| 00:07 | <Hixie> | zewt: you're making your life too complicated. just googled [structured clone], no site: operator, no quotes, and the first hit for me is the whatwg spec. |
| 00:07 | <Hixie> | admittedly, the wrong page of it :-) |
| 00:08 | <Hixie> | but that's what "find in page" and the single page version are for is for |
| 00:09 | <Hixie> | s/is for// |
| 00:09 | <zewt> | Hixie: my first hit is http://www.whatwg.org/specs/web-apps/current-work/multipage/urls.html, which is entirely wrong |
| 00:10 | <zewt> | the entire reason i'm searching is to avoid destroying my browser with the single page spec. heh |
| 00:10 | <Hixie> | use a better browser :-) |
| 00:11 | <zewt> | there are no browsers that can quickly load that. heh |
| 00:11 | <Hixie> | i leave it open in a tab in chrome all the time with no problems |
| 00:11 | <zewt> | i think it's entirely reasonable to expect "site:whatwg.org/specs structured clone" to find the right page :) (and it used to!) |
| 00:11 | <zewt> | no idea what changed |
| 00:12 | <Hixie> | oh i'm not saying it shouldn't work :-) |
| 00:13 | <ginger> | yay! modal dialogs! |
| 00:45 | <Yuhong> | Also horrible: using "standard-compliance" when "boat anchor" is meant. |
| 01:58 | <zewt> | if i hear one more person that should know better claim that "#! breaks the web" ... |
| 02:01 | <heycam> | twitter.com/#!/bin/sh |
| 03:11 | <onr> | lol'd |
| 03:12 | <onr> | iirc, google requires #! to crawl ajaxed pages |
| 06:09 | <annevk> | every Typed Array thread ends up with lots of email |
| 06:10 | <annevk> | but I guess as long as people don't expect me to read through it all, I'm good |
| 06:15 | <heycam> | you need an TypedArrayThreadView to show you just the important ones |
| 06:15 | <annevk> | my filter is just reading what bz writes |
| 06:16 | <annevk> | and glenn |
| 06:16 | <annevk> | well zewt |
| 06:16 | <heycam> | good filter |
| 06:17 | <annevk> | per my filter we're going to change send() to only take ArrayBufferView |
| 06:17 | <annevk> | but responseType = "arraybuffer" is perfectly acceptable |
| 06:21 | <annevk> | btw hsivonen, do you still think we should change it to "full screen" given that Mozilla now names everything "fullscreen" (modulo the API) |
| 06:22 | <zcorpan> | Hixie: if scripting is disabled, should <form method=dialog> submit anyway? why is it called "dialog"? |
| 06:23 | <Hixie> | it's called dialog because it closes dialogs |
| 06:23 | <Hixie> | and if you want to support script-disabled cases, then don't use it, just cancel the submit even |
| 06:23 | <Hixie> | t |
| 06:23 | <annevk> | does create a weird asymmetry with XMLHttpRequest where method exclusively means HTTP method |
| 06:24 | <Hixie> | (or whatever it is you have to do) |
| 06:24 | <annevk> | but action="about:dialog" seems somewhat ugly |
| 06:25 | <nesta_> | good day! :) |
| 06:25 | <annevk> | good morning |
| 06:25 | <Hixie> | annevk: using action="" would mean you couldn't do legacy fallback |
| 06:26 | <annevk> | ah yeah, I saw you mentioning that |
| 06:26 | <Hixie> | not that i think anyone will really ever do that, but anyway |
| 06:29 | <Hixie> | abarth: btw the table at http://www.whatwg.org/specs/web-apps/current-work/#browsing-context-names may be of interest |
| 06:29 | <Hixie> | abarth: though it doesn't answer your question in this case |
| 06:30 | <Hixie> | abarth: since you're not targetting anything, you're just starting a navigation |
| 06:30 | <abarth> | looking |
| 06:30 | <abarth> | this is a nice table |
| 06:30 | <abarth> | i will convert it to a test suite |
| 06:30 | <Hixie> | heh |
| 06:31 | <Hixie> | abarth: that table suggests that seamless overrides sandbox, actually |
| 06:31 | <Hixie> | (that table is non-normative btw) |
| 06:31 | <abarth> | is my case different from "none specified, for links and form submissions " |
| 06:31 | <abarth> | ? |
| 06:32 | <Hixie> | ok, i've now read the spec |
| 06:32 | <Hixie> | the sandboxing override happens _before_ the seamless redirect |
| 06:32 | <Hixie> | so the table is correct |
| 06:33 | <Hixie> | and window.location = foo is equivalent to the top row, but scripted |
| 06:33 | <abarth> | so, we should navigate the master |
| 06:33 | <Hixie> | right |
| 06:33 | <Hixie> | if scripting is enabled |
| 06:33 | <abarth> | thanks! |
| 06:33 | <abarth> | yeah |
| 06:33 | <abarth> | will fix |
| 06:33 | <abarth> | I'm glad I asked |
| 06:34 | <Hixie> | i'll reply on the list as well in case anyone else cares |
| 06:35 | <Hixie> | Content-Type: text/plain; charset=UTF-8 |
| 06:35 | <Hixie> | that's right, no? |
| 06:36 | <Hixie> | wtf |
| 06:36 | <abarth> | I think so |
| 06:37 | <Hixie> | why the hell is apache stripping my charset |
| 06:37 | <Hixie> | oh, it's not |
| 06:37 | <Hixie> | ok now i'm even more confused |
| 06:38 | <hsivonen> | should I reward the use of a bug tracker as a helpdesk by giving the answer: http://bugzilla.validator.nu/show_bug.cgi?id=918 ? |
| 06:42 | <zcorpan> | i'd say something like "don't use bugzilla for this. use <appropriate forum> instead. but the answer is <foo>. RESOLVED INVALID" |
| 06:44 | <Hixie> | or just create a component for helpdesk questions, they're gonna come anyway :-) |
| 06:47 | <Hixie> | gah, i hate all this crap in libraries and languages that tries to hide the difference between byte strings and unicode strings |
| 06:55 | <annevk> | what do you mean? |
| 06:56 | <Hixie> | like, if you read something from the console, it should just be a byte stream, and it should be impossible to concatenate it to a string "foo" |
| 06:56 | <Hixie> | because "foo" is a bunch of characters, not a byte stream |
| 06:57 | <Hixie> | unless you have an encoding to use to convert the byte stream to a character stream, it should remain a byte stream |
| 07:20 | <zcorpan> | so i guess websocket needs to s/ArrayBuffer/ArrayBufferView/ as well |
| 07:23 | zcorpan | finds out that the feedback form in the websocket spec on dev.w3.org doesn't work |
| 07:34 | <annevk> | who are involved in Fullscreen for the various browsers? |
| 07:34 | <annevk> | oh well |
| 07:35 | <annevk> | https://www.w3.org/Bugs/Public/show_bug.cgi?id=16709 |
| 07:35 | <heycam> | annevk, chris pearce for us |
| 07:35 | <annevk> | ah |
| 07:35 | <annevk> | thanks, addedh im |
| 07:44 | <annevk> | bz being too short for W3C Bugzilla sucks |
| 07:44 | <annevk> | they could at least special case him |
| 07:55 | <annevk> | thanks foolip for http://lists.w3.org/Archives/Public/public-html-ig-zh/2012Apr/0007.html and other emails! |
| 07:55 | <annevk> | foolip: I'm following along with Google Translate :) |
| 07:56 | <foolip> | annevk, as you might have guessed, they're a bit worried about other Big5 extensions |
| 07:56 | <zcorpan> | foolip: any progress on the gb data? |
| 07:56 | foolip | checks how well Google Translate understands his Chinese... |
| 07:56 | <annevk> | I looked for Taiwanese browser market share, and it's IE + Chrome, then Firefox |
| 07:56 | <foolip> | zcorpan, no, I haven't had time yet |
| 07:57 | <annevk> | Firefox having like 10% |
| 07:57 | <annevk> | and IE about 60% |
| 07:57 | <foolip> | annevk, the only think I could see making a difference is if there is some font using PUA that is installed on virtually all Taiwanese windows machines |
| 07:57 | <annevk> | which seems to me that if those pages use Firefox extensions, they're not aimed at the majority |
| 07:58 | <annevk> | foolip: true |
| 07:58 | <foolip> | But so far I haven't heard of such a thing, so I hope it is not common... |
| 07:58 | <annevk> | foolip: Chrome does match IE pretty closely for PUA |
| 07:58 | <foolip> | Indeed, which is why it could potentially be a problem. |
| 08:00 | <annevk> | god I hate PUA |
| 08:00 | <annevk> | so much |
| 08:00 | <foolip> | I hope we succeed in killing it for Big5 |
| 08:00 | <foolip> | Are there other encodings that use it? |
| 08:02 | <zcorpan> | utf-8? :-) |
| 08:02 | <foolip> | Yeah, of course :) |
| 08:02 | <annevk> | macintosh is the only index |
| 08:03 | <annevk> | http://dvcs.w3.org/hg/encoding/raw-file/tip/index-macintosh.txt |
| 08:03 | <annevk> | 112 0xF8FF (<Private Use>) |
| 08:04 | <foolip> | Huh, what's it usually used for? Can it be destroyed? |
| 08:04 | <annevk> | it's Apple's logo |
| 08:04 | <annevk> | it's used all over |
| 08:05 | <annevk> | http://en.wikipedia.org/wiki/Private_Use_(Unicode)#Example_code_point_U.2BF8FF |
| 08:06 | <foolip> | Oh well, nothing to do about it I guess |
| 08:06 | <zcorpan> | that code point should just be grandfathered in unicode to become a proper character |
| 08:06 | <annevk> | the more scary trend is these icon fonts that use PUA all over |
| 08:07 | <annevk> | but when copy and pasted lose all original meaning |
| 08:07 | <annevk> | the technique has awesome benefits such as icons scaling with the text, etc. |
