..., which form is the input associated with? [09:01:03.0000] ,

,
-
,
, etc. [09:01:04.0000] Philip`, the second gets ignored because the form pointer is already associated with something [09:04:00.0000] Philip`, are you revising tests for all HTML5 spec changes? [09:05:00.0000] more cases where legacy encoding labels are de facto aliases for newer encodings keep creeping out of the woodwork [09:06:00.0000] i saw that in #webkit and asked him to mail whatwg⊙wo, i'm glad he did :) [09:06:01.0000] annevk: Only for the tokeniser [09:07:00.0000] (and I can't guarantee I haven't missed any changes) [09:08:00.0000] hopefully enough fresh implementations keep coming to sort out all the mistakes... [09:08:01.0000] or test contributions for that matter [09:09:00.0000] aside: great Python on JVM news: http://fwierzbicki.blogspot.com/2008/02/jythons-future-looking-sunny.html [09:09:01.0000] I've been updating the Python html5lib to follow the spec, but gave up on the Ruby one after finding that it already had some non-trivial bugs (plus a trivial bug that hid all the others) [09:10:00.0000] Well, at least it had one non-trivial bug [09:10:01.0000] and maybe it was actually trivial, but I didn't try looking because I worried it might not be [09:11:00.0000] did you leave the bug exposed? [09:11:01.0000] if you did someone else will prolly fix it [09:11:02.0000] (this being the case or something like that) [09:11:03.0000] annevk: Yes, it'll fail if someone runs the tokeniser tests [09:11:04.0000] ah, that was annoying to fix on the python side too... [09:12:00.0000] and the python side was actually hiding some bugs there too, i remember, hmm [09:12:01.0000] since ruby was a port, i guess that's what went wrong... [09:12:02.0000] It seems easy to handle all the entities just with a regexp [09:12:03.0000] (plus another regexp for entities in attributes) [09:13:00.0000] though that's not so good if your input is a stream rather than a string [11:19:00.0000] hsivonen_: The obvious question is, what were fmt=0 up to fmt=5? [11:50:00.0000] Krzysztof Żelechowski's e-mails read like poetry [11:51:00.0000] or haikus [11:51:01.0000] e.g.: [11:51:02.0000] --- [11:51:03.0000] I am not sure I understand you correctly [11:51:04.0000] but if this introduces the ability [11:51:05.0000] to make the user agent [11:51:06.0000] report a different URL than the effective target, [11:51:07.0000] I should try sending an email to public-html or whatwg that's a pantoum sometime, just to see if someone notices [11:51:08.0000] it is going to be a sweet candy for phishers. [11:51:09.0000] (Newer browsers made this effect unavailable to scripts). [11:51:10.0000] --- [11:51:11.0000] actually, a pantoum is too obvious. too much repetition. [11:51:12.0000] Maybe a sonnet? [12:27:00.0000] gsnedders, it sets it from missing to the empty string... [12:27:01.0000] annevk: where? [12:27:02.0000] gsnedders, if you don't see that you're reading too much into it [12:27:03.0000] gsnedders, I quoted those bits in my e-mail... [12:27:04.0000] I searched the entire spec for "missing" [12:27:05.0000] missing is not mentioned again [12:28:00.0000] it does not need to be [12:28:01.0000] annevk: I'm an asshole, per markp's definition. what do you expect? :P [12:29:00.0000] annevk: if it is marked as missing, how does it not need to be marked as not missing? [12:29:01.0000] /me shrugs [12:29:02.0000] From what I can see in the spec, it is marked as missing, and is never marked as _not_ missing [12:30:00.0000] it needs to be set in the "Before DOCTYPE public identifier state" and the "Before DOCTYPE system identifier state" [12:53:00.0000] "public identifier, and system identifier must be marked as missing (which is a distinct state from the empty string)" [12:54:00.0000] then it says "Set the DOCTYPE token's system identifier to the empty string" [13:19:00.0000] Hmm, ICU4J's gb2312 seems to act identically to iso-8859-1 [13:19:01.0000] (gb2312-1980 works like it should, though) [13:21:00.0000] i think i'll add window.scroll/scrollTo/scrollBy to CSSOM View [13:22:00.0000] Oops, I'm wrong [13:24:00.0000] gb2312 does seem to work, but values outside the special range (about 0xA0-0xFF) are treated like in iso-8859-1 [13:25:00.0000] and gb2312-1980 is something totally different and not ASCII compatible [13:26:00.0000] after all, it has scroll* [13:28:00.0000] ok so i parsed over 6 billion files with the parser change to the insertion mode thing (testing that it never hit a table-related element) and it didn't crash [13:29:00.0000] so i figure it's safe [13:31:00.0000] annevk: that to me doesn't mean actually changing the marker [13:33:00.0000] Exception in thread "pool-1-thread-1201" java.lang.NoClassDefFoundError: Could not initialize class nu.validator.htmlparser.impl.EncodingInfo [13:33:01.0000] Hmm... [13:34:00.0000] Philip`: did you update ICU4J without updating the parser, too? [13:34:01.0000] I'm not sure what version of the parser I'm using [13:34:02.0000] Philip`: the new ICU4J has a b0rked UTF-7 decoder that crashes the old EncodingInfo [13:34:03.0000] so that's quite possible [13:35:00.0000] or, rather, EncodingInfo crashes the UTF-7 decoder [13:36:00.0000] gsnedders, there's no marker, there's just states [13:36:01.0000] Things work better when I use a more recently compiled version of the parser - thanks :-) [13:37:00.0000] annevk: "marked as missing" — a marker. [13:38:00.0000] "must be marked as missing (which is a distinct state from the empty string)" -- a state [13:38:01.0000] ... [13:38:02.0000] /me -> back to cssom-view [13:38:03.0000] so the marker is a state. ergh. [13:40:00.0000] I see reported encoding errors on 1.5% of pages with reported encodings [13:41:00.0000] (counting things like charset=ISO-8559-1 as an error) [13:41:01.0000] (but unknown charsets are only about a quarter of the errors) [13:43:00.0000] (18% of gb2312 pages have errors) [13:47:00.0000] http://www.narkasabasi.com/v2/ [13:47:01.0000] (Urgh - s/ /\n/) [13:47:02.0000] Looks like they couldn't quite make their mind up [13:49:00.0000] - that's not good if it gets incorrcetly interpreted as the page's charset [13:50:00.0000] /me wonders why dave hodder is asking the same question on public-html-comments that I already answered on the whatwg list [13:50:01.0000] http://jellybelly.com/International/Japanese/home.html is an interesting test case [13:50:02.0000] (Opera 9.2 fails) [13:51:00.0000] (Opera 9.5 fails too, though it gets the layout correct) [13:51:01.0000] Philip`: check out the JS in there, too [13:51:02.0000] jgraham, I think he's asking for a pointer to that e-mail again, he apparently forgot :) [13:51:03.0000] (Safari 3 fails too) [13:52:00.0000] /me goes to answer again [13:52:01.0000] SadEagle: What's unusual about the JS? [13:53:00.0000] nothing unusual --- but it's sniffing for netscape >= 3, ie >= 4 [13:54:00.0000] Ah [13:55:00.0000] That should degrade gracefully in other browsers, so it's not much of a problem :-) [13:58:00.0000] oh, Acid3 is announced [14:00:00.0000] /me sees 88582 pages with , 25175 with an HTTP Content-Type: text/html; charset=..., and 171 with [14:00:01.0000] (out of 130K) [14:01:00.0000] how many pages with content="" and without a valid value for http-equiv="" ? [14:01:01.0000] What's a "valid value"? [14:02:00.0000] content-type ascii case-insensitive [14:02:01.0000] What about all the other http-equiv values? [14:02:02.0000] where content contains the word charset [14:03:00.0000] content=""* [14:03:01.0000] Hmm, I suppose I could look for that [14:04:00.0000] but I won't bother doing that now, since it takes 20 minutes to run [14:04:01.0000] k [14:21:00.0000] Philip`: LOL, just stumbled on a webpage with 2 encoding headers, neither of which is right [15:15:00.0000] Hixie: In the outline algorithm, in the conditon "When exiting a sectioning content element, if the stack is not empty" [15:15:01.0000] yes? [15:15:02.0000] it's not clear to me what "Let current section be the last section in the outline of the current outlinee element. [15:15:03.0000] Insert its outline at the end of the current section. (This does not change which section is the last section in the outline.)" means [15:16:00.0000] Specifically the last line. [15:16:01.0000] oops [15:16:02.0000] "its" refers to the sectioning content element being exited [15:16:03.0000] let me fix that [15:21:00.0000] ok fixed [15:21:01.0000] is that clearer? [15:25:00.0000] I think that helps. I just need to work through and see if I have a sensible mental model of what's happening [15:25:01.0000] k [15:25:02.0000] i know you will, but, let me know what i can do to improve it [15:26:00.0000] the current algorithm should be way better than what was there before, yet get mostly the same results [15:28:00.0000] Hixie, you're still going through parser feedback right? [15:28:01.0000] yes [15:28:02.0000] i'm stalled right now trying to figure out how to handle spaces in x
[15:29:00.0000] flip a coin :) [15:30:00.0000] between what and what? i have no options so far :-) [15:30:01.0000] between " x
" and "x

" [15:30:02.0000] annevk: Do you mean the algorithm for handling spaces should involve flipping a coin? [15:31:00.0000] Philip`, that could be interesting, but requiring specific hardware for HTML 5 might be too much [15:32:00.0000] /me suggests asking the user to decide [15:33:00.0000] annevk: Any implementation is acceptable as long as it acts the same as flipping a coin [15:33:01.0000] annevk: oh i've decided it's "x

", the question is how to get there. [15:33:02.0000] "You have encountered a space inside a table. Would you like to move it outside (Y/n)" [15:34:00.0000] There should be a web service that provides a stream of random bits, called Flipr [15:34:01.0000] i'm thinking a flag on the table that decides whether spaces are sent out or not [15:34:02.0000] that gets set as soon as you send anything out [15:34:03.0000] the problem is nested tables in the innerHTML case makes this relatively hard to specify [15:34:04.0000] so i think the current spec covers it [15:34:05.0000] because you consume characters until the end [15:34:06.0000] which makes "x " a character block [15:35:00.0000] and the space before it gets treated specially [15:35:01.0000] the current spec has no concept of "character blocks" :-) [15:35:02.0000] the append the character stuff [15:35:03.0000] but e.g. " x

" should become "x

" [15:36:00.0000] so it's not that simply [15:36:01.0000] simple [15:36:02.0000] for real? [15:36:03.0000] /me didn't think that trailling space would be placed in front too [15:37:00.0000] " foo bar
" shouldn't display "foobar" [15:37:01.0000] it should display "foo bar" [15:37:02.0000] that's the bug :-) [15:38:00.0000] a flag sounds easiest then, yes [15:49:00.0000] Hixie: I think I'm still confused about what it means to insert a section "at the end of" another section. Do you mean "as the last child section of" the other section? [15:50:00.0000] (so if you have
foo you end up with a section for the
