman: (the list of things for acid2 that is) |
| 21:22 | <gsnedders> | Hixie: but there again, the filename list implies data URI support |
| 21:23 | <jruderman> | "Change Description: Reverse integration from green branch. Includes full implementation of ACID2" hmm |
| 21:23 | <jruderman> | oh, so that's what "dataprot" means |
| 21:27 | <jruderman> | it will be interesting to see what triggers "IE8 standards mode" |
| 21:27 | <gsnedders> | jruderman: see html-wg logs (linked from /topic) |
| 21:27 | <gsnedders> | jruderman: or join yourself |
| 21:28 | <jruderman> | ? |
| 21:28 | <jwalden> | Lachy: re <http://www.alistapart.com/articles/previewofhtml5>;, could you update it to add a link for Mozilla's experimental video builds? a decent link target would be <http://www.double.co.nz/video_test/>; |
| 21:28 | <gsnedders> | jruderman: http://krijnhoetmer.nl/irc-logs/html-wg/20071219 |
| 21:28 | <Lachy> | jwalden, sorry, no can do. It can't be updated |
| 21:28 | <Lachy> | I wish someone had told me about them before |
| 21:29 | <jwalden> | Lachy: why not? I wasn't aware ALA was anything beyond the web presentation where fixedness would matter |
| 21:30 | <jruderman> | an opt-in meta tag? eww |
| 21:34 | <hsivonen> | zcorpan: bad stuff. the source location bug on Musings is not reproducible when I copy the doc to another server |
| 21:34 | <hsivonen> | might be something bad around buffer boundaries |
| 21:34 | <jwalden> | now hope for sufficient evangelism that it can be made clear that the tag grants permission for IE to break your layout in the name of standards compat |
| 21:34 | <hsivonen> | hard to debug |
| 21:34 | <jwalden> | i.e. what doctype was supposed to be |
| 21:39 | <gsnedders> | Hixie: watching the video, it's the real acid2 test with data URI et all |
| 21:40 | <Hixie> | cool |
| 21:40 | <Hixie> | man, i agree with shawn's rant in http://www.w3.org/mid/994fc8d00712190734y3dea47derd9ba296d5232d541⊙mgc |
| 21:42 | <gsnedders> | they have worn printed copies of the CSS 2.1 :P |
| 22:59 | Philip` | wonders if the IE8 Acid2 thing makes Opera's antitrust complaint significantly less valid |
| 23:12 | <roc> | not really |
| 23:12 | <roc> | Acid2 is just a slice of a very large standards pie |
| 23:24 | <othermaciej> | Microsoft themselves used to downplay the importance of Acid2 quite a bit |
| 23:24 | <Hixie> | tehy still do |
| 23:25 | <Hixie> | even in that post they (correctly) point out it's just one test of many |
| 23:25 | <Hixie> | and not an especially good test from a standadrs-compliance perspective |
| 23:25 | <Hixie> | acid3 is mildly better in that regard, as constructed at the moment |