| 08:08 | <annevk> | but as far as semantics goes it's spacer gifs and tables for layout all over again |
| 08:08 | <zcorpan> | icon fonts should use ligatures and real words |
| 08:09 | <zcorpan> | like a ligature for "twitter" being an icon |
| 08:09 | <hsivonen> | zcorpan: indeed |
| 08:09 | <foolip> | zcorpan, do you mean that the Apple logo should be added to a new code point, or that U+F8FF should be redefined? |
| 08:10 | <zcorpan> | foolip: the latter |
| 08:10 | <hsivonen> | zcorpan: does OpenType place limits on the number of code points that can ligate into one glyph? |
| 08:11 | <zcorpan> | don't think so, there's one ligature that's a whole word in some font. don't recall what it is |
| 08:12 | <foolip> | zcorpan, I don't disagree, but Unicode politics seem pretty hard to navigate |
| 08:13 | <hsivonen> | I've seen whole-word ligatures in AAT fonts but I haven't seen them in OpenType fonts |
| 08:14 | <zcorpan> | oh. i don't recall whether this was OpenType or not |
| 08:14 | <hsivonen> | AAT Zapfino has a ligature for "Zapfino" and, IIRC, some AAT fonts when set to maximum ligatures ligate "Apple" to the Apple logo |
| 08:15 | <zcorpan> | Zapfino was probably what i was thinking of |
| 08:34 | <kennyluck> | Does anyone know why the CSS specs tend to have very few statements of conformance requirement? They are usually full of sloppy statements of facts. |
| 08:35 | <annevk> | writing specs is hard |
| 08:35 | <annevk> | doing so in committee doubly so |
| 08:35 | kennyluck | was using http://ln.hixie.ch/?start=1140242962&count=1 to categorize statements. |
| 08:36 | <kennyluck> | Well, using "must" or not seems more like a cultural problem than a technical one. |
| 08:36 | <annevk> | you don't use "must" in statements of fact |
| 08:37 | <Ms2ger> | Yeah, CSS doesn't bother with conformance requirements |
| 08:37 | <kennyluck> | annevk, I know. What I am saying is that you usually see fewer 'must's in CSS specs. |
| 08:39 | <annevk> | in terms of quality CSS is somewhere in the middle between HTML4 and HTML, erring towards HTML4 if not many people are looking or if they give up on the hard problems when people are looking |
| 08:39 | <annevk> | they're adding a bunch of new layout modules without understanding table layout |
| 08:40 | <annevk> | that seems irresponsible to me, but nobody cared much when I brought it up |
| 08:40 | <Ms2ger> | New stuff is more fun than old stuff, or so I hear |
| 08:41 | <annevk> | no man, legacy encodings are the shit :) |
| 08:41 | <Ms2ger> | s/the// :) |
| 08:41 | <annevk> | :p |
| 08:54 | <Von_Davidicus_> | Validation question: How do I get the HTML5 validator to recognize that I'm using XHTML5, and thus the XML processing instruction isn't an error? |
| 08:54 | <annevk> | by not using text/html as MIME type |
| 08:54 | <annevk> | but since you didn't do that yourself, it's probably best to stick to HTML |
| 09:01 | <Von_Davidicus_> | Must be the way I'm validating it (by file upload) |
| 09:02 | <annevk> | file extension? |
| 09:03 | <Von_Davidicus_> | .xhtml |
| 09:03 | Von_Davidicus_ | has a habit of using that file extension. |
| 09:07 | <annevk> | dunno then |
| 09:08 | <annevk> | you could set a Parser override just in case |
| 09:08 | <annevk> | to XML that is |
| 09:08 | <annevk> | assuming you're using http://validator.nu/ |
| 09:09 | <Von_Davidicus_> | Whoops, my bad. I had changed the extension to .html. |
| 09:09 | Von_Davidicus_ | switches file back to "Nerd Mode" :) |
| 09:18 | <kennyluck> | i think asking Firefox to drop certain big5 mapping is just like asking browsers to drop support for their proprietary CSS extension that is already used by some sites. It's neither beneficial to Firefox as a browser vendor nor to its users. |
| 09:19 | <annevk> | Firefox currently renders several big5 encoded sites incorrectly |
| 09:20 | <Von_Davidicus_> | Before I publish a statement, may I ask your opinions as to its accuracy? |
| 09:22 | <Von_Davidicus_> | I'm doing some writing on website coding. The statement is: "XHTML 2.0 was virtually booed off the web." |
| 09:22 | <kennyluck> | You can try data:application/xhtml+xml as a workaround. |
| 09:24 | <annevk> | kennyluck: also, fyi, Firefox has dropped proprietary extensions in the past to make the web better |
| 09:25 | <annevk> | kennyluck: but I'm curious what data you have that suggests Firefox's big5 behavior is better than that of other browsers because all the data I've looked at suggests otherwise |
| 09:29 | <kennyluck> | annevk, because it displays the Web archive of a BBS I visited everyday correctly → http://www.ptt.cc/bbs/C_Chat/M.1334222500.A.047.html |
| 09:30 | <kennyluck> | It gives me "いま俺の顔生涯最高にキモい自信がある" in Firefox and "俺生涯最高自信" in other browsers. |
| 09:31 | <kennyluck> | I almost never use HK sites so I might be selfish saying so though. |
| 09:33 | <kennyluck> | by the way, big5-hkscs gives "いま俺の𡟺生涯最高にキモい自信がある" which is a bit different too. |
| 09:36 | <annevk> | but that looks a lot closer already |
| 09:36 | <foolip> | kennyluck, how should "big5" be interpreted to fix the most number of sites? |
| 09:39 | <kennyluck> | foolip, I doubt we should fix that to be locale-dependent. As a non-HK person, the best for me seems to be "big5" maps to "big5-uao" in a zh-TW browser and "big5-hkscs" in a zh-HK browser. |
| 09:40 | <foolip> | kennyluck, so you think it *should* be locale-dependent? |
| 09:41 | <kennyluck> | foolip, probably just like how the default encoding is specced. |
| 09:41 | <kennyluck> | Yes |
| 09:41 | <kennyluck> | (though ironically the default encoding of zh-TW Firefox is UTF-8 instead of big5, which is not following the spec.) |
| 09:42 | <foolip> | kennyluck, do Windows machines in Taiwan have fonts using PUA that make this work in other browsers than Firefox, and how common is that? |
| 09:43 | <Von_Davidicus_> | annevk, may I get your opinion as to the accuracy of a statement I was hoping to put into an essay about HTML? |
| 09:43 | <kennyluck> | foolip, no data from me :( |
| 09:44 | <foolip> | kennyluck, do you know if .tw sites depending on Big5-UAO are common? FWIW, I use a few Taiwan sites regularly and have not encountered this problem in Opera |
| 09:46 | <kennyluck> | foolip, no I don't think it's common. But still, there's better chance for me to visit www.ptt.cc, which is the archive of the BBS I talked about (currently 80000 people online), than any HK site ever. |
| 09:47 | <foolip> | kennyluck, does that site only work properly in Firefox? |
| 09:51 | <kennyluck> | foolip, if you mean "the Japanese is correctly rendered", then Yes. But in reality not many people use Japanese in Taiwan anyway. I visit that forum, which is for discussions around Anime (admittedly *not* the WWW archive) quite often though. |
| 09:51 | <kennyluck> | I think it just somehow reaches a point that nobody cares too much. If the site cares about compatibility, it would just use utf-8. |
| 09:52 | <kennyluck> | But really, displaying more characters is like rendering effects. If there's script that blocks the whole functionality of a site than it indeed sounds more serious. |
| 09:53 | <kennyluck> | s/script/script that relies on big5 decoded as big5-hkscs/ |
| 10:03 | <annevk> | kennyluck: sites written towards one browser will of course start rendering incorrectly |
| 10:04 | <annevk> | kennyluck: that's why you shouldn't code towards one browser, because browsers will typically evolve to work with most sites, rather than a few coded towards them |
| 10:06 | <kennyluck> | annevk, I think those who are knowledged about this problem would just tell you that those who create site like www.ptt.cc are not targeting specific browser, they just dump their database, which happened to be 'big5-uao' in reality, and tag the page as 'big5'. |
| 10:07 | <kennyluck> | I believe it's a fact that 'big5-uao' is more popular than 'big5-hkscs' in Taiwan, but I don't have data. |
| 10:08 | <annevk> | so how does IE render those sites correctly? |
| 10:08 | <annevk> | it seems you're saying people have installed specific fonts |
| 10:09 | <annevk> | or that Firefox's market share is misrepresented maybe |
| 10:11 | <kennyluck> | annevk, I have no idea. I have no idea how many percent of Taiwanese people install the 'big5-uao' package (which is irrelevant to fonts) either. |
| 10:13 | <kennyluck> | (Those who install the 'big5-uao' package on Windows will see 'big5' as 'big5-uao'.) |
| 10:13 | <annevk> | doesn't sound like it's irrelevant to fonts to me |
| 10:15 | <annevk> | http://bugs.ruby-lang.org/issues/show/1784 only mentions ptt.cc |
| 10:19 | <annevk> | https://www.w3.org/Bugs/Public/show_bug.cgi?id=16303 *sigh* |
| 10:19 | annevk | goes to work on something else |
| 10:20 | <annevk> | kennyluck: do you have a reference for that package btw? |