as a child of the section for the [15:50:01.0000] in "When entering a heading content element" i used the term "append it to /candidate section/" [15:50:02.0000] is that clearer? [15:50:03.0000] sounds like appendChild()... [15:52:00.0000] Yeah, that bit is clearer. [15:52:01.0000] ok i'll use that terminology [15:52:02.0000] /me regens 2008-03-04 [17:14:00.0000] if anyone gets the reason behind the reference in the title of my latest blog, i'll be impressed [17:18:00.0000] hsivonen_: does validator.nu have a way to make it validate the page from the Referer field? [17:19:00.0000] hsivonen_: btw #presets in the "the pitch" section of your about page is a broken link [17:45:00.0000] Hixie Does it have anything to do with http://www.whatwg.org/issues/data.html ? :) [17:48:00.0000] no [17:59:00.0000] BTW, that page is pretty slow in Firefox. Pretty smooth in Opera. [17:59:01.0000] The mouse movement on canvas that is. [17:59:02.0000] Has everyone here read this already? "IE8 standards mode" will be the default "standards mode" http://blogs.msdn.com/ie/archive/2008/03/03/microsoft-s-interoperability-principles-and-ie8.aspx [18:08:00.0000] yeah, i even blogged about it [18:08:01.0000] great news [18:11:00.0000] sorry, I'm too slow [18:11:01.0000] :-) [18:11:02.0000] i didn't mean it that way :-) [18:25:00.0000] /me is disconcerted by a picture of Osama bin Laden next to the post about Acid 3 [18:25:01.0000] uri? [18:26:00.0000] http://www.flickr.com/photos/41089232@N00/114103905 [18:29:00.0000] either flickr is down or my proxy is acting up [18:30:00.0000] wtf the web just died for me [18:32:00.0000] Didn't you hear about the scheduled maintenance tonight? The web has to be turned off for two hours while it's all upgraded [18:32:01.0000] wouldn't that be nice [18:33:00.0000] mpt: i don't see anything about acid3 on there? *confused* [18:34:00.0000] Hixie, that was the photo "randomly selected from Flickr using Yahoo! Pipes based on the words in the latest Web log entry on this site" [18:35:00.0000] oh haha i didn't realise that was the image that had been selected [18:35:01.0000] wow that's funny [18:38:00.0000] so... [18:38:01.0000] if we make stay in the table... [18:38:02.0000] what the heck should x
do [18:38:03.0000] /me studies [18:39:00.0000] "This decision comes on the heels of a recent leak of a classified NSA National Security Assessment which ranked the Bush Administration itself as the #2 threat to U.S. national security interests." [18:39:01.0000] i hadn't heard about that [18:40:00.0000] Hixie: does html 5 parsing require that where tags go in the DOM depend on attribute values? [18:40:01.0000] it's about to start going down that slippery slope, yes [18:40:02.0000] bz and othermaciej have asked that be given special status in [18:41:00.0000] hmm [18:41:01.0000] jruderman: [citation needed] [18:41:02.0000] https://bugzilla.mozilla.org/show_bug.cgi?id=327796#c5 makes me think that will be unpopular [18:41:03.0000] Philip`: hehe [18:42:00.0000] I don't exactly love the idea of that rule but it seems to matter for compatibility [18:42:01.0000] jruderman: https://bugzilla.mozilla.org/show_bug.cgi?id=390565 [18:53:00.0000] Hixie, where do bug reports in acid3 go? [18:53:01.0000] ian⊙hc or the wasp alias, if you can find it [18:53:02.0000] i think that's acid3⊙wo [18:53:03.0000] not sure [18:53:04.0000] but ian⊙hc is fine [18:54:00.0000] and i try to give quick turnaround, so if you don't get a reply within 12 hours, ping me :-) [18:54:01.0000] dbaron: what bug did you find? [18:55:00.0000] othermaciej, I think test 42 has an incorrect selector [18:55:01.0000] Fx trunk should pass that test [18:55:02.0000] and it looks wrong by inspection [18:55:03.0000] what failure number do you get? [18:55:04.0000] although I didn't test the modificication [18:55:05.0000] Hixie, the very last failure [18:56:00.0000] 39? [18:56:01.0000] /me looks [18:56:02.0000] no, "rule did not start matching after change" [18:56:03.0000] oh the one between 12 and 14? ok [18:56:04.0000] /me looks at that instead :-) [18:57:00.0000] oh, I see, the test continues after that function [18:57:01.0000] I think "#div1 ~ div div + div > div" should be "#div1 ~ div + div div > div" [18:59:00.0000] after 310 is in the tree, you get this DOM: [18:59:01.0000] [18:59:02.0000] oh [18:59:03.0000] hrm [19:00:00.0000] ignore me [19:00:01.0000] for now, anyway [19:00:02.0000] and "#div1 ~ div div + div > div" should match 3111 the same way that "div1 ~ div3 div310 + div311 > div3111" does [19:00:03.0000] seems valid to me :-) [19:01:00.0000] yeah [19:01:01.0000] for some reason I forgot to process the insertBefore mentally even though I saw it [19:01:02.0000] no idea why we fail, though [19:01:03.0000] I guess I'll have to ask someone to write a testcase [19:02:00.0000] it's basically a dynamic variant of your greedy matching test from css1, extended to use the css3 combinators [19:02:01.0000] though maybe it's not too hard to extract just the relevant bits [19:04:00.0000] http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C!DOCTYPE%20html%3E%0A%3Cstyle%3E%20div1%20~%20div3%20div310%20%2B%20div311%20%3E%20div3111%20%7B%20border%3A%20solid%3B%20%7D%20%3C%2Fstyle%3E%0A%3Cbody%3E%3Cdiv1%3E%3C%2Fdiv1%3E%3Cdiv2%3E%3C%2Fdiv2%3E%3Cdiv3%3E%3Cdiv31%3E%3Cdiv311%3E%3Cdiv3111%3E%3C%2Fdiv3111%3E%3C%2Fdiv311%3E%3C%2Fdiv31%3E%3C%2Fdiv3%3E%3C%2Fbody%3E%0A%3Cscript%3E%0Avar%20div31%20%3D%20document.getElementsByTagName('div31')%5B0%5D%3B%0Adiv [19:05:00.0000] if that got truncated, go to http://software.hixie.ch/utilities/js/live-dom-viewer/ and hit "download" then "permalink" to get the uri [19:05:01.0000] i think that's the same test [19:05:02.0000] and it works fine if you do the mutation manually [19:06:00.0000] so it must be a dynamic thing [19:06:01.0000] yeah, I see the bug [19:06:02.0000] it's a one-liner [19:06:03.0000] cool [19:06:04.0000] now I need to find an excuse to check it in [19:06:05.0000] If only this test had been easier to parse the results of I'd have used it before I checked in my fix for those bugs, and been able to get it in easily... [19:07:00.0000] /me hates overly complex tests [19:07:01.0000] i tried to make the tests as self-contained as possible [19:07:02.0000] but really acid tests aren't meant to be processed directly by engineers [19:07:03.0000] when acid2 came out i made minimised tests for each bug that opera had in acid2 [19:08:00.0000] i'd expect any qa team to do the same for any acid test [19:08:01.0000] er, no, I don't see the bug [19:10:00.0000] I better leave splitting these bugs up to people who actually think they have enough time to do it, instead of rushing... [19:11:00.0000] i wouldn't recommend trying to do it this late in a release cycle anyway [19:11:01.0000] best to work on standards bugs at the start of a cycle [19:13:00.0000] yeah, except you always release these acid tests a few months before the end of our release cycles [19:14:00.0000] and then we ship a not-much-improved release and look stupid [19:15:00.0000] what would you like me to do? wait til all the browser vendors' release schedules line up? [19:15:01.0000] acid3 has been in production since last april, and mostly came to a head recently because of the IE acid2 announcement [19:15:02.0000] i recommend shorter release cycles, that way you never have a problem like this :-) [19:17:00.0000] in fact, i'd argue that this is the best time for you for an acid test to come out, as it means that as the test finally gets all the bugs shaken out of it, you'll be at the best possible time to start working on standards bugs again [19:17:01.0000] (i.e. it'll be stable just as your release cycle starts) [19:17:02.0000] yeah, I've been pushing for shorter release cycles [19:18:00.0000] We may well do one. [19:18:01.0000] dare I ask where the function expect is defined? [19:19:00.0000] for which test? [19:19:01.0000] er, never mind [19:20:00.0000] The closures confused me about what was calling what [19:20:01.0000] I'm just annoyed about this particular test because I thought I fixed all the bugs that would have caused it to fail. [19:20:02.0000] heh [19:29:00.0000] ah, it's a matching bug, not a dynamic change handling bug [19:35:00.0000] the next Safari release won't look so good on Acid3 [19:35:01.0000] not as good as current nightlies [19:39:00.0000] dbaron: really? it only seems to occur with a dynamic change. [19:40:00.0000] Hixie, not sure why you think that -- I tried removing and reinserting the body and the bug was still present [19:40:01.0000] Hixie, simple testcase in https://bugzilla.mozilla.org/show_bug.cgi?id=420814 [19:41:00.0000] odd, i couldn't get the test to fail without dynamic changes [19:41:01.0000] oh well [19:41:02.0000] oh i guess i didn't have elements named the same thing in my version of the test [19:41:03.0000] so maybe there are two bugs here [23:02:00.0000] if this upcoming change doesn't break something, i'll be impressed [23:23:00.0000] so anyone got any hot ideas on how to handle
in a compatible way? [23:23:01.0000] Philip`? [00:21:00.0000] annevk: regarding "UA-specific crap" on the tracker: I'm using the annotations. [00:26:00.0000] I think the "ordered" is a red herring when it comes to
and
. It really is about showing the counters. A bulleted list can still have a deliberate order. [00:43:00.0000] hsivonen_: don't let tantek hear you say that [00:45:00.0000] heh [00:51:00.0000] I haven't read it thoroughly yet, but did Microsoft just turn completely around on the x-ua-compatible stuff? [00:54:00.0000] virtuelv: Not completely - just on the default [00:54:01.0000] :/ [00:55:00.0000] (which is the part people were most complaining about) [00:57:00.0000] /me sees the Osama image at http://ln.hixie.ch/?start=1066145333&count=1 [01:00:00.0000] hsivonen_: :D :D :D [01:01:00.0000] the image is based on Hixie's last entry, so you'll see it on all pages [01:12:00.0000] wow!!!! "We’ve decided that IE8 will, by default, interpret web content in the most standards compliant way it can." [01:12:01.0000] http://blogs.msdn.com/ie/archive/2008/03/03/microsoft-s-interoperability-principles-and-ie8.aspx [01:14:00.0000] Lachy: yes, very good news [01:17:00.0000] Hixie: dammit, you're working too fast. i can't keep up :P [01:22:00.0000] zcorpan: :-D [01:23:00.0000] i guess i'll have to have a "last insertion mode" variable to do the
x y
thing [01:23:01.0000] anyway bed time [01:25:00.0000] aha! the precedent is set for looking at attribute values in the tree builder [01:27:00.0000] scary [01:27:01.0000] ah. it's type=hidden [01:28:00.0000] that's good [01:28:01.0000] why is it good? [01:31:00.0000] zcorpan: compat. and in the case of input, the conceptual reality is that the type attribute creates practically separate element types [01:31:01.0000] hsivonen_: ok. and yep [01:53:00.0000] Hixie: Maybe handle it by copying what WebKit or Gecko does? [01:53:01.0000] I don't really know what they do, though [01:54:00.0000] (except I vaguely remember WebKit running the normal parsing sequence but with an 'is being inserted into table' flag that changes the processing of table elements) [01:59:00.0000] '
Ignore the token.