| 10:21 | <annevk> | kennyluck: the data I've seen thus far suggests hkscs is a better mapping, with the notable exception of ptt.cc |
| 10:25 | <kennyluck> | annevk, or to be more precise, the forums in ptt.cc that use Japanese regularly, which probably accounts for less than 5%. |
| 10:26 | <kennyluck> | The best link I can get is http://zh.wikipedia.org/wiki/Unicode%E8%A3%9C%E5%AE%8C%E8%A8%88%E7%95%AB . |
| 10:33 | <Philip`> | hsivonen: I believe OpenType allows ligatures matching on sequences of up to 65535 glyphs (the only limit is the uint16 in the file format) |
| 10:34 | <Philip`> | hsivonen: (http://www.microsoft.com/typography/otspec/GSUB.htm#LSF1 is the relevant thing) |
| 10:34 | Philip` | doesn't know if any implementations are stricter than that |
| 10:34 | <foolip> | kennyluck, if ptt.cc is the only notable site that would suffer, would approaching them and asking them to fix the problem be an option? |
| 10:34 | <annevk> | kennyluck: that wikipedia page seems to suggest the Windows big5 mapping table is altered, but says nothing about newer versions of Windows |
| 10:37 | <annevk> | foolip: we should probably have a site compat issue on that |
| 10:37 | <foolip> | annevk, on Big5-UAO? |
| 10:37 | <foolip> | oh, on ptt.cc |
| 10:37 | <foolip> | I guess you mean Open the Web? |
| 10:38 | <annevk> | yeah, dunno about Unicode-at-on in general |
| 10:38 | <foolip> | annevk, I'm writing an email about this, one sec |
| 10:41 | <hsivonen> | Philip`: thanks |
| 10:46 | <annevk> | oh great |
| 10:47 | <annevk> | so HTML says when URL encode for query gets a character that cannot be encoded a "?" is to be used instead |
| 10:47 | <annevk> | Gecko uses UTF-8 bytes instead |
| 10:47 | <annevk> | Chrome uses &#...; instead! |
| 10:47 | <annevk> | Opera follows spec |
| 10:47 | <annevk> | Safari follows Chrome |
| 10:48 | <zcorpan> | iirc the spec matches (old?) ie |
| 10:49 | <annevk> | IE9 emits raw bytes (after I changed test to use standards mode to make it work at all) |
| 10:49 | <annevk> | source code |
| 10:49 | <annevk> | <!doctype html><meta charset=tis-620><a></a><script>var a = document.getElementsByTagName("a")[0]; a.href = "?\ufffd"; alert(a.href) </script> |
| 10:49 | <annevk> | not sure what raw bytes means actually |
| 10:49 | <annevk> | the alert just had the character there |
| 10:50 | <annevk> | if I encodeURI() it I get utf-8... |
| 10:51 | <zcorpan> | i guess you should try following the link and see what the server gets |
| 10:52 | <annevk> | pretty sure IE will just put raw bytes on there |
| 10:54 | <annevk> | unsolved problems from 2009 http://annevankesteren.nl/2009/03/urls |
| 10:55 | <annevk> | oh yes |
| 10:55 | <annevk> | http://lists.w3.org/Archives/Public/www-style/2009Mar/0321.html |
| 10:55 | <annevk> | "There are some small interop issues. IE6 (not sure about newer versions) sends the query string as raw UTF-8 bytes rather than having them percent-escaped." |
| 11:01 | <kennyluck> | foolip, I think that's an option. I would guess not many people like this option though. Don't ask me for a reason because I truly don't know. But I think what would be more realistic and useful is to get IE people's commitment on shipping 'big5' to mean 'big5-hkscs'. Perhaps they have clear reasons for not doing so. |
| 11:03 | <kennyluck> | As my friend (Timothy Chien in the mailing list) told me, this might not be doable because IE might refuse to use anything besides the system's mapping tables. |
| 11:05 | <annevk> | hmm http://stackoverflow.com/questions/6763799/utf-8-encoding-issue-in-ie-query-parameters#comment8020739_6763799 |
| 11:05 | <annevk> | sounds like this is some kind of WTF area |
| 11:09 | <zcorpan> | should we make the spec look at namespaced attributes? http://software.hixie.ch/utilities/js/live-dom-viewer/saved/1460 |
| 11:11 | <annevk> | zcorpan: because? returning null there is fine no? |
| 11:14 | <zcorpan> | we look at elements with prefixes, but not at attributes with prefixes |
| 11:14 | <zcorpan> | seems inconsistent |
| 11:15 | <annevk> | that's only so you don't have to look for xmlns attributes |
| 11:15 | <annevk> | and only at xmlns:* attributes |
| 11:15 | <zcorpan> | the spec looks at xmlns attributes |
| 11:16 | <zcorpan> | the spec even picks up xmlns="" in no namespace; maybe that's bogus |
| 11:16 | <annevk> | you previously argued to do the simplest thing possible here |
| 11:16 | <zcorpan> | yeah, i did |
| 11:18 | <kennyluck> | annevk, I did some URL in CSS testing last week too → http://lists.w3.org/Archives/Public/www-style/2012Apr/0204 . I was very amused by how IE9 handles the url "%" :p |
| 11:18 | <zcorpan> | but innerHTML uses these methods |
| 11:19 | <zcorpan> | but i guess looking at namespaced attributes is overkill |
| 11:21 | <zcorpan> | maybe innerHTML shouldn't be using these methods in the first place |
| 11:22 | <Ms2ger> | You tell me :) |
| 11:23 | <annevk> | URLs doing decoding so massively different is annoying for testing decoders |
| 11:23 | <annevk> | well, the problem is mostly IE I guess |
| 11:24 | <annevk> | I wish Microsoft hang out in an IRC channel somewhere |
| 11:33 | <annevk> | oh sweet |
| 11:33 | <annevk> | IE is the worst |
| 11:34 | <annevk> | it does indeed use "?" over the wire |
| 11:35 | <annevk> | but to script it exposes the character itself |
| 11:36 | <annevk> | so some kind of normalization happens at the network layer, whereas other browsers do normalization upfront |
| 11:39 | <annevk> | so basically testing IE would involve network loads |
| 11:40 | <annevk> | fffffffuuuuuuuuuuuu |
| 12:04 | <nesta_> | annevk++ :) |
| 12:22 | <zcorpan> | and another quirk gets some tests http://simon.html5.org/test/quirks-mode/table-cell-width-calculation.html |
| 12:22 | <zcorpan> | time for coffee |
| 12:23 | <hsivonen> | Who has time to WONTFIX https://www.w3.org/Bugs/Public/show_bug.cgi?id=16711 with strong enough arguments to prevent escalation and reopening? |
| 12:25 | <annevk> | another question, who can make this fast: |
| 12:25 | <annevk> | results = [] |
| 12:25 | <annevk> | for(i = 0; i < 0x10FFFF; i++) { |
| 12:25 | <annevk> | t.href = "?" + cp_str(i) |
| 12:25 | <annevk> | results.push(uhex(i) + "\t" + t.search) |
| 12:25 | <annevk> | } |
| 12:25 | <annevk> | r.textContent = results.join("\n") |
| 12:25 | <Ms2ger> | zcorpan++ |
| 12:28 | <annevk> | until 0x10000 goes fast enough, but then CPU starts spinning mad |
| 12:28 | <zcorpan> | hsivonen: how about "Or else...?" :-) |
| 12:30 | <zcorpan> | annevk: what's cp_str? |
| 12:30 | <annevk> | function cp_str (cp) { |
| 12:30 | <annevk> | if(cp < 0x10000) |
| 12:30 | <annevk> | return String.fromCharCode(cp) |
| 12:30 | <annevk> | cp -= 0x10000 |
| 12:30 | <annevk> | return String.fromCharCode(0xD800 + (cp >> 10), 0xDC00 + (cp & 0x3FF)) |
| 12:30 | <annevk> | } |
| 12:32 | <Ms2ger> | return String.fromCharCode(0xD800 + (cp >> 10)) + String.fromCharCode(0xDC00 + (cp & 0x3FF)), maybe? |
| 12:34 | <annevk> | that is faster? |
| 12:35 | <annevk> | hsivonen: added a comment |
| 12:36 | <annevk> | hsivonen: and resolved INVALID for good measure |
| 12:36 | <Ms2ger> | It could be, up to you to test :) |
| 12:38 | <annevk> | hmm |
| 12:39 | <gsnedders> | annevk: What browser are you testing in? |
| 12:40 | <gsnedders> | Actually, nvm, won't hit that bug anyway. |
| 12:40 | <annevk> | Firefox/Opera Next/Chrome |
| 12:40 | <annevk> | CPU is spinning mad |
| 12:41 | <annevk> | Chrome / Opera done |
| 12:41 | <annevk> | Opera responsive |
| 12:41 | <annevk> | in Chrome the page cannot really be used |
| 12:41 | <annevk> | Firefox is not done, but has some kind of responsive page |
| 12:42 | <annevk> | hmm, looks like Firefox quit at E5F33 ?%F3%A5%BC%B3 |
| 12:42 | <hsivonen> | annevk: thanks |
| 12:44 | <annevk> | I guess once I add a way to filter out results that encode per default error handling ("?", "&#...", URL encoded utf-8 bytes) this might be okay |
| 12:44 | <zcorpan> | annevk: maybe you can have two nested loops to generate all possible surrogate pairs |
| 12:44 | <annevk> | Internet Explorer would still not be tested but bah |
| 12:45 | <annevk> | zcorpan: seems so unlogical that simple math would be the bottleneck here |
| 12:45 | <annevk> | illogical* |
| 12:46 | Ms2ger | points at the topic |
| 12:46 | <foolip> | kennyluck, are you sure that ptt.cc is actually purely Big5-UAO? Isn't it more likely that it's a random mix of Big5-HKSCS and Big5-UAO depending on what software/browser was used to post? |
| 12:47 | <hsivonen> | passing through bytes provided by whoever posted seems scary |
| 12:48 | <foolip> | hsivonen, indeed, but surely it's what 99% of the Web does? |
| 12:48 | <hsivonen> | 99% of the Web is scary |
| 12:49 | foolip | nods |
| 12:49 | <kennyluck> | foolip, indeed. It would depend on what software is used to post, which this is case are telnet clients. As I said, only 5% of the sub-forums are affected by this. |
| 12:49 | <hsivonen> | if there's something good about Java, it's that it converts to UTF-16 upon input, so you don't get to pass through bytes |
| 12:50 | <Ms2ger> | That's a pretty big 'if' :) |
| 12:50 | <foolip> | kennyluck, do you have any contact with the site operators? it would be interesting to analyze a random selection of pages that use the byte sequences in question... |
| 12:53 | <zcorpan> | var high, low; |
| 12:53 | <zcorpan> | for (high = 0xd800; high < 0xdc00; ++high) { |
| 12:53 | <zcorpan> | for (low = 0xdc00; low < 0xe000; ++low) { |
| 12:53 | <zcorpan> | doStuff(String.fromCharCode(high, low)); |
| 12:53 | <zcorpan> | } |
| 12:53 | <zcorpan> | } |
| 12:54 | <foolip> | zcorpan, was that for me? |
| 12:54 | <zcorpan> | for annevk |
| 12:55 | <kennyluck> | foolip, yes. I do, actually, a friend of a friend. What would be best way to do this? Let me ask for a mail address for me. But what kind of analysis do you want to do? If you just want a random selection of the pages, the subpages of http://www.ptt.cc/bbs/C_Chat/index.html that has Japanese in it would do. Or do you want to get a list of softwares that are used to post to PTT? |
| 12:56 | <kennyluck> | I think a bunch of them are open source telnet clients. |
| 12:56 | <kennyluck> | s/for me/for you. |
| 12:57 | <foolip> | kennyluck, ideally, a list of every single page that uses any of the byte sequences that differ between hkscs and uao, but if that's hard to produce I guess a big enough random selection should allow me to find them myself |
| 12:57 | <annevk> | I thought Firefox was supposed to have a non-blocking UI? |
| 12:57 | <foolip> | kennyluck, even better would of course be to ask if they're aware of the problem and if they would be prepared to make some changes to fix it |
| 12:59 | <kennyluck> | foolip, that is… well, not very easy I would assume. It is basically a non-profit site, partly run by National Taiwan University, and I never think they actively maintain the underlying BBS software. |
| 13:00 | <foolip> | kennyluck, so perhaps just scraping it ourselves would be easier? |
| 13:00 | <foolip> | kennyluck, are there sub-forums in or about Cantonese? |
| 13:00 | <kennyluck> | foolip, good point. Yes. |
| 13:01 | <kennyluck> | It's basically a community site for everything. |
| 13:01 | <foolip> | kennyluck, do I need a special version of Windows to test if these pages display properly in non-Firefox browsers, with special font support? |
| 13:02 | <annevk> | even the U+0000 - U+007F range has some incompatible stuff, but that's due to using URLs I guess |
| 13:04 | <kennyluck> | foolip, I have no idea. I've been not using Windows for a while, but I would guess you can install the big5-uao package on every Windows. It pretty much just replaces the system's big5 encoding table, or this is what I heard. |
| 13:05 | <foolip> | kennyluck, do you have any idea where one would find this package? Googling "big5-uao" mostly finds Mozilla-related things... |
| 13:05 | <foolip> | In any case, unless this is installed on >50% of Taiwan Windows computers, I really don't see how it could change matters much... |
| 13:06 | <kennyluck> | foolip, you can google the phrase "unicode補完計畫". This is one page teaching this → http://www.techbang.com/posts/3350-let-win7-perfect-display-japanese-web-page |
| 13:07 | <foolip> | kennyluck, ah, that's the "Unicode 補完" you mentioned in your mail (which was very helpful, thank you!) |
| 13:08 | <kennyluck> | Wow, the steps are crazy. I wonder how many people go throughout these. |
| 13:10 | <foolip> | kennyluck, yeah, replacing CP950 in 18 steps really doesn't seem like something many would do... |
| 13:11 | <annevk> | that it's available does suggest the problem might be more widespread :/ |
| 13:12 | <foolip> | kennyluck, also, the example used on that page (forum.tw.fdzone.org/viewthread.php?tid=324495) looks like it's only Japanese kana, which big5-hkscs would fix too |
| 13:12 | <foolip> | I think we should try to find a better sample of Taiwanese pages, but using what source? |
| 13:13 | <kennyluck> | Yeah, I wonder why people don't teach big5-hkscs instead. MS people are clearly not doing good evangelism. |
| 13:14 | <annevk> | all you get back from MS people is "use utf-8 or utf-16" |
| 13:14 | <annevk> | even if you ask questions specific to big5 :) |
| 13:14 | <foolip> | kennyluck, any advice you can give about how to proceed in order to minimize breakage for Taiwan users would be most helpful... :/ |
| 13:15 | <foolip> | annevk, I assume that we don't have any locale-dependent mappings and that you won't consider spec'ing that? |
| 13:16 | <annevk> | if that's what's required we'll do that |
| 13:16 | <annevk> | currently we don't have that |
| 13:16 | <foolip> | Mr. Pragmatic :) |
| 13:17 | <foolip> | I know I don't want it, because I use both Hong Kong and Taiwan sites with an English locale... |
| 13:18 | <annevk> | it sounds pretty sucky indeed |
| 13:19 | <annevk> | but we already have local-dependent defaults |
| 13:20 | <kennyluck> | Yeah, i am pretty sure what they (Yuan and Timothy) want is what I said: "big5"="big5-uao" for zh-TW browsers. |
| 13:24 | <kennyluck> | foolip, ok, so in case you are interested. PTT has a UTF-8 gateway "ssh bbsu⊙pc", which transcodes the content correctly for the page I gave you. This means that changing http://www.ptt.cc/ shouldn't be very hard *in theory*. |
| 13:26 | <kennyluck> | But really, I think the best way to make progress on this is to consult MS people working in Taiwan. They probably understand the problem better, esp. why they don't try to stop the grassroot big5-uao effort... |
| 13:34 | <annevk> | I reached out to their encoding expert, but he couldn't offer much help |
| 13:35 | <annevk> | kennyluck: see http://lists.w3.org/Archives/Public/www-archive/2012Mar/thread.html#msg46 |
| 13:38 | <kennyluck> | annevk, foolip, hold on. I need to understand this a bit more. When you say IE treat 'big5'='big5-hkscs'. Do they render the Kanas or not? (Kanas are not rendered correctly in by Win7 Laptop, I wonder if this is locale-dependent.) |
| 13:39 | <foolip> | kennyluck, they have a single mapping that uses PUA, what it renders like really depends on the installed fonts |
| 13:39 | <annevk> | for windows-874 Chrome maps fullwidth ASCII back to ASCII, for windows-1252 they don't |
| 13:39 | <foolip> | even though the MingLiu_HKSCS font comes with Win7 by default, I'm not sure if it's used by IE |
| 13:40 | <kennyluck> | foolip, OK. now I see why people want to install the big5-uao package. |
| 13:41 | <foolip> | I think that a lot of these pages don't render correctly in most browsers, but we want a proper mapping to Unicode to make them work as well as they can... |
| 13:41 | <kennyluck> | Is there a similar package that does the *correct big5-hkscs* which doesn't map to PUA (HK-2008?) |
| 13:43 | <foolip> | kennyluck, Microsoft has stopped providing the hacked code page and just ship MingLiu_HKSCS with Win7, but I'm not exactly sure if any special steps are needed to make IE use it |
| 13:43 | <foolip> | A Hong Kong user would probably know. |
| 13:44 | <kennyluck> | foolip, yeah. I mean, as a normal user, I would be happy to learn something simple just to turn the Kanas on instead of the 18 steps 'big5-uao'. |
| 13:55 | <kennyluck> | Or I have the feeling that fixing this part of evangelism might be just too difficult. I wonder how bad it is if we just do locale-dependent 'big5'='big5-uao'. If we don't fix the whole thing, we'll just go into a suboptimal that "いま俺の顔生涯最高にキモい自信がある" (uao) is rendered as "いま俺の𡟺生涯最高にキモい自信がある" (hkscs) |
| 13:56 | <annevk> | unless special fonts are used big5-uao is already broken in Chrome |
| 13:56 | <annevk> | and Internet Explorer |
| 13:57 | <annevk> | combined they have about 80% of the Taiwan market |
| 13:57 | <annevk> | Chrome about 15% |
| 13:57 | <annevk> | Firefox 10% last I looked |
| 13:58 | <annevk> | and Chrome does not use the table from Windows |
| 13:58 | <foolip> | kennyluck, the question is really if that page already works for a majority of Taiwan users. If it only works in Firefox, I'm not sure we should introduce locale-dependent mappings in order to fix it |
| 13:59 | <annevk> | yeah, it seems better that all browsers break it so that other pages become less broken and there's an incentive to fix the content |
| 14:00 | <foolip> | does anyone have a good source for the top million sites or something that I can scrape for taiwanese sites? |
| 14:00 | <foolip> | My hunch is that even in Taiwan, using the big5-hkscs mapping will be correct more often. |
| 14:01 | <annevk> | there's http://s3.amazonaws.com/alexa-static/top-1m.csv.zip but it's only front pages |
| 14:02 | <kennyluck> | I fear that it's not just the content that's broken. The fact that we still see pages educating people to install the big5-uao package is not neglectable. |