' :) [02:49:00.0000] Philip`: it's not that simple; has to do different things than
[02:50:00.0000] are we now figuring out the parser architecture is not good enough? [02:50:01.0000] /me hopes not [02:50:02.0000] no, it can be hacked to do this in various ways [02:50:03.0000] i'm just trying to work out which way [02:50:04.0000] i can remember the last mode and treat
a sectioning root element. My guess is that that will make the algorithm useless for a significant amount of real-world content [15:19:01.0000] the algorithm is already useless for that if it only takes into account
-
[15:20:00.0000] for real world content you want to base this stuff on computed style information [15:20:01.0000] annevk: A non-trivial number of pages use
-
in a semi-sane way [15:20:02.0000] so given that you might as well discourage table for layout madness [15:21:00.0000] jgraham, inside ? [15:21:01.0000] jgraham: i would agree, if it wasn't for the problem of legitimate uses of headers inside cells being cases where you _would_ want it treated a sectioning root [15:21:02.0000] Hixie: Example? [15:22:00.0000] /me can't really think of why you would want headers in tables at all [15:22:01.0000] a character sheet where one of the cells is the characters's background story [15:23:00.0000] a page using tables for layout [15:23:01.0000] And the story has multiple subheadings? [15:23:02.0000] yeah [15:23:03.0000] jwalden: we already know that's non-conforming :-) [15:24:00.0000] just stating the practical response :-) [15:24:01.0000] jwalden: That's the case that is currently unsupported by the algorithm [15:24:02.0000] er, pragmatic [15:24:03.0000] and that I think needs to be supported [15:25:00.0000] if i made it a sectioning element it still wouldn't work [15:25:01.0000] for real content [15:26:00.0000] I think for real content it would have to be ignored. [15:27:00.0000] right, which isn't compatible with conforming content [15:27:01.0000] I'm not sure how to deal with the character sheet example other than to say "don't use a table for that", which seems a stretch as a table isn't obviously wrong [15:27:02.0000] and as much as we want to make legacy pages work, valid semantic pages have to be a higher priority imho [15:29:00.0000] I guess it would be nice to look at how headings as descendants of table content are actually used in the real world [15:31:00.0000] On an (not) entirely different topic, is there a good reason to call the current outline element /current outlinee/ rather than, say, /current outline element/ [15:33:00.0000] which looks less like a typo and prevents me from having to make the same translation every time in my head [15:33:01.0000] heh [15:33:02.0000] well it's not the "current outline element" [15:34:00.0000] it's the "element that is currently having an outline created for it" [15:34:01.0000] i don't mind changing it to something else if you have a suggestion that's correct :-) [15:34:02.0000] so supporting
turned out way simpler than i expected [15:35:00.0000] i must have misunderstood something [15:39:00.0000] Oh I hadn't read outlinee like that at all. I guess that's wwhat happens when you use made up words even if they do follow standard morphological rules :) [15:40:00.0000] :-) [15:42:00.0000] closes when found in "in select" [15:47:00.0000] (tested Firefox and Opera) [15:47:01.0000] nn [15:47:02.0000] /me -> bed [15:49:00.0000] doesn't the spec already say that? [15:51:00.0000] heh, i think my script handles "
a
should be processed according to the rules for the in table mode, which would be the anything else section [09:41:00.0000] jgraham__: If a transition only happens after a parse error, then it's red; otherwise it's blue for reprocess-as-if, and black for set-insertion-mode [09:41:01.0000] it's currently putting it as a child of the tr, but, AIUI, it should be as a child of the div [09:41:02.0000] (in that above demo I linked to) [09:41:03.0000] Lachy: That's the same conclusion I had reached [09:41:04.0000] ok [09:41:05.0000] Philip`: Thanks [09:41:06.0000] jgraham__, see the PM i sent you [09:42:00.0000] Lachy: Dunno if I can reply to PMs on freenode [09:42:01.0000] oh, isn't your nick registered? [09:45:00.0000] No. I think technically someone else owns it which means I'm being really bad by using it but otoh, no one has ever complained [09:46:00.0000] /me avoids the name reuse problem by appending random significant punctuation [09:47:00.0000] /me doesn't have that problem :) [10:40:00.0000] Lachy: The whitespace thing should be fixed now [10:41:00.0000] /me has discovered a bunch of lxml-related breakage that needs to be fixed [11:49:00.0000] whoa, @microsoft.com on @whatwg.org [11:49:01.0000] oh wait, that was not a first [11:50:00.0000] anyone been doing any testing with betas regarding how figure/legend (or unknownelement/legend) works? [11:51:00.0000] last i heard it didn't work wel [11:51:01.0000] l [11:51:02.0000] gecko now is generating a fieldset :( [11:53:00.0000] whole doc after gets included in the fieldset....ouch! [11:56:00.0000] should there end up being a legend element in the DOM, or should it just generate a text child of figure? (sounded like the latter last i read the spec closely) [11:57:00.0000] the former [11:58:00.0000] that makes more sense to me [11:59:00.0000] opera seems to do the latter? [12:00:00.0000] looked that way anyway....though i couldn't get the dev toolbar working for some reason...to view the dom [12:00:01.0000] could be, we're not doing what HTML5 wants at this point [12:01:00.0000] anyway, got to go [12:01:01.0000] take care [12:39:00.0000] there is some irony to the way that microsoft suggests xxx has a security problem when in fact xdr is the one increasing the attack surface [12:46:00.0000] Philip`: there's a line missing from InFrameset to AfterFrameset [12:46:01.0000] unless the spec itself is wrong somehow [12:46:02.0000] let me know if i'm being inconsistent or something [12:47:00.0000] /me notes that #writing is not in sync with #parsing [12:47:01.0000] Hixie: Oops, that's because my spec-to-code regexp thing is incomplete and fails to understand that bit of text [12:48:00.0000] It's missing after-after-(body|frameset) -> in-\1 for the same reason [12:48:01.0000] /me is sad that kinda-sorta-half-updated postMessage [12:48:02.0000] ah [12:48:03.0000] it's on window [12:48:04.0000] zcorpan_: oh? [12:49:00.0000] but it uses uri/domain [12:49:01.0000] Hixie: It would be much easier for me if you wrote the spec in OCaml instead of English [12:49:02.0000] Philip`: that's your objective opinion? [12:49:03.0000] ba-dum-ch [12:50:00.0000] Philip`: i believe you [12:50:01.0000] namely entities in escaping text spans (in rcdata) [12:51:00.0000] Philip`: i'm happy to make my text more self-consistent (i did a lot of changes to that effect recently so you should be able to simplify your regexps) but it's staying in english :-) [12:53:00.0000] i also found that #writing didn't say anything about the LF and , but then i remembered that is not valid in the first place :) [12:54:00.0000] but #serialising should talk about that [12:56:00.0000] speaking of serializing, i think the algorithm should change for (r)cdata elements to only emit the children text and cdata nodes [12:56:01.0000] s/ren// [12:57:00.0000] (including plaintext) [12:58:00.0000] less checking (hence better perf) and better roundtripping (comments are not turned into text but are instead just dropped) [12:58:01.0000] and it's what firefox does, i think [13:00:00.0000] zcorpan_: please send e-mail, i'm working on tables at the moment [13:19:00.0000] (othermaciej: is the windows safari 3.1 still labelled a beta?) [13:19:01.0000] Hixie: no [13:24:00.0000] cool [13:25:00.0000] zcorpan_: the perf aspect is purely an implementation detail [13:26:00.0000] zcorpan_: how should the XML "<div> text</div>" be serialised by this algorithm? [13:35:00.0000] [13:37:00.0000] for anyone interested in tracking: gecko figure/legend bug is 423721 [13:38:00.0000] (assming the elements are in the HTML namespace :) ) [13:40:00.0000] a-ja: doesn't only imply fieldset if it's foudn directly in body? [13:41:00.0000] a-ja: and is otherwise just dropped? [13:41:01.0000] /me wants [13:43:00.0000] zcorpan: well....i think it used to be that way....in body or unknown element...but since the post-b4 parser update to allow unknown (e.g. html5) elements to contains blocks, this has cropped up [13:44:00.0000] ah. haven't tested a build with the other fix in [13:44:01.0000] grab latest nitely if you wanna test....there was a related checkin yesterday [13:45:00.0000] Maybe someone should run the HTML5 parser tests in Firefox to try to find unintended consequences of that patch [13:46:00.0000] little late in beta cycle to replace ff's html parser with html5lib, i'd think :) [13:47:00.0000] sounds like a good idea though...since it hasn't made it to a beta release yet. though code freeze for "last" beta is tonite [13:48:00.0000] I'm not suggesting trying to make FF pass the tests - they're just a useful source of occasionally obscure HTML code that could be compared before/after the patch [13:51:00.0000] Hixie: ping [13:51:01.0000] yo [13:51:02.0000] hey [13:51:03.0000] pretty much, unknown elements can now have block children....instead of having to wrap them in an inline (e.g. span) to get em in DOM [13:51:04.0000] Hixie: so I'm looking at the application cache spec [13:53:00.0000] i'm assuming you have a question related to this section? :-) [13:54:00.0000] Hixie: actually, I have two! :) [13:54:01.0000] Hixie: first, I didn't see an applicationCache property on the window interface [13:54:02.0000] yeah, known bug [13:54:03.0000] OK [13:55:00.0000] Hixie: and a small nit in 4.6.4, could p3 and p4 switch places? [13:56:00.0000] p3 and p4? [13:56:01.0000] Hixie: step 3 and step 4 [13:57:00.0000] what difference does it make? [13:58:00.0000] step 3 doesn't have any visible side-effects [13:58:01.0000] Hixie: none really, it's just that you don't need to define what the cache is until after step 4 [13:58:02.0000] right [13:58:03.0000] i prefer to have all the definitions together [13:58:04.0000] makes it easier to find them [13:58:05.0000] blame my pascal upbringing :-) [13:59:00.0000] OK, told you it was a small nit :) [13:59:01.0000] :-) [14:01:00.0000] http://software.hixie.ch/utilities/js/live-dom-viewer/?%3Ca%20href%3Da%3Eaa%3Cmarquee%3Eaa%3Ca%20href%3Db%3Ebb%3C%2Fmarquee%3Eaa is different in FF3 vs FF2 [14:17:00.0000] looks like every single news article about Safari 3.1 mentions HTML 5 [14:18:00.0000] what HTML 5 features does 3.1 have (over 3.0)? [14:18:01.0000] the biggest are the video and audio elements and (nearly all of) their APIs [14:19:00.0000] (also some smaller things like getElementsByClassName) [14:21:00.0000] (of course, getElementsByClassName will probably get a lot more use in the near term) [14:29:00.0000] hmm... SSML? [14:29:01.0000] what's SSML? [14:30:00.0000] http://www.w3.org/TR/speech-synthesis/ [14:30:01.0000] looks like a weird version of HTML to me [14:33:00.0000] 3.1 also implements the Database spec [14:40:00.0000] oh yeah [14:44:00.0000] hehe, http://diveintomark.org/archives/2008/03/18/translation-from-ms-speak-to-english-of-selected-portions-of-joel-spolskys-martin-headsets [14:47:00.0000] /me wonders if he's intentionally misinterpreting "nobody has a way to test against the standard" as meaning testing browsers against standards, rather than testing web pages against standards [14:50:00.0000] hsivonen: when validating xhtml with resolving external entities, v.nu complains about shape attributes on , even though the markup didn't have shape=''. i understand that the dtd resolves them, but it's probably confusing and unhelpful for authors... [14:56:00.0000] Hixie: what happens if I setAttribute('manifest', ...) on the HTML element? [14:56:01.0000] nothing [14:56:02.0000] (as per the spec) [14:56:03.0000] the manifest attribute is only handled during parse time [14:59:00.0000] great [14:59:01.0000] :) [14:59:02.0000] Hixie: oh, you are right - now why didn't I notice that [15:00:00.0000] search for "run the application cache selection algorithm" [15:00:01.0000] those are the only times that the attribute has any effect [15:00:02.0000] the spec defines the handling for XML files, text/plain files, and HTML files (in the HTML parser) [15:01:00.0000] oh and for images and plugins [15:01:01.0000] and error pages [15:01:02.0000] man, i was thorough [15:02:00.0000] :) [15:02:01.0000] that's good [15:03:00.0000] Philip`: I'm not quite clear that Joel does mean "test web pages against standards", but he seems somewhat incoherent on that point [15:04:00.0000] i think joel doesn't understand that you can have pragmatic standards [15:05:00.0000] his argument seemed predicted on the concept of specs either being idealistic and unimplementable, draconian and unsuitable, or overly vague. [15:05:01.0000] he missed the "realistic and detailed" option [15:05:02.0000] I'm not sure what his point was in his big rant [15:05:03.0000] but I admit I didn't read the whole thing [15:05:04.0000] yeah me either [15:05:05.0000] i started to read it but realized it kept going [15:05:06.0000] to not understanding the point [15:05:07.0000] (i did read it) [15:06:00.0000] I think the point might have been to say he predicted in 2004 that Vista would suck [15:07:00.0000] his reasoning for why vista hasn't been the success that XP was seems somewhat flawed, imho [15:10:00.0000] jgraham__: do you have any discussion of the cell association algorithm