| 14:03 | <foolip> | kennyluck, I guess all we can do right now is speculate on how big the problem is. I'll try to get some kind of random scrape of the Taiwan Web... |
| 14:03 | <kennyluck> | foolip, I don't even find a clue on how to turn on big5-hkscs in IE as 'big5'. How could that be true? |
| 14:04 | <foolip> | kennyluck, it's always the case, big5 always maps to the same PUA code points |
| 14:04 | <foolip> | and the MingLiu_HKSCS font shipped with Win7 has glyphs for those code points |
| 14:05 | <foolip> | I don't know what font you need to make those PUA code points match Big5-UAO |
| 14:05 | <annevk> | kennyluck: I just read that article, did you see how in the end it says in Chrome you should simply select the big5-hkscs override? |
| 14:05 | <annevk> | kennyluck: and that it's pretty much only written for PTT |
| 14:06 | <kennyluck> | I need to go back to that page again... |
| 14:06 | <annevk> | so if even the Taiwanese suggest big5-hkscs in Chrome, it seems like using that by default would be an improvement |
| 14:06 | <annevk> | here is the URL http://www.techbang.com/posts/3350-let-win7-perfect-display-japanese-web-page |
| 14:08 | <hsivonen> | so sad that PUA is used for de facto standardization |
| 14:08 | <hsivonen> | not surprising of course |
| 14:08 | <foolip> | hsivonen, killing it for Big5 looks in the real of the possible, fortunately |
| 14:09 | <hsivonen> | PUA is the vendor prefixes of Unicode |
| 14:09 | <foolip> | yay, http://s3.amazonaws.com/alexa-static/top-1m.csv.zip has 2930 .tw domains |
| 14:10 | <zcorpan> | now do a site: search for each domain to get the top n pages of each domain :-) |
| 14:11 | <foolip> | zcorpan, can that be automated? |
| 14:11 | <zcorpan> | dunno, i guess most search engines have anti-DOS measures |
| 14:11 | <foolip> | obviously, ptt.cc is not on the list of top .tw sites, but I don't know how else to make a list like this :/ |
| 14:12 | <zcorpan> | hmm, don't some search engines have a feature to search for pages in a particular language? |
| 14:12 | <foolip> | I would love a proxy server that saved all requests to disk so I could just let my web browser visit these pages, follow a few links, and then I have the pages to analyze on disk... |
| 14:17 | <Philip`> | zcorpan: Pick a common word in that language and then search for it? |
| 14:21 | <annevk> | weehee, all encoders are slightly different |
| 14:22 | <annevk> | ^^ sarcastic |
| 14:28 | <hsivonen> | looks like Qt got memes before Opera or IE: http://qtmemes.tumblr.com/ |
| 14:30 | <Philip`> | It seems to be getting quite metamemetic |
| 14:30 | <zcorpan> | well that'll be today's quirk testing |
| 14:31 | <annevk> | Opera: [red penguin] Follows standards. [blue pinguin] Alone. |
| 14:33 | <kennyluck> | foolip, so this is what I think as a user. 1) I open http://www.toysdaily.com/discuz/forum-24-2.html 2) I can't see the Kana presumably because of reasons related to fonts 3) I happen to know and only know 'big5-uao' as the way to turn Kana on 4) I start outputting Japanese content because I can read it now 5) I now output 'big5-uao'. |
| 14:33 | <Ms2ger> | http://qtmemes.tumblr.com/post/20183979051/we-have-qtwebkit-that-means-we-count-right |
| 14:33 | <kennyluck> | So yeah, intercepting 2) seems workable, but it really depends on IE. |
| 14:34 | <kennyluck> | Otherwise, I would still be tempted to install 'big5-uao' and then create incompatible content. |
| 14:35 | <annevk> | you cannot output big5-uao in Gecko I think |
| 14:35 | <annevk> | their encoder is restricted |
| 14:35 | <kennyluck> | annevk, ah. Good point. |
| 14:41 | <annevk> | also, Chrome users are apparently advised to just use big5-hkscs |
| 14:42 | <annevk> | Firefox users are not mentioned... |
| 14:45 | <kennyluck> | annevk, in any case I agree that expanding big5 to big5-hkscs or the intersection of big5-hkscs and big5-uao is making good progress, but I am not sure if it's optimal. |
| 14:47 | <annevk> | foolip: did you look at "firefox" vs "firefox-hk"? |
| 14:47 | <annevk> | foolip: ignoring PUA |
| 14:47 | <smaug____> | rniwa: ping |
| 14:48 | <smaug____> | rniwa: where should I send comments about undomanager |
| 14:48 | <annevk> | whatwg⊙wo has been used thus far |
| 14:49 | <kennyluck> | (re. you cannot output big5-uao in Gecko I think) But I might still output 'big5-uao' from IE. I think the extended mapping installed by the package isn't unidirectional like Gecko, but I am not very sure. |
| 14:50 | <annevk> | oh yeah, with that package IE can do damage |
| 14:55 | <annevk> | I guess we should figure out a) what's incompatible between hkscs and uao and b) how widespread uao is. |
| 14:58 | <annevk> | but not today, I have some things to do |
| 15:05 | <gsnedders> | hsivonen: I wonder how much would break moving away from the PUA for Big5. |
| 15:24 | <annevk> | gsnedders: everything that would break would already be broken in Opera; and the pages we looked at would work better with hkscs |
| 17:01 | <jsbell_> | http://www.corp.google.com/~jsbell/rampart - added left/right mouse click to place/rotate (in addition to space/control); added territory capture. (algorithm requires you can't surround the "seed" location, hence bisecting water) |
| 17:02 | <jsbell_> | Whoops, that would be the wrong channel |
| 17:03 | <jsbell> | (curses, our team's top secret plans to dominate the world via an obsession with an obscure 90's arcade game have been leaked) |
| 17:16 | <[tm]> | jsbell: 'you can't surround the "seed" location, hence bisecting water' |
| 17:17 | <[tm]> | I'm pretty sure you stole that from a Wallace Stevens poem |
| 17:18 | <jsbell> | heh |
| 17:29 | <rniwa> | smaug____: hi, i just replied to your message |
| 17:29 | <rniwa> | smaug____: sorry, I forgot to close my IRC client :\ |
| 17:30 | <smaug____> | rniwa: thanks. It was just a first quick read-through |
| 17:33 | <rniwa> | smaug____: btw, since i've started hosting it on w3c repository, it might make sense for us to make public-webapps the place for discussion |
| 17:33 | <rniwa> | AryehGregor: what do you think? |
| 17:33 | <rniwa> | sadly, nobody had replied to my email about chartering undomanager in the editing community |
| 17:33 | <rniwa> | so not sure what's happening there |
| 17:33 | <rniwa> | but... |
| 17:34 | <smaug____> | rniwa: yeah, webapps should be ok |
| 17:34 | <Ms2ger> | rniwa, as long as you don't make it public-html :) |
| 17:34 | <smaug____> | (If I could get File system out from webapps, and undomanager in...) |
| 17:35 | <Ms2ger> | Did you comment on the charter? :) |
| 17:35 | <smaug____> | oh, yes, if the discussion happened in public-html, I would promise to not send any comments :) |
| 17:41 | <rniwa> | Ms2ger: I thought I did but maybe I didn't use that particular word or wasn't clear about it :\ |
| 17:42 | <Ms2ger> | rniwa, hmm? |
| 17:43 | <smaug____> | rniwa: so, input element and textarea could just always have undomanager |
| 17:44 | <smaug____> | rniwa: DOMTransactionEvent doesn't really make it more undo related... |
| 17:44 | <rniwa> | smaug____: i know. |
| 17:45 | <rniwa> | smaug____: we should rename it to something like UndoRedoEvent |
| 17:45 | <smaug____> | rniwa: where is the event dispatched? |
| 17:45 | <smaug____> | to the Undomanager itself? |
| 17:45 | <smaug____> | or to the element? |
| 17:45 | <rniwa> | smaug____: to the element |
| 17:45 | <smaug____> | hmm |
| 17:45 | <rniwa> | smaug____: "When the user agent is required to fire a DOM transaction event for a DOM transaction t at an undo scope host h, the user agent must run the following steps:" |
| 17:45 | <smaug____> | right |
| 17:45 | <rniwa> | smaug____: so that it can bubble |
| 17:46 | <smaug____> | why does it need to bubble? |
| 17:46 | <smaug____> | nm |
| 17:46 | <rniwa> | smaug____: so... an important use case is to do something in response to undo/redo |
| 17:46 | <smaug____> | yes, it should bubble |
| 17:46 | <rniwa> | smaug____: for that, you don't necessary want to attach event listeners on all elements with undoManager |
| 17:48 | <rniwa> | smaug____, Ms2ger: since you're already here... do you know if Mozilla imports W3C tests and create reference files for them? |
| 17:48 | <rniwa> | smaug____, Ms2ger: we want to import CSS test suite but don't want to add thousands of pixel results. |
| 17:49 | <Ms2ger> | I've written some for 2.1 a while back |
| 17:49 | <smaug____> | I don't know about css tests |
| 17:49 | <rniwa> | Ms2ger: oh yeah? |
| 17:49 | <rniwa> | Ms2ger: do you know where they're located? |
| 17:49 | <Ms2ger> | Boring ones, for padding, IIRC |
| 17:49 | <Ms2ger> | https://bitbucket.org/ms2ger/css-tests |
| 17:49 | <rniwa> | Ms2ger: if you already have reference files, ideally, we don't want to re-invent reference files ourselves |
| 17:50 | <rniwa> | since Mozilla folks surely have more experience writing reference files |
| 17:50 | <Ms2ger> | I need to figure out how to get them reviewed and into the WG's repo |
| 17:50 | <rniwa> | Ms2ger: oh, so they're not in Mozilla's repository? |
| 17:50 | <Ms2ger> | No |
| 17:50 | <rniwa> | okay. so i guess we have the same problem then. |
| 17:51 | <Ms2ger> | We don't import any non-reftest CSS tests, I don't think |
| 17:51 | <rniwa> | Ms2ger: okay. makes sense. |
| 17:51 | smaug____ | tries to figure out some non-DB-related synonym for transaction |
| 17:51 | <rniwa> | smaug____: we use UndoStep internally in WebKit |
| 17:52 | <smaug____> | that sounds ok |
| 17:52 | <rniwa> | annevk: hi annevk |
| 17:52 | <smaug____> | in gecko we do use transactions |
| 17:52 | <smaug____> | rniwa: undomanager is TransactionManager |
| 17:52 | <rniwa> | smaug____: yeah, but it's getting quite confusing in the world where we have IDB's transaction :\ |
| 17:53 | <rniwa> | smaug____: we could just throw in "undo" prefix as well |
| 17:53 | <smaug____> | yeah. undo/redo/undoupdated ... |
| 17:53 | <rniwa> | so like undo-transaction |
| 17:53 | <annevk> | rniwa: good evening |
| 17:53 | <rniwa> | annevk: does opera import CSS2.1 test suite as pixel tests? (i.e. generate images)? |
| 17:54 | <rniwa> | annevk: we want to import newer css test suite but we've realized that they don't have reference files |
| 17:54 | <rniwa> | and we don't really want to generate thousands of png files :( |
| 17:54 | <annevk> | yeah, I objected to the CSS WG doing that |
| 17:54 | <annevk> | and then the whole group got mad |
| 17:54 | <annevk> | and glazou blamed Opera |
| 17:54 | <annevk> | and then they went ahead with their pixel tests instead of doing the test suite right... |
| 17:54 | <rniwa> | annevk: :( |
| 17:55 | <annevk> | gsnedders prolly knows what we do internally |
| 17:55 | <rniwa> | annevk: what if we said we don't want to import tests that are not reftests? |
| 17:55 | <annevk> | I think TabAtkins actually said it didn't matter to Google |
| 17:55 | rniwa | wonders if webkit community's decision will have an impact on CSS WG |
| 17:55 | <annevk> | but that would be a wise change of position I think |
| 17:56 | <rniwa> | annevk: yeah, pixel results make very little sense |
| 17:56 | <Ms2ger> | rniwa, the policy for CSS3 is reftest-only, IIRC |
| 17:56 | <rniwa> | annevk: it incurs way too much maintenance cost. |
| 17:56 | <rniwa> | Ms2ger: good! |
| 17:56 | <annevk> | Microsoft didn't care, and Gecko/WebKit people were kind of holding back because glazou blew up |
| 17:56 | <rniwa> | TabAtkins: do you know the details? |
| 17:56 | <astearns> | Ms2ger: it's reftest or testharness.js now |
| 17:56 | <annevk> | astearns: sweet |
| 17:57 | <rniwa> | annevk: I see. |
| 17:57 | <rniwa> | Ms2ger, annevk: it might make sense for Gecko/WebKit people to push CSS WG to have reference files |
| 17:57 | <Ms2ger> | astearns, well, requiring reftests for JS tests would be silly |
| 17:58 | <rniwa> | since neither of us want to import tests that are not reftests |
| 17:58 | <astearns> | Ms2ger: more silly than not allowing JS tests at all |
| 17:58 | <rniwa> | Ms2ger: testharness.js aren't reftests, right? |
| 17:58 | <astearns> | rniwa: correct |
| 17:59 | <astearns> | rniwa: and we just checked in testharness.js support in WebKit |
| 18:01 | <Ms2ger> | And so did I for Gecko :) |
| 18:01 | <Ms2ger> | rniwa, I can't speak for Mozilla, and I barely do anything to do with CSS, but you have my vote :) |
| 18:07 | <gsnedders> | rniwa: We have it as screenshot tests, yes |
| 18:08 | <gsnedders> | rniwa: We've converted some to reftests, but not entirely |
| 18:08 | <gsnedders> | rniwa: Never submitted to the WG after it became kinda obvious the WG didn't want them lest it delay REC |
| 18:08 | <rniwa> | gsnedders: :( |
| 18:08 | <rniwa> | gsnedders: that's my fear was well. |
| 18:09 | <gsnedders> | rniwa: The policy for CSS3 is reftest only, pretty much |
| 18:09 | <rniwa> | gsnedders: it seems like they won't accept our patches to add reference files even if we submitted them |
| 18:09 | <rniwa> | gsnedders: do you have your reference files publicly available somewhere? |
| 18:09 | <gsnedders> | rniwa: You can try, and argue that these tests will likely become part of CSS3 module test suites |
| 18:09 | <gsnedders> | rniwa: I don't think so. |
| 18:10 | <Ms2ger> | Now that 2.1 is a rec, I guess we can get them in |
| 18:10 | <gsnedders> | rniwa: But group the tests by their screenshot, and you realize CSS 2.1 test suite is in large part identical references. |
| 18:10 | <Ms2ger> | (Some of the refs in my repo are gsnedders's, I should note) |
| 18:10 | <gsnedders> | Ms2ger: Yeah, I guess, but I've moved on from caring about it now. |
| 18:11 | <gsnedders> | I've argued for this before, and I can't be fucked fighting to get them in. |
| 18:11 | <Ms2ger> | Anyway, if the WG doesn't want them, I'd prefer a shared fork, though |
| 18:11 | <rniwa> | Ms2ger: yeah, that's what i'm getting at |
| 18:12 | <rniwa> | Ms2ger: there's no point for each of us to re-do all the work |
| 18:12 | <gsnedders> | If you don't want to take the work that people have done, even though your requirements insist on them in future, then that's your own damned problem. |
| 18:12 | <rniwa> | Ms2ger: if we can just share the results |
| 18:12 | <gsnedders> | rniwa: That's the exact argument I made to the mailing list two years ago. |
| 18:12 | <gsnedders> | Nobody cared. |
| 18:12 | <gsnedders> | Literally. |
| 18:12 | <rniwa> | gsnedders: :( |
| 18:12 | <rniwa> | gsnedders: that's astoundingly annoying |
| 18:12 | <gsnedders> | Not even Mozilla people, who were the only others at the time to have a running reftest system. |
| 18:13 | <gsnedders> | WebKit people were inteested, but mainly in an in-the-future way. |
| 18:13 | <rniwa> | gsnedders: it's possible that future has come :D |
| 18:13 | <rniwa> | gsnedders: since we DO support reftests now |
| 18:13 | <astearns> | the future is here! rniwa has added support for reftests :) |
| 18:13 | <gsnedders> | Some of my refs may well assume 96dpi on tests that shouldn't. |
| 18:13 | <gsnedders> | astearns: Yeah, I'm well aware. :) |
| 18:14 | <rniwa> | gsnedders, astearns: we've had our own support for reftests, but we only added the support for W3C style reftests last winter |
| 18:14 | <gsnedders> | rniwa: Search for emails from me to public-css-testsuite and w3c-css-wg if you want the background |
| 18:14 | <rniwa> | gsnedders: okay. |
| 18:14 | <gsnedders> | rniwa: Yeah, you didn't even have that when I was working on this :) |
| 18:15 | <rniwa> | gsnedders, Ms2ger: anyway, it'll be great if we could share reference files even if W3C doesn't accept them |
| 18:15 | <rniwa> | gsnedders, Ms2ger: for webkit, I want to make sure we don't invent our own reference files that don't adequately exercise tests |
| 18:15 | <gsnedders> | rniwa: https://lists.w3.org/Archives/Member/w3c-css-wg/2010JulSep/0222.html is where I gave up |
| 18:15 | <gsnedders> | (Member only, etc) |
| 18:15 | <rniwa> | gsnedders: oh... it's member only |
| 18:16 | <rniwa> | gsnedders: i hear member-only mailing lists are terrible places to live in |
| 18:17 | <Ms2ger> | I hear some people are more polite if their emails are archived in public |
| 18:18 | <jamesr> | i hear some people are more polite if they aren't glazou |
| 18:18 | <Ms2ger> | No comment |
| 18:18 | <gsnedders> | rniwa: I may have sent some to public-css-testsuite |
| 18:19 | <gsnedders> | (chaals's response to that email is the truth, for those of you who can read them) |
| 18:19 | <gsnedders> | (references, that is) |
| 18:19 | <Ms2ger> | rniwa, I'm happy to help out writing some references, btw, but you'll probably have to poke me :) |
| 18:20 | <gsnedders> | rniwa: http://lists.w3.org/Archives/Public/public-css-testsuite/2010Sep/0030.html |
| 18:20 | <gsnedders> | "Attached is a diff to convert 830 tests to reftests (with a mere four |
| 18:20 | <gsnedders> | references!)." |
| 18:21 | <rniwa> | gsnedders: thanks! |
| 18:21 | <rniwa> | gsnedders, Ms2ger: I'll probably post something back on webkit-dev about this if you guys don't mind |
| 18:21 | <gsnedders> | The whole situation was ridiculous, really. |
| 18:21 | <rniwa> | we're having a big debate on how to import css tests |
| 18:21 | <gsnedders> | rniwa: tl;dr of the debate? |
| 18:21 | <Ms2ger> | rniwa, gsnedders, those are in my repo too |
| 18:21 | <rniwa> | and this reftest vs. pixel test thing is one major issue |
| 18:22 | <rniwa> | Ms2ger, gsnedders: so... we want to import tests as reftests |
| 18:22 | <gsnedders> | You don't want to go through all 10k tests and label them as pass or fail (or in your case, pass or xfail). |