anywhere? i'm curious to know which cases you were trying to solve [15:10:01.0000] vista is in many ways nicer than XP [15:10:02.0000] it just has some hickups [15:10:03.0000] Hixie: I don't know if we have anything written down that's not in email [15:11:00.0000] i didn't see much discussion in the e-mails, mostly just algorithms :-) [15:12:00.0000] Basically it's optimized for tables like http://annevankesteren.nl/2007/09/tmb-overview and http://sitesurgeon.co.uk/tables/tides/minimal.html [15:14:00.0000] (which seems to be a fairly common general pattern) [15:16:00.0000] so a column (row) header's scope stops when it hits another header with the same width (height) and starting x (y) position? [15:17:00.0000] Just the same with/height iirc although I guess maybe it should also consider the starting position [15:18:00.0000] i guess i'm going to have to just look at a lot of tables and determine the best algorithm that covers those tables [15:18:01.0000] yay for ben's work looking at tables [15:19:00.0000] k. bbiab. [15:19:01.0000] On Joel's article: http://my.opera.com/hallvors/blog/show.dml/1818385 [15:19:02.0000] Hixie: Yeah. [15:21:00.0000] /me feels sorry for Hixie having to wade through all the crap in the table-related discussion [16:03:00.0000] Yay for search engine crawlers that DoS my server [16:06:00.0000] (I suppose I shouldn't provide an infinite number of URLs in a system that takes a large fraction of a second to generate each page with no caching or anything clever, but still they should be more considerate of bad server design...) [16:25:00.0000] /me wonders when Searchme will realise it's just getting Forbidden responses and stop requesting two pages a second [16:37:00.0000] Philip`: thank you for chardet lengths. I think I'll probably run chardet on the first 512 bytes and running it on more than 1024 bytes doesn't appear to make much sense [16:40:00.0000] (Hmm, looks like Searchme uses pretty much the whole of 208.111.154.0/24, so I'll have to block all that...) 2008-03-19 [17:03:00.0000] Philip`: At some point it would be nice to have the data behind the phase transition diagram for the tree construction (the .dot file or whatever) [17:03:01.0000] But right now I'm going to sleep [17:04:00.0000] jgraham__: http://canvex.lazyilluminati.com/svn/tokeniser/ has a possibly out of date version of the code for generating the graph (and more) [17:16:00.0000] (graph_gen.ml in particular produces the .dot file, based on treeconstructot_spec.ml which is generated by treeconstructor.pl from section-tree-construction.html which is an XMLified version of the section from the HTML5 spec) [17:16:01.0000] s/t/r/ [03:16:00.0000] /me will use this channel to note URLs to tables [03:16:01.0000] i hope no-one minds [03:16:02.0000] http://broads-authority.gov.uk/boating/navigating/tide-tables.html [03:17:00.0000] http://www.usability.com.au/resources/tables.cfm [03:25:00.0000] http://www.gutenberg.org/files/17128/17128-h/17128-h.htm [04:07:00.0000] http://www.gutenberg.org/files/21341/21341-h/21341-h.htm#metrical_feet [04:11:00.0000] http://www.gutenberg.org/files/19598/19598-h/19598-h.htm [04:11:01.0000] http://broads-authority.gov.uk/boating/navigating/tide-tables.html [04:11:02.0000] http://www.socialsecurity.gov/policy/docs/statcomps/eedata_sc/2004/table01.html [04:12:00.0000] http://joeclark.org/dossiers/PDFUA/PDFUA-tables-1.html [04:43:00.0000] I'm trying to find a way to reprhase this requirement in selectors api: "If the user agent also supports some level of CSS, the implementation should support the same set of selectors in both these APIs and CSS." [04:44:00.0000] I want it to say that UAs should support the same selectors, but allow for exceptions in case some selectors are introduced that don't work for CSS, but do for selectors api, or vice versa [04:45:00.0000] any suggestions? [04:59:00.0000] to me it sounds like something that should be non-normative... [05:04:00.0000] yeah, that's what I was thinking [05:29:00.0000] I think this is better "For user agents that support of CSS, it is recommended that they support the same set of selectors for use in both these APIs and CSS." [05:32:00.0000] so IE8 uses a bugus mime type application/x-hatom in WebSlices discovery... [05:48:00.0000] Lachy: isn't "RECOMMENDED" an rfc2119 keyword? [05:49:00.0000] yes, it is [05:50:00.0000] but it allows for the requirement to be ignored [05:50:01.0000] if there are valid reasons to do so [05:50:02.0000] same as "SHOULD" [05:50:03.0000] yes [05:50:04.0000] so your change is merely editorial :) [05:51:00.0000] well, I couldn't think of a non-normative way to phrase it that didn't use either "should" or "recommended" [05:51:01.0000] but the existing draft in SVN was actually a MUST level requriement [05:51:02.0000] so it's more than editorial [05:51:03.0000] ah [05:52:00.0000] "User agents are strongly encouraged to..." [05:52:01.0000] if you want it non-normativ [05:52:02.0000] e [05:52:03.0000] (or s/strongly //) [05:52:04.0000] that might work, though what are the advantages or disadvantages of making it non-normative? [05:53:00.0000] is it something you'd put in a testsuite? [05:53:01.0000] (though "should"s are often not in testsuites anyway...) [05:53:02.0000] it would be difficult to test with a test suite [05:54:00.0000] though, not impossible [05:54:01.0000] indeed [05:54:02.0000] might be a useful thing to test [05:54:03.0000] dunno [05:54:04.0000] don't really have an opinion on the matter [05:54:05.0000] just trying to help with wording :) [08:13:00.0000] hmm. Pingback interop doesn't look so great [08:13:01.0000] pingback to WordPress requires It looks like it probably requires lowercase a, but it doesn't care about the href [08:32:00.0000] i.e. and would work just as well [08:39:00.0000] perhaps we should make parse into [08:40:00.0000] Why? [08:40:01.0000] convenient? [08:41:00.0000] Doesn't seem that hard to write href="" :-p [08:41:01.0000] particularly compared to the cost of updating tutorials and tools [08:42:00.0000] I see one person writing but that's about it [08:43:00.0000] Also would be kind of ambiguous [08:43:01.0000] this charset alias stuff is crazy but fascinating in a way [08:44:00.0000] Sounds like the web [08:45:00.0000] here's an idea for a testing project: extracting all the charset aliases supported by Gecko, generating test pages for each and checking which aliases are actually supported by IE, Safari and Opera [08:46:00.0000] the aliases for the old Apple encodings seem to be particularly inconsistent in different implementations [08:47:00.0000] according to http://philip.html5.org/data/charsets.html almost no one uses them anyway... [08:48:00.0000] There's a million times more pages on the web than what I looked at, so "almost no one" multiplies into quite a lot of people [08:48:01.0000] another research item: figuring out the requirements for the visual and logical iso-8859 Hebrew and Arabic variants [08:49:00.0000] Philip`: well, according to your numbers, <$mtpublishcharset$> is more popular than x-mac-thai :-) [08:50:00.0000] According to my numbers, the difference is not significant given the sample size, so you can't tell which is more popular :-) [08:54:00.0000] implementations don't agree whether the chaset name for MacUkranian should contain the word ukraine or ukranian, whether there should be the x- prefix and whether there should be a hyphen after mac [08:55:00.0000] (Also my sample is quite strongly biased towards certain languages that are not Thai) [08:55:01.0000] (Er, my sample isn't, but my population is) [08:56:00.0000] (compared to the entire web) [08:56:01.0000] (I'm probably messing up the statistical terms anyway) [08:57:00.0000] my local build of Validator.nu parser now knows about 1194 charset names (when certain ones have been banned) [08:57:01.0000] that's just crazy when utf-8 should be the one name to rule them all [08:59:00.0000] Until n years in the future when a new character encoding that solves all the problems with UTF-8 will become popular and should rule over all others? [09:00:00.0000] problems, what problems? :-) [09:00:01.0000] (utf-8-nfc :-) [09:01:00.0000] There needs to be a URI-based namespacing system for private extensions, otherwise people will end up conflicting and it'll all go horribly wrong [09:02:00.0000] It's no good just allocating a flat block of numbers and telling people to do what they want in that range [09:28:00.0000] fixing all these charset details is a mess [10:21:00.0000] hsivonen: http://hsivonen.iki.fi/rdf/ - s/Martial/Martian/ [10:22:00.0000] Philip`: thanks fixed [10:29:00.0000] hsivonen: also, s/on error (last paragraph) [10:45:00.0000] krijnh: thanks fixed. [10:53:00.0000] Looks like half the SVG files on Wikipedia come from Inkscape [10:54:00.0000] and 25% from Adobe Illustrator [10:55:00.0000] Also one of them says [10:55:01.0000] I guess vandalism isn't restricted to content pages :-( [10:56:00.0000] (Mostel_HUAC.svg, in case anyone cares) [10:57:00.0000] heh [11:01:00.0000] someone described MS as having 'red state' developers [11:03:00.0000] ie lots of in-house intranet devs, who are their primary customers, and who have built sites predicated on everyone having the same MS browser, as that's what IT installs [11:11:00.0000] http://philip.html5.org/data/svg-xmlns.txt [11:11:01.0000] There's a lot of xmlns:svg (from Inkscape, I think) but no at all [11:18:00.0000] Hmm, Opera renders http://upload.wikimedia.org/wikipedia/en/2/23/Yonkers_seal.svg [11:18:01.0000] and Firefox doesn't [11:18:02.0000] Saf doesn't [11:19:00.0000] Opera 9.5 doesn't [11:19:01.0000] (9.2 does) [11:19:02.0000] I guess that's just a bug [11:20:00.0000] namespaces are tough [11:21:00.0000] and it affects about 2-3% of the SVG images on Wikipedia [12:22:00.0000] Hixie: ping! [13:08:00.0000] Philip`: the SVG is wrong :) [13:08:01.0000] Philip`: it needs a namespace [13:08:02.0000] glad to see 9.5 has become more strict [13:08:03.0000] andersca: here [13:10:00.0000] Hixie: hi! is this a typo? [13:10:01.0000] A browsing context can be associated with an application cache. A child browsing context is always associated with the same browsing context as its parent browsing context, if any. [13:10:02.0000] should it be "A child browsing context is always associated with the same application cache..." ? [13:11:00.0000] er yes [13:11:01.0000] that makes things more clear :) [13:11:02.0000] please do e-mail the list about that one :-) [13:11:03.0000] will do [13:12:00.0000] thanks [13:13:00.0000] sent [13:50:00.0000] w3.org is now ahead of whatwg.org for html 5 on google [13:50:01.0000] indeed [13:50:02.0000] unsurprising [13:59:00.0000] Hixie: if confidence is Certain and a meta is seen for a different encoding (after alias resolution), shouldn't that be a parse error? [13:59:01.0000] I though it already was but now I don't see it in the tree builder spec [14:09:00.0000] hsivonen: it's an error anyway, since you can't have more than one encoding declaration [14:11:00.0000] Hixie: that doesn't cover the case where it became confident due to HTTP or BOM and meta disagrees [14:11:01.0000] i thought i'd made that non-conforming too [14:11:02.0000] yes, it is [14:11:03.0000] 3.7.5.4. Specifying the document's character encoding [14:11:04.0000] first bullet point [14:13:00.0000] Hixie: in the live dom viewer you can get ie8 working by using try-catch with the ie8 approach in the catch block [14:14:00.0000] there's a separate ie8.html file in that directory at the moment [14:14:01.0000] Hixie: ok. the natural way to implement that is as parse error [14:15:00.0000] hsivonen: that makes sense [14:15:01.0000] yes, but it would be more useful if it worked in the normal version [14:15:02.0000] giving 2 duplicate links every time doesn't scale [14:15:03.0000] s/2// [14:15:04.0000] well, get them to fix ie8 :-) [14:15:05.0000] i don't want to fix the main one to work around bugs in unrelased products [14:16:00.0000] ok [14:17:00.0000] I'm starting to suspect that we are going to need a charset registry 5 so that so that each implementor doesn't need to figure out the magic alias mess [14:18:00.0000] can someone suggest a good behavior-based heuristic for testing if a character decoder is ebcdic-based? [14:18:01.0000] what mess? [14:20:00.0000] Hixie: mapping TIS-620 to Windows-874. mapping GB_2312-80 to GBK. KS_C_5601-1987 to windows-949-2000. x-x-big5 to big5, etc. [14:20:01.0000] some of those will presumably end up in the spec, but i'm hoping that most are not widespread enough to require support [14:20:02.0000] also, if the Gecko alias file and the ICU aliases are any indication of reality, the IANA registry of aliases is not the whole story [14:41:00.0000] http://www.flickr.com/photos/joeclark/192878174/ [14:41:01.0000] ...and other such tables [14:41:02.0000] seem to use indenting as a way of having "subrows" [14:42:00.0000] i wonder if we should support that explicitly somehow [14:44:00.0000] http://www.flickr.com/photos/joeclark/185786265/ is an insane table [14:46:00.0000] how is x-user-defined supposed to work? [14:48:00.0000] WebKit seems to implement x-user-defined as mapping to the first 256 code points of the PUA [15:08:00.0000] this indenting thing is really common [15:09:00.0000] Hixie: URLs specified in the manifest attribute are relative to the document, right? [15:10:00.0000] I sucks considerably that Java doesn't have a rewindable InputStream in the standard library [15:10:01.0000] andersca: yes, in html [15:11:00.0000] one that doesn't require telling it how much to buffer