| 18:22 | <rniwa> | but some of us (mainly me) don't want to invent our own reference files because there is a chance that our own reference files don't exercise the tests adaquately |
| 18:22 | <Ms2ger> | rniwa, isn't http://28.media.tumblr.com/tumblr_m284vvtgQT1rqvy12o1_1280.jpg argument enough in favour of reftests? :) |
| 18:23 | <rniwa> | e.g. there could be a bug that affects both the test and the reference file same way and end up hiding the bug |
| 18:23 | gsnedders | may have borrowed summer interns for a day to label them |
| 18:24 | <annevk> | Ms2ger: brilliant |
| 18:24 | <Ms2ger> | Note to self: don't do internship at Opera the summer after MS dumps its tests |
| 18:24 | <Ms2ger> | Also: http://28.media.tumblr.com/tumblr_m1yh3dbdHf1rqvy12o1_r1_1280.jpg |
| 18:24 | <gsnedders> | Ms2ger: Heh, we're almost certainly never going to bring in a large dump of screenshot-based tests ever again |
| 18:27 | <matjas> | could some of you take a look at https://bugs.ecmascript.org/show_bug.cgi?id=277#c2? /cc zcorpan |
| 18:27 | <matjas> | IIRC all browsers except Firefox respect http://wiki.whatwg.org/wiki/Web_ECMAScript#Identifiers but it’s been a while since I tested |
| 18:27 | <Ms2ger> | v\u0061r x = 0 |
| 18:27 | <Ms2ger> | Orly |
| 18:28 | <smaug____> | Ms2ger: that is good one |
| 18:28 | gsnedders | cringes at the thought of all that |
| 18:28 | rniwa | goes to his office |
| 18:28 | <Ms2ger> | matjas, unsurprisingly, Allen wasn't too happy to see you mention that :) |
| 18:29 | <matjas> | hey, I never said this should become part of ECMAScript |
| 18:29 | <matjas> | just how most browsers seem to do it |
| 18:30 | <gsnedders> | matjas: Then that's not relevant for es-discuss :P |
| 18:32 | <matjas> | I just mentioned it casually :) It was a Spidermonkey-specific thread anyway (as no other browser allows `function function() {}` etc.) |
| 18:32 | matjas | nerdrages |
| 18:32 | <gsnedders> | jgraham: Any idea whether you found the identifier nonsense needed for web compat? |
| 18:37 | <matjas> | var tru\u0065; /* "Expected identifier" error in IE9 */ console.log(fals\u0065) /* "Syntax error" in IE9 */ |
| 18:37 | <matjas> | so Fx and IE agree |
| 18:53 | <Ms2ger> | Looks like pointer lock landed in Gecko |
| 18:58 | <smaug____> | yup |
| 19:38 | <Hixie> | abarth: any movement on http://www.w3.org/2011/webappsec/track/issues/6 ? |
| 19:39 | <abarth> | Hixie: we talked about it in the telecon yesterday |
| 19:39 | <Hixie> | so i should still do the refactoring? |
| 19:39 | <abarth> | as far as I can tell, everyone wants it, but moz hasn't implemented it and is trying to block the spec from progressing until they finish |
| 19:39 | <Hixie> | just making sure :-) |
| 19:40 | <Hixie> | abarth: ok so my plan with sandbox flags is as follows: |
| 19:40 | <Hixie> | - make it so that browsing contexts have the set of flags |
| 19:41 | <Hixie> | - make it so that documents have two copies of the set of flags: one called something like "the CSP flags" and one called the "effective flags" |
| 19:41 | <Hixie> | the effective flags is the boolean-or of the flags from the CSP flags, the browsing context flags, and all ancestor browsing context flags |
| 19:41 | <Hixie> | hm no that doesn't work |
| 19:41 | <Hixie> | csp needs to also affect decendant browsing contexts |
| 19:41 | <Hixie> | let me try again |
| 19:42 | <Hixie> | plan: |
| 19:42 | <Hixie> | hmm |
| 19:43 | <Hixie> | the basic idea is i move the real set of flags to Document |
| 19:43 | <Hixie> | so all the security checks use that |
| 19:43 | <Hixie> | and those are set by being fed from the various other sources |
| 19:43 | <Hixie> | CSP, containing <iframe>, and ancestor documents |
| 19:44 | <Hixie> | ok there we go, that's how we do it |
| 19:44 | <Ms2ger> | Die, IDBDatabaseException and IDBDatabaseException, die |
| 19:44 | <Hixie> | abarth: does that work for you? |
| 19:44 | <abarth> | (Sorry, I was replying to ojan on the list. reading now) |
| 19:45 | <Hixie> | no worries |
| 19:45 | <abarth> | yes |
| 19:45 | <abarth> | that works exactly the same was as the code |
| 19:46 | <Hixie> | excellent |
| 19:46 | <Hixie> | ok |
| 19:46 | <Hixie> | that should be easy to spec |
| 19:46 | <abarth> | when the document is created, it boolean-ors together a bunch of sandbox flags from various sources |
| 19:46 | <Ms2ger> | Also, getting script crossorigin specced would be nice |
| 19:46 | <Hixie> | (though confusing for people to read) |
| 19:46 | <abarth> | it seems pretty clean |
| 19:47 | <Hixie> | yeah it's just confusing because in specs it's hard to make multiple copies of the same thing understandable |
| 19:47 | <Hixie> | much easier to do in code |
| 19:47 | <annevk> | Ms2ger: indexed DB, the slightly better DB (but is it really?), with lots of minor API issues |
| 19:48 | <Ms2ger> | annevk, hmm? |
| 19:49 | <annevk> | geez, BlobBuilder is still around |
| 19:49 | <annevk> | !summon arun |
| 19:50 | <Ms2ger> | Is that in HG already? |
| 19:51 | <annevk> | oh it's not Arun |
| 19:51 | <annevk> | it's from Eric |
| 19:51 | <annevk> | http://dev.w3.org/2009/dap/file-system/file-writer.html |
| 19:52 | <annevk> | but I guess the idea was that it was going to http://dev.w3.org/2006/webapi/FileAPI/ |
| 19:52 | <annevk> | which has a constructor defined |
| 19:52 | <Ms2ger> | Ah, /dap/ |
| 20:10 | <jgraham> | gsnedders: No idea |
| 20:12 | <jgraham> | rniwa: If you never learn anything else from Opera, learn that pixel based testing is a special form of hell |
| 20:13 | <rniwa> | jgraham: we also do have pixel tests :) |
| 20:13 | <rniwa> | jgraham: i know it's hard to maintain and all |
| 20:13 | <jgraham> | Probably not in the same way we do :) |
| 20:13 | <rniwa> | jgraham: but reftests have its own faults :\ |
| 20:13 | <rniwa> | jgraham: okay. |
| 20:13 | <rniwa> | jgraham: we have pixel results for each port + platform though |
| 20:14 | <rniwa> | jgraham: so we have something like 6-8 PNGs per test :( |
| 20:14 | <Ms2ger> | rniwa, I think Opera has more platforms to test :) |
| 20:14 | <jgraham> | We have a flexible system that allows an arbitary number of results per test :) |
| 20:15 | <rniwa> | Ms2ger, jgraham: http://trac.webkit.org/browser/trunk/LayoutTests/platform/ |
| 20:15 | <Ms2ger> | Oh, I know |
| 20:15 | <rniwa> | Ms2ger, jgraham: each one of them will have the same dir. structure as http://trac.webkit.org/browser/trunk/LayoutTests/ |
| 20:15 | <rniwa> | (except platform of course) |
| 20:16 | <rniwa> | and contains -expected.png and/or -expected.txt |
| 20:19 | <jgraham> | rniwa: I don't really understand the concern about not making "good enough" references |
| 20:20 | <jgraham> | What are you testing here (I didn't read all the backscroll) |
| 20:21 | <rniwa> | jgraham: so if we import a test from CSS test suite that doesn't have a reference |
| 20:21 | <rniwa> | jgraham: and create our own home-brew reference file |
| 20:21 | <rniwa> | jgraham: then there's a possibility that the reference file doesn't address all intents of the test |
| 20:21 | <rniwa> | jgraham: i.e. it could fail to detect some bugs that the test author intended to test |
| 20:31 | <rniwa> | jgraham: to be honest with you, i'm not even sure if this is a practical concern or not. |
| 20:31 | <rniwa> | jgraham: we simply lack adequate experience with reftests |
| 20:32 | <Ms2ger> | Fwiw, fantasai would be very happy to accept references |
| 20:32 | <rniwa> | Ms2ger: for css 3? |
| 20:32 | <Ms2ger> | 21 |
| 20:32 | <TabAtkins> | For all. |
| 20:33 | <Ms2ger> | TabAtkins, also, I'm looking forward to you speccing ::first-letter :) |
| 20:33 | <TabAtkins> | Sigh. I'll get there at some point. |
| 20:33 | <TabAtkins> | If fantasai doesn't beat me to it. |
| 20:34 | <Ms2ger> | * fantasai didn't want to touch pseudo elements |
| 20:34 | <rniwa> | Ms2ger: okay. |
| 20:34 | <Ms2ger> | I'm afraid you're out of luck on that count |
| 20:34 | <TabAtkins> | Damn. All right. |
| 20:34 | <TabAtkins> | Well, I'll do Box Model at some point. |
| 20:34 | <TabAtkins> | I'm thinking end of year. |
| 20:34 | <rniwa> | Ms2ger, TabAtkins: can we convince CSS WG to convert the existing tests to reference tests? |
| 20:34 | <rniwa> | it's a no laughing matter for us |
| 20:35 | <TabAtkins> | rniwa: They don't need convincing, just manpower. |
| 20:35 | <Ms2ger> | I believe the open source answer to that is "patches accepted" :) |
| 20:35 | <TabAtkins> | We *hate* our existing non-reftests. |
| 20:35 | <rniwa> | because we don't want to add a couple thousand * 8-10 PNG files to our repository |
| 20:35 | <rniwa> | TabAtkins: but we've already had a proposal: http://lists.w3.org/Archives/Public/public-css-testsuite/2010Sep/0030.html |
| 20:35 | <rniwa> | TabAtkins: why did we not do this back then? |
| 20:36 | <Ms2ger> | "It would delay REC" |
| 20:36 | <rniwa> | Ms2ger: so since it's it's already REC'ed, its okay to change it now? |