ahead of time, that is [15:11:01.0000] fortunately, I'm not the first person to need it, so I don't have to write it [15:11:02.0000] andersca: in xhtml, if there's an xml:base on the root element, that is honoured [15:11:03.0000] oh, interesting [15:11:04.0000] (assuming you support xml:base) [15:11:05.0000] ok my macbook pro is CLEARLY having overheating issues [15:12:00.0000] it keeps locking up when it gets even remotely warm [15:12:01.0000] must be fun with manifest and [15:12:02.0000] i wonder if the fans are busted or something [15:13:00.0000] Hixie: how do you empty an application cache? is that something that the user agent decides when/how to do? [15:15:00.0000] andersca: I think "Empty Cache" in the UI should do it [15:15:01.0000] yeah [15:16:00.0000] empty cache does it [15:16:01.0000] there's also a way to trigger it [15:16:02.0000] but i forget what it is [15:18:00.0000] swapCache doesn't clear it, or does it? [15:22:00.0000] ok so the way to clear a cache is to simply empty the manifest, but leave it on the server [15:23:00.0000] (and to remove the manifest attributes) [15:24:00.0000] bbiab [15:24:01.0000] right 2008-03-20 [17:24:00.0000] is the idea that the offline cache should only be accessed when navigator.onLine is false? [17:24:01.0000] no [17:24:02.0000] the offline cache is always accessed [17:24:03.0000] is the spec not clear about that? [17:25:00.0000] "The navigator.onLine attribute must return false if the user agent will not contact the network when the user follows links or when a script requests a remote page" [17:25:01.0000] that's what got me confused [17:26:00.0000] ah, yeah, that should probably be clarified [17:26:01.0000] please send mail :-) [17:26:02.0000] will do [17:26:03.0000] that text predates the offline cache [17:26:04.0000] ah, got it [17:26:05.0000] Hixie: also, can an application cache group ever have more than two caches? [17:26:06.0000] yes [17:27:00.0000] if there are many browsing contexts all using the same cache group [17:27:01.0000] and the cache group gets upgraded a number of times [17:27:02.0000] with only one browsing context getting upgraded each time [17:28:00.0000] ah, that makes sense [17:29:00.0000] so I should be able to get rid of older caches as contexts referencing them go away [17:32:00.0000] yeah [17:32:01.0000] cool [17:33:00.0000] this is starting to make sense [19:18:00.0000] i suppose we could handle the indent thing by using multiple colums [19:18:01.0000] but that feels like a hack [22:17:00.0000] ok i think i've looked at enough tables [22:17:01.0000] now to study the ones i marked as being interesting [00:49:00.0000] /me pokes othermaciej [00:49:01.0000] bewes1: yes? [00:49:02.0000] oh, I wanted to ask you about geolocation [00:50:00.0000] I am at your service [00:50:01.0000] I didn't think each HTTP request would re-consume the resources necessary to do geolocation [00:51:00.0000] is it unreasonable to assume the browser has some cached version of the geolocation results? [00:52:00.0000] if there's a cached version it could be completely wrong [00:52:01.0000] without triggering new lookups? [00:52:02.0000] let me use iPhone as an example [00:52:03.0000] so it's a cache invalidation issue? [00:52:04.0000] I believe its geolocation has been discussed a little publicly [00:52:05.0000] it uses a combination of WiFi (looking for nearby networks) and triangulation from cell towers [00:52:06.0000] both of these fire up the radio [00:53:00.0000] and may need to use high power signals [00:53:01.0000] it does not do this all the time [00:53:02.0000] only when you ask [00:53:03.0000] if you put it in your pocket and drive across town, it has no idea if you are still near where you were [00:53:04.0000] and it would have to suck a lot of battery to find out [00:53:05.0000] so it's far better to only provide geolocation info on demand [00:53:06.0000] instead of sending it to every server all the time [00:53:07.0000] yeah, I guess I was imagining there'd be some kind of reliable cache in between [00:54:00.0000] how would you know if your cache is valid without geolocating again? [00:54:01.0000] keep in mind, many sites (perhaps most) will not even use this data [00:54:02.0000] yeah [00:55:00.0000] I would also add that it is awkward to get a request header to script code, and often the interesting location-based things you want to do will be in script [00:55:01.0000] yeah, I guess I was missing some perspective on devices for which it's really really expensive and a pretty unreliable cache [00:55:02.0000] although your points on usage are good too [00:55:03.0000] I think what I said may be true for just about any phone with location support but no hardware GPS [00:56:00.0000] yeah [00:58:00.0000] how do you know where the wifi signals are? do you always require at least one cell tower? [00:58:01.0000] I don't know all of the nitty gritty technical details [01:06:00.0000] fwiw, I use location-based services from both from mobile browsers and other apps on devices here in Japan quite a lot, and location interaction from browsers deployed here is always as othermaciej describes [01:07:00.0000] that is, nothing is cached [01:08:00.0000] but some non-browser apps on mobile devices here are capable of dynamically updating as you move [01:09:00.0000] e.g., they can show you your update position on a map as you move [01:09:01.0000] in near real-time [01:09:02.0000] like the navigation systems in cars [01:10:00.0000] another thing is that the mechanism being used to determine the position isn't really exposed to apps [01:11:00.0000] there's just an API for making a location query, and the device makes a determination about what the optimal mechanism is for determining your location [01:11:01.0000] yeah, I was just curious [01:17:00.0000] MikeSmith: what kinds of apps use the near real-time movement? [01:17:01.0000] bewes1 - apps for trip-planning/trip-routing [01:18:00.0000] most widely deployed one is this: [01:19:00.0000] http://www.navitime.com/howtouse.html [01:19:01.0000] oh, do you mean the ones in cars? I thought you meant devices you'd carry with you all the time [01:19:02.0000] Navitime [01:19:03.0000] not in cars, but on mobile handsets/ mobile phones [01:19:04.0000] but basically the same functionality as car navigation systems [01:19:05.0000] ah [01:20:00.0000] expect that since here in Tokyo most people don't drive cars, the apps are for train commuting and pedestrians [01:20:01.0000] use case is that you just fire up the app and tell it where you want to go [01:21:00.0000] pretty convenient [01:21:01.0000] e.g., give it an address or name of a business or something [01:21:02.0000] then the app shows you the route: tells what train station is closest, the shows you an interactive map of how to get to the train station [01:22:00.0000] that's pretty good stuff [01:22:01.0000] then same thing after you get off the train at the closest station to your destination [01:22:02.0000] does it include when the next train arrives? [01:22:03.0000] yep [01:22:04.0000] it has access to complete train schedules for all train lines in Japan [01:23:00.0000] how often are trains late? [01:24:00.0000] bewes1 - rarely [01:25:00.0000] typically only in cases of really bad weather [01:25:01.0000] or if somebody jumps in front of one of them [01:26:00.0000] btw, the actual Navitime app and others like it is either a Java/J2ME or BREW app, depending on which handset it's on [01:26:01.0000] but there's no real reason that a remote Web-based app could not provide the same features [01:26:02.0000] except that we don't have any standard scripting APIs that would enable Web developers to write such apps [01:59:00.0000] looks like headers='' hasn't registered on html4all yet [02:43:00.0000] hmm. the restriction against entities in the internal character encoding decl is pretty annoying [02:51:00.0000] Hixie: is "# ] [02:51:01.0000] # The encoding name must be serialised without the use of character entity references or character escapes of any kind. " [02:52:00.0000] based on browser behavior? [02:52:01.0000] I assume that's to make the preparse thing work correctly? [02:52:02.0000] the way the spec is written, the tree builder's encoding switching thing works even with escaping [02:53:00.0000] Philip`: presumably [02:54:00.0000] but I have to wonder if this kind of layering-breaking restriction is right [02:55:00.0000] Hixie: assuming that the restriction stays, shouldn't it apply not only to the encoding name but the entire attribute? [02:56:00.0000] bah. I'll leave implementation for another day and send email for now [02:57:00.0000] in a way, this restriction is even more annoying than the old 512-byte restriction [02:57:01.0000] mainly because I implemented that one already [03:03:00.0000] /me only sees half a dozen cases of & in charsets [03:03:01.0000] like [03:03:02.0000] and [03:04:00.0000] and similar things [03:39:00.0000] Philip`, or Hixie: it would be very useful to have data about pages that use " before EOF) [03:40:00.0000] we're a bit scared that we'll break many pages when fixing that [03:53:00.0000] or ... --> (with no --> before ) [03:54:00.0000] (and no further script blocks afterwards) [04:40:00.0000] /me needs a better grep [04:51:00.0000] has anyone tested if meta can cause browsers to switch to a non-UTF-* non-US-ASCII superset encoding an reparse? [05:50:00.0000] zcorpan: http://philip.html5.org/data/pages-with-unclosed-comments.txt shows the pages that have unclosed comments, assuming my regexp was about right [05:52:00.0000] what does the spec change about that relative to Opera's existing behavior? [05:52:01.0000] http://parsetree.validator.nu/?doc=http://brianyeedds.com/ - the HTML5 way doesn't seem to work very well [05:54:00.0000] http://parsetree.validator.nu/?doc=http://clubleonberg.free.fr/ loses content too [05:54:01.0000] http://parsetree.validator.nu/?doc=http://cyrilvictor.photo.free.fr/ too [05:54:02.0000] So it seems a reasonably common problem [05:54:03.0000] Philip`: spec bug or validator.nu bug? [05:55:00.0000] hsivonen: Spec, since it doesn't reparse comments if it finds EOF inside them [05:55:01.0000] reparsing is evil, though [05:55:02.0000] (as far as I'm aware) [05:56:00.0000] Why is it evil in this case? [05:56:01.0000] (when it's only reparsing comment text) [05:56:02.0000] (so it's already got the text saved in memory, and it hasn't been inserting elements that are hard to undo) [05:56:03.0000] 1) I makes crafted network errors (EOF) cause scripts to run [05:57:00.0000] Oh, http://james.html5.org/cgi-bin/parsetree/parsetree.py?uri=http%3A%2F%2Fcyrilvictor.photo.free.fr%2F says something different [05:57:01.0000] 2) It is a world of pain for parser developers [05:59:00.0000] having to rewind the byte stream is seriously bad [05:59:01.0000] http://james.html5.org/cgi-bin/parsetree/parsetree.py?source=%3Cscript%3E%3C!----x%3E%3C/script%3Efoo - html5lib bug? [05:59:02.0000] taking the decoded comment buffer and emulating document.writing it is also bad but not quite as bad [06:19:00.0000] jgraham: Is it intentional that you disabled all the Python tokeniser tests? [06:20:00.0000] Oh, looks like just a typo [06:24:00.0000] /me fixes the --x> thing in html5lib [06:28:00.0000] (Not sure it's the most sensible fix, since it's only really the data='-' case that causes issues, but it should be safe enough...) [06:52:00.0000] http://lists.w3.org/Archives/Public/www-style/2008Mar/0279.html [06:56:00.0000] Philip`: wow thanks! [06:57:00.0000] the first has --!> ... [07:05:00.0000] http://parsetree.validator.nu/?doc=http://www.graniteschools.org/jr/churchill/ [07:05:01.0000] That seems to be the )([^<]|<(?!/script))*-->([^<]|<(?!/script))*$/ which hopefully is something like what you meant by " ... --> (with no --> before " [07:12:00.0000] zcorpan: http://philip.html5.org/data/pages-with-unclosed-scripts-and-comment-stuff.txt [07:13:00.0000] zcorpan: does Opera mitigate the script running risk upon reparse? [07:15:00.0000] Philip`: yep, thanks [07:16:00.0000] hsivonen: i don't understand the question [07:17:00.0000] /me notes that firefox and safari close comments at --!> [07:17:01.0000] and doing so has better compat with pages i've looked at so far [07:18:00.0000] zcorpan: the security characteristics of the bytes in the "comment" changes radically depending on whether they are comments or markup that can run scripts and instatiate iframes [07:19:00.0000] zcorpan: In quirks or standards? [07:19:01.0000] Philip`: both [07:19:02.0000] hsivonen: yeah. the reparsed comment can run scripts and instatiate iframes... [07:20:00.0000] zcorpan: is only special in quirks [07:20:02.0000] i'm testing in firefox 3 [07:21:00.0000] http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C!doctype%20html%3E%3C!--x--!%3E%20x%20--%3E [07:22:00.0000] http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C!doctype%20html%3E%3C!--x--foo%3E%20x%20--%3E [07:22:01.0000] /me notes that some pages use // -- > [07:22:02.0000] and [07:23:00.0000] http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C!--x--foo%3E%20x%20--%3E [07:23:01.0000] http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C!--x--!%3E%20x%20--%3E [07:23:02.0000] Philip`: ah [07:25:00.0000] s/for any foo/for any foo not containing an odd number of '--'/ [07:26:00.0000] and is not > :) [07:26:01.0000] gotta love the simplicity of comments [07:27:00.0000] foo = '>' will still cause the comment to close when writing " the same as "" (where N = number of columns) [11:10:01.0000] instead of treating the former in a separate way [11:14:00.0000] data:text/html;base64,PHRhYmxlPjx0cj48dGg%2BSDx0aD5IPHRoPkg8dHI%2BPHRoPkg8dHI%2BPHRoPkg8dGQ%2BRDx0ZD5EPHRyPjx0aCBjb2xzcGFuPTM%2BSDx0cj48dGg%2BSDx0ZD5EPHRkPkQ8dHI%2BPHRoPkg8dHI%2BPHRoPkg8dGQ%2BRDx0ZD5EPHRyPg%3D%3D [11:14:01.0000] what should the default interpretation of that table be? [11:24:00.0000] /me trips on some passing tumbleweed [11:25:00.0000] I think the browser should add a elements of

b" as we'd want because it pushes a <#bucket> node onto the stack when doing foster parenting [15:59:00.0000] jgraham: i pasted the html5 spec into the outliner algorithm and it didn't do anything (After hanging safari for a few minutes) [15:59:01.0000] just said "updating..." 2008-03-05 [16:00:00.0000] Hixie: The HTML5 spec seems to kill lxml, which I;m using to construct the tree [16:00:01.0000] ah [16:01:00.0000] (I haven't got as far as making the updating... thing go away if there is an error) [16:01:01.0000] specifically lxml 1.3.6 doesn't like the doctype [16:08:00.0000] heh [16:10:00.0000] /me fixes the bug in html5lib; will deploy tomorrow [16:10:01.0000] /me wonders if jgraham is the author of html5lib [16:15:00.0000] eseidel: I am one of the authors along with annevk and various others [16:18:00.0000] Hixie: It doesn't work quite right on the spec atm [16:19:00.0000] but that's another problem for tomorrow [16:19:01.0000] /me -> bed [16:26:00.0000] nn [16:32:00.0000] /me looks at some data and discovers surprisingly that the web is a mess [16:33:00.0000] It looks like it's actually about as broken as the rest of the internet [16:55:00.0000] On the positive side, 100% of pages that declare themselves as iso-8859-1 can be decoded with no errors [16:56:00.0000] On the less positive side, only 94% of pages declared as utf-8 can be decoded without errors [16:56:01.0000] and 84% of gb2312 [17:02:00.0000] http://validator.nu/?doc=http://tools.ietf.org/html/rfc303 - some people can't even get us-ascii right [17:39:00.0000] How does MetaSniffer think that pages like http://fqq.com/ have charset "x-windows-874"? [17:40:00.0000] There's 58 of them, and none seem to actually say "x-windows-874" anywhere inside them... [17:43:00.0000] Oh, that's because I'm asking it for the canonical name of the sniffed charset [20:47:00.0000] these fixes are proving remarkably easy to do [23:05:00.0000] Hixie: would you consider the pause of a WebKit release build on Acid3 (around test 26) to constitute a failure? [23:06:00.0000] (I think it pauses less than any other browser but it's still noticeable) [23:09:00.0000] yes, though obviously nowhere as serious as anything else [23:10:00.0000] I'm not sure it is possible to pass then [23:11:00.0000] but I guess we'll look at it once we fix everything else [23:11:01.0000] (up to 90/100 now) [23:53:00.0000] Hixie, not, the current spec treats a [23:56:00.0000] hm? [23:57:00.0000] how do you mean? [23:58:00.0000]
X
element saying "ERROR: Table is too complex" [11:28:00.0000] heh [11:30:00.0000] It seems helpful for authors if the algorithm is simple and predictable enough that they can tell whether their slightly-fancy table is going to be automatically headerised correctly, or if they're going to have to manually annotate it [11:30:01.0000] at least for authors who know a bit about the table thing, and care about it, but can't be bothered to use some awkward ugly tool to tell them how the table really is processed [11:31:00.0000] (rather than trying to make the algorithm really complex so it handles lots of weird edge cases) [11:46:00.0000] Hixie: "If the resource is not being loaded as part of navigation of a top-level browsing context" [11:47:00.0000] Hixie: that statement would be true when loading something in an iframe, right? [12:03:00.0000] andersca: or a stylesheet, or an image, or XMLHttpRequest, etc, yeah [12:03:01.0000] or script [12:03:02.0000] or any number of other things [12:04:00.0000] basically anything except the maii document :-) [12:04:01.0000] main [12:04:02.0000] although I guess that only xmlhttprequests can cause the cache selection algorithm to be invoked [12:05:00.0000] i think the cache selection algorithm can only be invoked when there's a browsing context [12:05:01.0000] ah, yeah [12:46:00.0000] http://www.pcworld.com/article/id,140408/article.html is funny [12:47:00.0000] (given safari 3.1) [12:47:01.0000] ...nothing about safari :( [12:48:00.0000] yeah i love that the only browser to actually _ship_
[07:40:01.0000] (and that it should be allowed and perhaps that the html parser shouldn't imply -- i've sent email about this if i recall correctly) [07:41:00.0000] zcorpan_: re: attributes defaulted from a DTD: If the user chooses DTD processing, Validator.nu processes the DTD. If that sucks for the user, the users shouldn't ask for DTD processing [07:41:01.0000] hsivonen: fair enough :) [07:41:02.0000] zcorpan_: however, I have considered cloning Gecko's DTD behavior and offering that as an option [07:42:00.0000] but I don't want to do it unless there's agreement that Gecko's DTD behavior should be grandfathered into the Web platform [07:58:00.0000] hsivonen: I haven't tried using getProbableCharsets at all [07:59:00.0000] Philip`: ok [08:00:00.0000] csarven: is there any point other than self-promotion with pictures and minibios? [08:01:00.0000] BenMillard: I think tha's a fair summary although I guess Hixie can come up with an entirely novel algorithm [08:03:00.0000] hsivonen My thoughts exactly! [08:16:00.0000] jgraham_mibbit, thanks and that's a possibility zcorpan mentioned. [10:30:00.0000] hsivonen: if you want to let the html4all people know, there will be a gigantic e-mail coming out relatively soon discussing all the table edits in great detail [10:30:01.0000] hsivonen: i just didn't want to delay the edits until i'd done editing the spec [10:31:00.0000] er, until i'd done replying to the e-mail, i mean [11:54:00.0000] http://wiki.whatwg.org/wiki/Special:Recentchanges - Mpsingh looks like a spammer [12:24:00.0000] Hixie: found a tiny typo in the spec [12:24:01.0000] Explicit entries can also be marked as foreign, which means that they have an manifest attribute but that it doesn't point at this cache's manifest. [12:24:02.0000] an manifest => a manifest [12:30:00.0000] Hixie: mentioned as part of replying on html4all [12:48:00.0000] Hixie: also, ping [12:48:01.0000] :) [12:53:00.0000] /me summons hixie too, for good measure [12:54:00.0000] cool [13:25:00.0000] andersca, Pavlov_: here [13:26:00.0000] andersca: and sorry to be a pain, but if you could, please mail the typo to me :-) (either the list or ian⊙hc if you don't want to spam the list) [13:26:01.0000] Hixie: I would rather have all those typos exposed on the whatwg list ;) [13:27:00.0000] sure thing :-) [13:27:01.0000] Hixie: also, I was wondering why null characters in the manifest file should be converted to fffd? [13:28:00.0000] to avoid any problems with implementors wanting to use null-terminated strings [13:29:00.0000] it's been a source of many problems in other parts of the web platform [13:30:00.0000] ap: did you see that [13:31:00.0000] Hixie: the problems in other parts were fixed without changing decoding of nulls - I think it would be better to have a consistent approach [13:31:01.0000] the html5 parser handles nulls in much the same way [13:31:02.0000] Hixie: hmm, do any browsers follow that? [13:32:00.0000] Hixie: for all I know, WebKit just treats nulls as normal characters [13:32:01.0000] some do. others treat them as nulls, and yet others have security bugs (they just strip out the nulls, or truncate at nulls, both of which can be used for xss or similar behaviours) [13:33:00.0000] to be honest it really doesn't matter much what we do with nulls here, since none of the magic strings contain nulls, and URIs aren't allowed to contain nulls. [13:33:01.0000] (or fffds) [13:33:02.0000] i don't really understand why it's a problem [13:33:03.0000] so long as the behaviour is indistinguishable from treating nulls as FFFDs [13:33:04.0000] Hixie: I'd just prefer to avoid introducing yet another mode in the decoder [13:34:00.0000] the spec doesn't say anything about how you implement it [13:34:01.0000] Hixie: hmm, I think andersca was talking about this phrase from the spec: "All U+0000 NULL characters must be replaced by U+FFFD REPLACEMENT CHARACTERs." [13:35:00.0000] right, so long as the result is equivalent to this, it doesn't matter how you do it [13:35:01.0000] in fact i think that line is effectively redundant, so long as you correctly detect whether URIs are valid or not [13:36:00.0000] why not just wait for Opera to fix their null character handling bugs [13:36:01.0000] since both FFFD and 0000 are invalid in URIs [13:36:02.0000] Hixie: ok, I guess I didn't have enough context, and thought that this had practical consequences [13:36:03.0000] oh, sorry about that :( [13:36:04.0000] othermaciej: this isn't about those bugs [13:36:05.0000] othermaciej: but the null->fffd stuff in the parser is there because nulls have caused far many more problems than just in opera [13:37:00.0000] e.g. the status bar bug several years back [13:37:01.0000] and since there really is no use case for U+0000 on the Web, it seems far safer to just remove it altogether [13:37:02.0000] nulls are certainly risky when you mix with C APIs [13:37:03.0000] but converting them in some places but not others is an annoying burden [13:37:04.0000] Hixie: sounds like a warning could have better effect than a somewhat misguiding requirement in a sample algorithm [13:38:00.0000] and you really want your code to be safe against nulls in strings anyway [13:38:01.0000] ap: send mail :-) [13:38:02.0000] Hixie: k [13:38:03.0000] thanks [13:38:04.0000] but if the conversion doesn't affect results of the algorithm then it does not matter much in this case [13:39:00.0000] othermaciej: in the manifest case, i don't think there's any noticeable effect, and we can probably remove it [13:39:01.0000] othermaciej: in the html parser case, i agree that it is unfortunate that you can still introduce nulls through the DOM APIs [13:39:02.0000] othermaciej: not much i can do about that, though [13:40:00.0000] /me blames the Unicode consortium [13:40:01.0000] othermaciej: still seems far safer and better to make nulls become FFFDs, since you already have to have code to turn a number of other things into FFFDs and have to do newline normalisation anyway [13:40:02.0000] anyway [13:40:03.0000] please discuss this on a mailing list :-) [13:40:04.0000] /me is working on tables [13:42:00.0000] eww! you can't use tables! they're presentational! divs ftw! [13:46:00.0000] /me likes \0 to FFFD substitution in the parser [13:52:00.0000] hixie, is there anything I can be doing to help your work on tables? [13:58:00.0000] /me sees the last paragraph of http://html4all.org/mailman/archives/list_html4all.org/2008-March/000663.html and conducts some research [13:58:01.0000] Out of about 130,000 random pages from dmoz.org, the 'axis' attribute is used on http://www.cdsm.co.uk/ and http://centerville.lionwap.org/ and http://gazianteplionsclubtr.lionwap.org/ [13:58:02.0000] /me concludes that it's quite rare, and that he has insufficient data to conclude anything else [13:59:00.0000] Philip`: I assume you keep a local cache of 130000 pages? [13:59:01.0000] hsivonen: Yes [14:00:00.0000] how much disk space does the cache take? [14:00:01.0000] I wouldn't want the cost of downloading ~3GB of content every time I wanted to find a little more information :-) [14:00:02.0000] hsivonen: ~3GB [14:00:03.0000] (having clamped each page to 256KB) [14:01:00.0000] (since the (unclamped) mean size of a page is infinite) [14:02:00.0000] If disk space was a concern, I imagine compression could help significantly [14:05:00.0000] (http://triin.net/archive/kool/webstat/figure-8.png suggests almost everything is <=256KB, so that seems a reasonable limit to impose) [14:05:01.0000] (though actually I picked the value pretty arbitrarily) [14:07:00.0000] If I have understood axis correctly (not a safe assuption), http://gazianteplionsclubtr.lionwap.org/ uses axis in a useless way [14:07:01.0000] as does http://centerville.lionwap.org/ [14:08:00.0000] and http://www.cdsm.co.uk/ [14:09:00.0000] /me is pretty sure axis is useless [14:12:00.0000] At least it's abused far less frequently than many other attributes [14:24:00.0000] /me agrees with jgraham [14:29:00.0000] I'm off for some supper, some TV and then bed. bye all [14:29:01.0000] ye [14:29:02.0000] bye even [14:30:00.0000] cya [15:11:00.0000] http://james.html5.org/graph_2.svg -> possibly all the state transitions made in the html5lib treebuilder testcases (assuming I didn't screw up) [15:12:00.0000] /me will work out how to distinguish "real" transitions from "reprocess as if in" transitions next (that graph should have both) [15:13:00.0000] how do you go from AfterBodyPhase to InFramesetPhase? [15:14:00.0000] oh that's the "process using the rules for" thing [15:14:01.0000] nevermind [15:14:02.0000] there are iirc only four insertion modes that can be subject to that, btw [15:14:03.0000] if that helps optimise code [15:49:00.0000] jgraham_: That's quite nice :-) [15:52:00.0000] - doesn't it need to quote the string? [15:53:00.0000] no [15:53:01.0000] Oh [15:53:02.0000] it does, however, need units on the font-size [15:54:00.0000] Oh [15:54:01.0000] Hixie: in SVG? [15:54:02.0000] yup [15:54:03.0000] unless you use the font-size="" attribute [15:55:00.0000] in retrospec, it might have been good to make unitless default to px in