| 20:36 | <TabAtkins> | rniwa: I presume Ms2ger is right. |
| 20:36 | <TabAtkins> | And yes, now that 2.1 is back into errata mode, just poke at it again. |
| 20:36 | <Ms2ger> | Unless Microsoft invents a new objection |
| 20:39 | <Ms2ger> | Anyway, off for today |
| 20:39 | <rniwa> | Ms2ger: LOL |
| 20:39 | <rniwa> | Ms2ger: hopefully not. |
| 20:40 | <rniwa> | TabAtkins: okay, let's do that. |
| 20:40 | <rniwa> | TabAtkins: it's totally worthwhile our time. |
| 20:40 | <rniwa> | TabAtkins: since Ms2ger and others already have reference files! |
| 20:40 | <Ms2ger> | Not a lot ;) |
| 20:45 | <jgraham> | rniwa: Mostly it doesn't seem like a real issue |
| 20:45 | <rniwa> | jgraham: WHAT IS? |
| 20:45 | <jgraham> | The only problem is cases where there are multiple possible acceptable renderings |
| 20:45 | <rniwa> | what is?* |
| 20:45 | <rniwa> | jgraham: I see. yeah, that'll be tricky i guess. |
| 20:45 | <jgraham> | The reference not capturing the full intent of the test |
| 20:45 | <TabAtkins> | Nah, you can provide tests with multiple acceptable refs. |
| 20:45 | <jgraham> | I mean you could have the same problem with a pixel reference |
| 20:46 | <jgraham> | TabAtkins: Depends how your test system is set up |
| 20:46 | <TabAtkins> | Shepherd is set up to allow that. |
| 20:46 | <TabAtkins> | Since we're discussing CSS tests. ^_^ |
| 20:46 | <jgraham> | The Opera system doesn't cope with that at the moment |
| 20:46 | <jgraham> | Shepard is irrelevant |
| 20:46 | <jgraham> | (sorry peter) |
| 20:47 | <TabAtkins> | Is it Shepard or Shepherd? |
| 20:47 | <jgraham> | Specifically it's irrelevant to browser vendors who all have home-grown testing solutions designed to meet the requirements of running tests every day |
| 20:47 | <jgraham> | Ho you were right |
| 20:47 | <rniwa> | jgraham: I'd appreciate it if you could give us some insights into pros and cons of reftests |
| 20:47 | <jgraham> | I just can't spell and can't type |
| 20:47 | <jgraham> | so I have errors**2 |
| 20:48 | <rniwa> | jgraham: in general, we totally lack any sort of expertise working with ref tests |
| 20:48 | <rniwa> | since we hadn't had support 'til very recently |
| 20:48 | <rniwa> | and even today, people don't write as much ref tests as they should |
| 20:49 | <jgraham> | rniwa: Pros - They don't require changes when you do things like add a new platform or change your text rendering in some way |
| 20:50 | <jgraham> | Cons - it can be more effort to think of a good reference |
| 20:50 | <jgraham> | Sometimes the ref and test can match for the wrong reasons (but it is usually possible to catch this either with a single pixel based test or a well chosen negative ref) |
| 20:51 | <jgraham> | (e.g. both could be blank) |
| 20:51 | <jgraham> | The pros massively outweigh the cons (at least for us) |
| 20:52 | <rniwa> | jgraham: ok. |
| 20:52 | <rniwa> | jgraham: can I quote you on that? |
| 20:52 | <jgraham> | rniwa: Of course |
| 20:52 | <rniwa> | jgraham: (webkit-dev) |
| 20:52 | <rniwa> | jgraham: great. i might also cc you on the thread. |
| 20:52 | <rniwa> | webkit needs someone with reftest knowledge badly :( |
| 20:54 | <jgraham> | rniwa: For example we recently changed some of our font rendering bits and had to re-label 13,500 screenshots |
| 20:54 | <rniwa> | jgraham: LOL |
| 20:54 | <rniwa> | jgraham: yeah... we have a similar issue. |
| 20:54 | <jgraham> | Happily we had already converted 50% of all screenshot tests to reftests |
| 20:55 | <rniwa> | jgraham: that's very nice. |
| 20:55 | <rniwa> | jgraham: we definitely need to do that for some of our regression tests |
| 20:55 | <jgraham> | Yeah, 25,000 tests wouldn't have been better :) |
| 20:58 | <rniwa> | jgraham: right... |
| 20:59 | <rniwa> | jgraham: although i think a good portion of our tests are js tests (i.e. text based) |
| 20:59 | <rniwa> | jgraham: so i think we have something like 10,000-15,000 pixel texts |
| 20:59 | <annevk> | rniwa: "Daniel Glazman from Opera has tried adding reference files for W3C's:" euh, that was Geoffrey Sneddon ;) |
| 20:59 | <rniwa> | annevk: oh oops :( |
| 20:59 | <rniwa> | annevk: will post the correction |
| 21:00 | <annevk> | rniwa: also, I hope you succeed! |
| 21:00 | <rniwa> | annevk: sorry about that :( |
| 21:01 | <jgraham> | rniwa: Yes, most of our tests are js-based these days |
| 21:01 | <rniwa> | annevk: i lost my IRC history prior to writing that email |
| 21:01 | <rniwa> | jgraham: yeah js-based tests are nice :D |
| 21:01 | <rniwa> | unless you introduce weird js regressions :( |
| 21:01 | <jgraham> | Overwhhelmingly so if you go per-result rather than per-file (screenshot/reftests return 1 result per file, js tests typically many) |
| 21:05 | <rniwa> | jgraham, annevk: posted correction :) |
| 21:06 | <rniwa> | jgraham: Thanks for the input. I really appreciate it. |
| 21:06 | <rniwa> | jgraham, annevk, TabAtkins: i think we should really posh the CSS WG to start accepting more reference files. |
| 21:06 | <rniwa> | FWIW, we already have some reference files |
| 21:07 | <rniwa> | and we're talking about creating one for each file we import |
| 21:07 | <rniwa> | so WebKit community can contribute a bit in this area :) |
| 21:07 | <rniwa> | (of course, nothing is set in stone but things are looking good) |
| 21:07 | <jgraham> | rniwa: I think that the plan is for all CSS3 tests to be reftests |
| 21:08 | <jgraham> | Certainly I will be unhappy if Opera start contributing non-reftests |
| 21:08 | <rniwa> | jgraham: but we still need CSS 2.1 tests, right? |
| 21:08 | <rniwa> | jgraham: or is the plan to replace the existing tests by new ones? |
| 21:08 | <jgraham> | rniwa: Well in reality yes, in W3C land, no |
| 21:09 | <TabAtkins> | Ideally we'll replace all the existing W3C CSS2.1 tests with refs. |
| 21:09 | <jgraham> | (that was an answer to your first question) |
| 21:09 | <TabAtkins> | Or at least, all the ones for which that is possible. |
| 21:09 | <rniwa> | TabAtkins: right. that's what we wanna do. |
| 21:09 | <rniwa> | jgraham: could you elaborate a little? |
| 21:09 | <rniwa> | jgraham: are you saying that people at CSS WG doesn't want to make any changes to CSS2.1? |
| 21:09 | <jgraham> | rniwa: We always need new tests that cover previously untested parts of a spec |
| 21:09 | <rniwa> | jgraham: or are they regarding CSS2.1 tests to be obsolete once CSS3 is REC'ed? |
| 21:10 | <rniwa> | jgraham: right. |
| 21:10 | <jgraham> | rniwa: But there is no Process requirement for further CSS 2.1 tests |
| 21:10 | <rniwa> | jgraham: but what about parts that are specified by the older spec? |
| 21:10 | <jgraham> | Since it is in Rec. |
| 21:10 | <jgraham> | (iirc) |
| 21:10 | <rniwa> | jgraham: so you're saying that W3C won't accept any changes? |
| 21:11 | <jgraham> | I think they likely will |
| 21:11 | <rniwa> | jgraham: okay. so we can just do that, no? |
| 21:11 | <jgraham> | But there isn't a great deal of motivation for people to do the work |
| 21:11 | <rniwa> | jgraham: I see. |
| 21:11 | <jgraham> | From the W3C side |
| 21:11 | <jgraham> | But TabAtkins will know more about the CSSWG |
| 21:12 | <rniwa> | jgraham: well, but i think we'll do the most of the work to create reference files though? |
| 21:12 | <rniwa> | jgraham: i guess W3C folks just need to review them |
| 21:12 | <jgraham> | Hopefully we can get ref equivalents to the existing tests, yes |
| 21:12 | <rniwa> | if you're saying that they're not willing to review those reference files, then that'll be a problem... |
| 21:12 | <TabAtkins> | Yes, the primary problem is having someone with the time and motivation to generate all the refs. |
| 21:13 | <rniwa> | but i'm hoping that if enough browser vendors show enough interests, we can convince the WG to spend some time on it |
| 21:13 | <rniwa> | TabAtkins: i think we're going to do it |
| 21:13 | <rniwa> | TabAtkins: at least that's what we've been discussing at webkit-dev |
| 21:13 | <rniwa> | TabAtkins: we're more concerned with how we upstream changes, etc... |
| 21:13 | <TabAtkins> | Cool. |
| 21:14 | <rniwa> | TabAtkins: I think everyone's agreeing that converting them on W3C repository is the best solution if we could do it without much overhead |
| 21:14 | <rniwa> | TabAtkins: althoug htere's some disagreement as to whether that should happen before or after we import tests in webkit |
| 21:28 | <jgraham> | rniwa: (my message to webkit-dev bounced since I'm not subscribed) |
| 21:28 | <rniwa> | jgraham: oh oops. |
| 21:28 | <rniwa> | i think someone needs to approve it |
| 21:28 | <rniwa> | (unfortuantely i don't have a privilege to do it) |
| 21:36 | <rniwa> | jgraham: I've just forwarded your email instead. |
| 21:36 | <rniwa> | jgraham: thanks for the replay btw |
| 21:40 | rniwa | is thrilled about the prospect that we can make a big step forward here |
| 21:44 | <jgraham> | rniwa: Thanks |