CSS even in the standards mode [15:55:01.0000] except that would clash with unitless numbers in line-height, font, and other properties [15:55:02.0000] unitless line-height is redundant with % [15:56:00.0000] nope [15:56:01.0000] they inherit differently [15:56:02.0000] too difficult :-) [15:56:03.0000] (numbers allow lines to expand to fit children with bigger font sizes, %s don't) [15:56:04.0000] i'm not defending the design of the language [15:57:00.0000] just explaining why the svg group got it wrong [15:59:00.0000] http://www.gerv.net/security/link-fingerprints/ - "half a loaf is better than no bread" - I think that's not true in general - the half loaf might be covered in mould, and it'd be better if you didn't have any bread at all [16:00:00.0000] i think we already discussed fingerprints on the list once [16:00:01.0000] http://mindforks.blogspot.com/2008/03/aria-templateid-explained.html [16:00:02.0000] and came to the conclusion it wasn't a good idea [16:00:03.0000] i was hoping someone would find the reference and reply to him before i had to :-) [16:00:04.0000] I wonder how aria-templateid scales on the time axis given the update cycle of AT [16:00:05.0000] compared to the update cycle of Web 2.0 apps [16:00:06.0000] /me wonders what happens if he sets aria-templateid="google.com/gmail" on his blog [16:01:00.0000] what's aria-templateid? [16:01:01.0000] http://mindforks.blogspot.com/2008/03/aria-templateid-explained.html [16:02:00.0000] /me tends to download files by copying the URL and pasting it into wget, and then getting bash errors and doing it again with quotes around the URL argument, which would avoid the benefits of link hashing [16:02:01.0000] that sounds like an amazingly bad idea [16:02:02.0000] othermaciej: you hate blind people! [16:03:00.0000] othermaciej: you are not the first person with that initial reaction [16:04:00.0000] I can't imagine AT vendors actually implementing support for it [16:04:01.0000] Hixie: If you did that, it would make your blog less usable to some people with AT, and there are much easier ways you can cause problems for those people, so I don't see why it matters [16:05:00.0000] Philip`: it seems like reducing the ways that authors can (accidentally) screw over their users would be a good thing [16:06:00.0000] Philip`: and, similarly, increasing the number of ways that authors can screw over their users would be a bad thing [16:06:01.0000] Why would someone accidentally set aria-templateid="google.com/gmail" on their site? [16:06:02.0000] the saddest thing about it is that it seems to presume that web sites themselves won't use ARIA to give a good enough accessibility experience, so that AT needs special-case hacks for different sites [16:06:03.0000] Philip`: why would someone accidentally put as their doctype? [16:06:04.0000] othermaciej: firevox shows that this is already happening [16:07:00.0000] othermaciej: though why a site would care enough to put an aria-templateid attribute but not enough to make their site usable, i dunno [16:07:01.0000] I find the AT upgrade cycle a much more persuasive argument :) [16:08:00.0000] we have indeed heard many times that the brutally slow AT upgrade cycle makes changing accessibility features fruitless [16:08:01.0000] Hixie: There's an infinite number of errors like that that people can accidentally introduce; but there's only a (pretty small) finite number of aria-templateid="..." strings that any AT would care about, so there's a vastly smaller chance of that happening by accident [16:08:02.0000] for instance that a new table header association algorithm was worthless, since it would be forever until AT supports it, and twice forever until users who need it have that version [16:09:00.0000] Philip`: except that the sites that would use it are those most likely to be held up as examples to learn from [16:09:01.0000] othermaciej: way to rain on what i'm doing on my other monitor :-P [16:10:00.0000] aria-templateid allows competition in (e.g.) the Gmail-usability space, because lots of vendors (and extension developers etc) can all try to make it work as well as possible; otherwise only Google could improve Gmail usability, so they're going to do less well than when there are many competing attempts [16:11:00.0000] nonsense [16:11:01.0000] Hixie: I'm not saying I agree with such sentiments [16:11:02.0000] Philip`: ATs could easily just hardcode the gmail uris [16:11:03.0000] Philip`: Replace templateid with the URL and you have a user script system that works today [16:12:00.0000] Maybe Google Maps is a better example, since that often gets embedded in other web pages and you can't hardcode the page URLs [16:12:01.0000] (Yay mashups!) [16:13:00.0000] that might hypothetically work, if templateid is on a per-subtree level, but that would need extremely careful definition [16:14:00.0000] (the visual part of Google maps is very visual though) [16:14:01.0000] horrah, ms feedback on storage [16:15:00.0000] I suppose the ARIA spec is a little vague on aria-templateid, but maybe someone will blog about it in an extremely careful detailed way [16:15:01.0000] man, i should use that model to write html5 [16:15:02.0000] that would be so much easier [16:16:00.0000] hsivonen: (There are many disabled people who aren't blind, so AT improvements on primarily visual content would still help some users) [16:16:01.0000] Philip`: you are aware of the wiki guide, right? [16:17:00.0000] hsivonen: I'm not, and Google just tells me about a science fiction manga [16:17:01.0000] Oh, is that the Mozilla one? [16:17:02.0000] Philip`: http://developer.mozilla.org/en/docs/ARIA_UA_Best_Practices [16:18:00.0000] "aria-templateid: XXX not sure what to do with this one" [16:18:01.0000] Hmm [16:22:00.0000] /me just sees http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2006-November/thread.html#7825 about the link fingerprint thing [16:23:00.0000] that's the one [16:26:00.0000] thanks Philip` [16:28:00.0000] /me thanks the "Search Mail" button [16:33:00.0000] /me wonders how strange a validator feature request would have to be before hsivonen would just reject it out of hand [16:33:01.0000] hsivonen: so i'm tempted to ask for support for an equivalent of the #line preprocessor directive... [16:33:02.0000] /me wonders what happens if the target of checksummed link is replaced by an HTML file containing a malicious script followed by enough stuff to make a browser want to incrementally render before it's finished downloading and can check the checksum [16:33:03.0000] (i edit a file that is then concatenated to another one before validation, so the line numbers are all 82 lines off for me) [16:34:00.0000] crap, gotta go [16:34:01.0000] if people want to check it out, i just regenned the spec with the new algorithm for tables [16:34:02.0000] bbl [16:36:00.0000] Are there any UAs that support inline pseudo-classes? [16:38:00.0000] Hixie: cool 2008-03-22 [19:26:00.0000] ping Hixie [19:26:01.0000] with a cherry on top [19:28:00.0000] /me yodels in hopes of attracting attention [19:31:00.0000] alas [01:09:00.0000] There are no bugs, only features. [02:37:00.0000] /me notices that http://www.robbyslaughter.com/blog/?2008-03-21 , saying "You might have noticed that not all websites work perfectly", does not work in Opera since it renders as white text on a white background [02:43:00.0000] i wonder why that happens [02:44:00.0000] I'd guess it's to do with style inheritance into tables in quirks mode, which no browsers are particularly consistent on [02:48:00.0000] Unicode and Math really messed up with φ. Not good. [02:49:00.0000] hsivonen: how so? [02:49:01.0000] Unicode swapped the glyphs for two code points around. [02:50:00.0000] then Math swapped the entities around to keep the entity-glyph mapping [02:50:01.0000] now φ means different things in Gecko depending on doctype [02:51:00.0000] oy vay [03:26:00.0000] FYI: http://groups.google.com/group/mozilla.dev.tech.mathml/browse_thread/thread/e7f7efbb5e161348/9fde74f46fb0b5d2#9fde74f46fb0b5d2 [03:28:00.0000] not sure if preventing grief for Safari is an argument that will fly :-) [03:28:01.0000] but that does sound like a good idea [03:31:00.0000] I read the excessive DTD traffic article a few days ago...found it funny in a tragic sort of way. [03:57:00.0000] othermaciej: I'm hoping it works in the "do unto others" sense and with the Mission of the Foundation [05:26:00.0000] looks like the TAG issue list has even older items than the WHATWG issue list... http://www.w3.org/2001/tag/issues.html?view=normal&closed=1&type=1 [05:41:00.0000] hsivonen - do you keep up with the httpbis list? [05:42:00.0000] /me sighs at the very mention of that list [05:42:01.0000] MikeSmith: I don't. [05:43:00.0000] MikeSmith: is there something going on there that I should be aware of? [05:43:01.0000] not that I know of. was just asking because I've got a bunch of unread messages from that and trying to figure out what threads to read [05:44:00.0000] I see IDNAbis in the subjects. That can't be good. [05:44:01.0000] heh [05:44:02.0000] yep [05:44:03.0000] hsivonen: the recent discussion is all idealistic as if compat. with current non-conforming impls is irrelevant [05:47:00.0000] eek. Are they really contemplating allowing non-ASCII on the HTTP "GET" line? [05:48:00.0000] I dunno. I gave up reading it. [05:48:01.0000] :) [05:49:00.0000] It's one thing to insist on a multitude of human scripts in user-readable prose and to insist on non-ASCII in computer-to-computer messages that are text-based for programmer convenience [05:50:00.0000] s/and/and another/ [05:51:00.0000] /me thinks people who advocate non-ASCII in programming language variable/method names or XML element/attribute names are misguided [06:35:00.0000] hmm. does SVG really have no IDREF or IDREFS attributes? [06:41:00.0000] do UAs actually implement SVG as a DOM tree clone or as an anonymous layout object tree? [06:41:01.0000] or something else? [06:42:00.0000] "Property inheritance, however, works as if the referenced element had been textually included as a deeply cloned [06:42:01.0000] child of the 'use' element." [06:42:02.0000] that sounds really bad [06:45:00.0000] hmm. looks like the SVG WG has created a need for a hashed ID reference in their spec, as well [06:46:00.0000] "● local URI references, where the URI reference does not contain an or and [06:46:01.0000] thus only contains a fragment identifier (i.e., # or #xpointer(id)) [06:46:02.0000] " [06:46:03.0000] is the xpointer alternative interoperable? [06:47:00.0000] also, in compound documents, can are "local" references local to the SVG subtree or to the entire document? [07:00:00.0000] are cross-document and supposed to work? [07:02:00.0000] hsivonen: http://krijnhoetmer.nl/irc-logs/whatwg/20080311#l-655 is relevant about implementation of SVG [07:04:00.0000] Philip`: yeah. I'm trying to figure out if I should try to validate it as a mere syntactic IRI or as a hashed ID reference whose referent is actually checked for sanity [07:05:00.0000] I do agree with othermaciej's assessment [10:23:00.0000] Just found thant input.click() in some input types elements doesn't call the onclick handler in Firefox. Strange. [10:23:01.0000] testcase http://a.met.cz/oreilly/flanagan/ [10:24:00.0000] it ignores type text, password, file, hidden - all textlike input elements [11:37:00.0000] Hixie: ping (when you're done chatting with Bernd....no hurry or anything) [11:45:00.0000] yo [11:46:00.0000] was working on html5'izing a site other day, & noticed i couldn't get valid dir listing pages on an apache 1.3 server...even using custom header and readme files. prob is that they get a pre with a hr inside. [11:47:00.0000] i presume that was valid html 3.2 [11:48:00.0000] can get em to work with apache 2.something options. worth a special case in 5? [11:49:00.0000] nah [11:49:01.0000] those pages don't have an html5 doctype anyway [11:49:02.0000] /me has to get a _real_ hoster :) [11:50:00.0000] not by default they don't anyway [11:51:00.0000] tend to agree......real edge case [11:53:00.0000] 1.3's kind of a dinosaur already (but plenty of em still around, i'd imagine) [11:57:00.0000] on unrelated matter...filed a html5-related moz 1.9 bug the other day you might wanna cc yourself on...which cropped up with the recent parser change [11:58:00.0000] bug 423721 legend child of figure or unknown element should not generate fieldset [11:59:00.0000] feel free to steal /modify the testcase there if you want [13:22:00.0000] a-ja: About "plenty of em still around": Counting 130K pages from dmoz.org, I see 33K Apache 1.3, 16K of 2.0, 9K of 2.2, 23K unlabelled Apache [13:22:01.0000] so I'd agree there's plenty [13:23:00.0000] (Also I see 22K IIS 6.0, 7K 5.0, negligible other versions) [13:41:00.0000] it irks me that the SVG 1.1 and 1.2 Tiny specs don't tell me if they expect and to work across documents. [13:42:00.0000] I guess I have to write a test case and find out. [13:52:00.0000] findings so far: Cross-doc doesn't work in Firefox, but the impl does URI resolution and isn't just like the hashed ID ref [13:53:00.0000] SVG still quits Safari on me. Odd [13:54:00.0000] hsivonen: do you have a specific SVG that does it? [13:54:01.0000] othermaciej: yes, I have [13:55:00.0000] Opera appears to implement real cross-doc [13:55:01.0000] hsivonen: can you file a bug, or mail it to me? [13:55:02.0000] othermaciej: I've already filed a bug [13:55:03.0000] hsivonen: I wouldn't be surprised if your impl has bugs [13:55:04.0000] I freaking hate [13:55:05.0000] hsivonen: thanks! [13:56:00.0000] othermaciej: unfortunately, the information I've provided isn't very useful [13:56:01.0000] hsivonen: if you included the file that crashes, that's useful enough [13:57:00.0000] oops. I've missed a new comment in the bug [13:59:00.0000] /me tries to get a backtrace now [14:03:00.0000] whoa. I have the Adobe Plug-in [14:03:01.0000] http://scripts.indisguise.org/2008/03/22/xhtml-versus-html/ [14:05:00.0000] hsivonen: oh - we'll use ASVG when it's installed [14:05:01.0000] I think maybe we should stop preferring it without explicit opt-in [14:05:02.0000] commented: http://bugs.webkit.org/show_bug.cgi?id=17073 [14:06:00.0000] sorry about blaming WebKit. I had no idea I had Adobe residue on the system [14:07:00.0000] looks like WebKit implement as a hashed ID reference [14:07:01.0000] Firefox does URI-resolution and if it points to a fragment in the same doc, uses the fragment [14:08:00.0000] Opera actually allows references to different HTTP resources [14:08:01.0000] yay for interop [14:08:02.0000] so should Validator.nu allow a full IRI, a full URI or require a hashed ID ref? [14:09:00.0000] actually, I should rerun the Opera and Firefox tests in case either loaded the Adobe plug-in [14:11:00.0000] those results hold [14:12:00.0000] Philip`: tks for #'s (was AFK) [14:14:00.0000] Hixie: I'd love to see some numbers proving an XML parser marginally faster especially in the case that it is reading a DTD (even an abridged local one) [14:16:00.0000] I am pretty sure that in WebKit the XML parser is slower than the HTML parser [14:16:01.0000] at least, I would be very very surprised if this was not true [14:18:00.0000] othermaciej: are you claiming that going from UTF-8 to UTF-16 to UTF-8 and back to UTF-16 is slow ;) [14:18:01.0000] weinig: that's kind of what I am assuming [14:19:00.0000] (and I hope we can fix it) [14:19:01.0000] weinig: why do you go once to UTF-8 in between? [14:20:00.0000] hsivonen: libxml only operates on UTF-8 [14:20:01.0000] weinig: oh so does your stream decoder always expand to UTF-16 first? [14:21:00.0000] hsivonen: yes [14:21:01.0000] I see. [16:40:00.0000] paging Dr. Hixie [16:41:00.0000] hsivonen, SVG Tiny 1.2 does say, in the Reference Restrictions section: http://www.w3.org/TR/SVGMobile12/linking.html#ReferenceRestrictions [16:42:00.0000] 1.1 doesn't though [16:44:00.0000] correction, 1.1 does: http://www.w3.org/TR/SVG11/struct.html#uriReferenceDefinition 2008-03-23 [18:07:00.0000] no Hixie? [18:28:00.0000] dglazkov: here now [18:28:01.0000] cool! [00:35:00.0000] anyone have IE? [00:35:01.0000] http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C!DOCTYPE%20html%3E%0D%0A%3Ctable%20border%3E%0D%0A%3Ctr%3E%3Ctd%20rowspan%3D1%3EX%3Ctd%20rowspan%3D2%3EY%3Cbr%3EY%3Ctd%20rowspan%3D3%3EZ%3Cbr%3EZ%3Cbr%3EZ%3Cbr%3EZ%0D%0A%3Ctbody%3E%0D%0A%3Ctr%3E%3Ctd%3Ezzz [00:35:02.0000] i need to know if the cells in the first row group are the same height [00:36:00.0000] or whether they do what safari does [00:39:00.0000] sec [00:39:01.0000] X YY and ZZZ? [00:40:00.0000] er, ZZZZ? [00:40:01.0000] if so, they are all the same height for me in IE6 [00:41:00.0000] oh i was hoping for IE8 [00:41:01.0000] sorry [00:41:02.0000] ah [00:41:03.0000] sorry [00:41:04.0000] np [00:41:05.0000] i doubt it changed anyway [00:42:00.0000] thanks [01:09:00.0000] maybe we should just drop xml:base altoegether [01:48:00.0000] heycam: thanks. [01:52:00.0000] np [01:55:00.0000] Is the error handling well enough specified these days for there to be a reference resulting parse tree for http://virtuelvis.com/download/162/evilml.html ? [01:55:01.0000] yes [01:59:00.0000] Ok. Results aren't too encouraging, then [02:00:00.0000] I can't test the DOM Viewer outpout in Midori (Webkit-based GTK app), since it crashes the browser [02:00:01.0000] heycam: would you consider it appropriate if I made Validator.nu complain about if pattern had a full URI instead of just a hashed ID ref? [02:00:02.0000] DOM Viewer handles it ok in Safari 3.1 [02:01:00.0000] http://james.html5.org/cgi-bin/parsetree/parsetree.py?source=%3C%21DOCTYPE+HTML+PUBLIC+%22-%2F%2FW3C%2F%2FDTD+HTML+4.01%2F%2FEN%22+%0D%0A%22http%3A%2F%2Fwww.w3.org%2FTR%2Fhtml4%2Fstrict.dtd%22%3E%3Chtml%3Chead%3Ctitle%3EWhat+does+your+HTML+%0D%0Aparser+make+of+this%3F%3C%2F%3E%3C%2F%3E%3Cbody%3Ch1%3Cem%3EEmphasized%3C%2F%3E+in+%26lt%3Bh1%26gt%3B%3C%2F%3E%3Cp%3Ca+%0D%0Ahref%3D%22http%3A%2F%2Fwww.example.com%22%3Cem%3EEmphasis%3C%2F%3E+in+links+as+well%3C%2F%3E.+%0D [02:01:01.0000] er [02:01:02.0000] http://james.html5.org/cgi-bin/parsetree/parsetree.py [02:01:03.0000] othermaciej: Midori isn't exactly stable, but it's the only reasonable way I can run it under Gnome [02:01:04.0000] is there a bug here: http://parsetree.validator.nu/?doc=http://virtuelvis.com/download/162/evilml.html [02:02:00.0000] apparently not [02:03:00.0000] it's just confusing that "<" can become part of an element name [02:03:01.0000] fwiw in WebKit we have made no attempt yet to better match HTML5 standard parsing [02:05:00.0000] Hixie: what's the deal with allowing "<" in an element name? is it an IEism? none of Gecko, WebKit and Opera do it [02:08:00.0000] hsivonen, that sounds fine for svg 1.1. but for tiny 1.2, i'd only do it if you also complain about other unsupported values, e.g. fill="blah" [02:08:01.0000] since it's the same class of error as that [02:08:02.0000] /me goes for dinner for sure this time [02:08:03.0000] heycam: thanks [02:14:00.0000] hsivonen: someone asked for < not to end tags, so, it doesn't [02:15:00.0000] send mail if you want to have it changed (but do let me know what it should be changed to) [02:25:00.0000] Hixie: email sent [02:26:00.0000] thanks [02:28:00.0000] /me is unhappy that SVG uses URIs instead of honest IDREFs for references that must be intra-document [02:43:00.0000] hsivonen: I believe it is on a theory of future extensibility [02:46:00.0000] othermaciej: in that case, the Safari impl. isn't forward-compatible [02:47:00.0000] you mean that we just treat is as an idref, not resolve as a URI and then check if it is the same document? [02:47:01.0000] othermaciej: yes [02:48:00.0000] probably a bug that we treat it that way, but SVG is not yet that huge an area of concern [02:49:00.0000] so besides requiring the parser to create elements with "<" in the name, does current html5 consider that an error? [02:59:00.0000] othermaciej: no, the error will be caught on the next layer. [03:44:00.0000] /me doesn't like ARIA's XSD datatype dependency [03:44:01.0000] /me would prefer Web Forms 2.0 datatypes [04:05:00.0000] the conformance requirements for ARIA need some work [09:23:00.0000] 3.4.3.1 Header Levels versus Nesting Levels in http://www.w3.org/TR/wai-aria-practices/ is just sad [09:28:00.0000] hsivonen, is there anything in that document that isn't sad? [09:29:00.0000] Lachy_: given that Ajax libraries are building widgets out of divs, annotating those with ARIA is better than putting one's head in sand and insisting that the div widgets aren't happening [09:29:01.0000] Lachy_: but yeah, even the div widgets are kinda sad [09:43:00.0000] http://gsnedders.html5.org/tests/content-type.php [11:28:00.0000] hsivonen: now's a good time to raise aria issues [11:28:01.0000] hsivonen: i agree with most of what you said [11:28:02.0000] hsivonen: i'd go even further, and say that you shouldn't make one element take the semantics of another when those semantics affect the wider page, e.g. making a paragraph into a header [12:00:00.0000] Hixie: ok. I've been thinking I should draft an ARIA-in-HTML5 integration proposal from the document conformance point of view [12:01:00.0000] that'd be very interesting [12:01:01.0000] once we converge on what's conforming, we can consider what behavior to prescribe for the non-conforming cases [12:01:02.0000] also from the point of view of a non-CSS UA [12:01:03.0000] well [12:01:04.0000] my perspective on things is that we should avoid giving authors rope to hang themselves with, generally [12:02:00.0000] hence how many of the things i design from scratch just don't have error conditions [12:02:01.0000] they just have well-defined semantics for everything possible [12:02:02.0000] Hixie: Unfortunately aria has lots of error conditions :) [12:02:03.0000] I'm particularly concerned that the "best practices" doc has examples that fail to use semantics that have been in HTML for over a decade and that wouldn't be disruptive for Ajax libraries [12:03:00.0000] also, from a selfish point of view, I'm not at all a fan of aria-datatype and aria-owns [12:04:00.0000] jgraham: do you recall what format i have to use for "ISSUE-20"-type markers in e-mails to make my e-mails get associated with tracker issues? [12:04:01.0000] "aria-owns"?! hah [12:04:02.0000] /me introduces "html5-is-da-hot-stuff" [12:05:00.0000] Hixie: No. I think you just put the ISSUE-20 in the subj. line [12:05:01.0000] seriously, though, ARIA containment conformance checking would be super-simple without aria-owns [12:05:02.0000] or with aria-owned-by [12:05:03.0000] but having the relation go from parent to child sucks big time for me :-( [12:05:04.0000] jgraham: oh i have to put it in the subject? that's unfortunate [12:06:00.0000] I don't know what happens if you put ISSUE-20 ISSUE-30 etc. all in the same subject. Possibly tracker explodes ;) [12:07:00.0000] (I assume that is the problem) [12:07:01.0000] /me thought they could be in the message body as well [12:07:02.0000] i'm just gonna mention them in the body and link them up later if required [12:08:00.0000] "The keywords can be anywhere in the subject or text of the message. However, the tool only reads the text portions of an email, not HTML or PDF, so make sure you send emails either in plain-text or with an equivalent plain-txt attachment." [12:08:01.0000] i hope it reads all of the e-mail [12:08:02.0000] hsivonen: Where is that? [12:08:03.0000] this is one big mail [12:08:04.0000] http://www.w3.org/2005/06/tracker/email [12:09:00.0000] /me didn't spot that link [12:12:00.0000] also, I think aria-templateid may have bad effects on innovation and compatition [12:13:00.0000] bad for innovation, because it makes changing sites break accessibility [12:13:01.0000] and bad for competition, because some participants get a special boost [12:13:02.0000] and as a competitor, you cannot get the same advantage by writing to an open spec [12:14:00.0000] instead, you have to use a magic string and reverse engineer its effect [12:17:00.0000] you don't have to [12:17:01.0000] you can just hardcode uris [12:17:02.0000] and that way you're not dependent on the author to provide a hook [12:19:00.0000] Hixie: I meant you are a competitor of a site whose templateid is getting preferential treatment [12:20:00.0000] [24~aah [12:20:01.0000] Hixie: how do you gain the same advantage? [12:20:02.0000] ure [12:20:03.0000] sure even [12:20:04.0000] it's anti-competitive! [12:20:05.0000] must be dropped [12:20:06.0000] :-) [12:21:00.0000] the innovation point is that you are yourself getting preferential treatment for you old version. can you change your app innovatively? [12:22:00.0000] yeah [12:28:00.0000] well i just closed a tracker issue with a rejection for the first time [12:28:01.0000] /me dons his asbestos suit [12:28:02.0000] /me just replied to 102 e-mails [12:28:03.0000] that should help the graph [12:32:00.0000] so i don't suppose anyone did go through my input-for-whatwg-html-parsing-rules-namespaces-discussion folder and summarise the problem descriptions, huh [12:36:00.0000] Hixie: Only 67kb. I'm disappointed ;) [12:45:00.0000] Hixie: btw, aria-owns is supposed to address the case where a table has sub-rows but the table markup only allows consecutive rows [12:48:00.0000] Hixie: did you talk markp into removing his summary attributes? [12:49:00.0000] /me is almost sure there were summary='' instances on diveintomark a while ago [14:05:00.0000] hsivonen: no, but i doubt he'd be hard to convince :-) [14:17:00.0000] well that didn't take long... [14:17:01.0000] (re: www-archive/html4all) [14:28:00.0000] wow, I'm surprised I ever said abbr="" should be retained. It's interesting how my opinion changed in the last 2 years [14:32:00.0000] abbr=""? what's that? [14:33:00.0000] when the top hit on google is actually the spec and not some tutorial, you know you're off to a bad start [14:33:01.0000] what a pointless attribute. [14:34:00.0000] Hixie: Does HTML5 have a way to a "provide a brief overview of how a data table is organized or a brief explanation of how to navigate the table", making "the information available to people who use screen readers" without displaying the information visually ( quotes from http://www.w3.org/TR/WCAG20-TECHS/H73.html )? [14:34:01.0000] gsnedders: If you're trying to reduce table verbosity abbr can be useful. [14:35:00.0000] webben_: I don't understand why you would want to discriminate against people who can see. fulfills the rest of the use case. [14:36:00.0000] any whatwgers got ssb brawl? [14:36:01.0000] Hixie: That assumes that the information is useful to people who can see. [14:36:02.0000] If it's useful to people who can see, it shouldn't be in summary. [14:36:03.0000] webben_: why would it not be? [14:36:04.0000] Why would it be? [14:38:00.0000] Hixie: e.g. in the WCAG example, is that information useful to people who can see? [14:38:01.0000] at least assuming "Schedule for Route 7 going downtown" is obvious looking elsewhere on the page [14:39:00.0000] Hixie, what's "ssb brawl"? [14:39:01.0000] sorry, playing brawl right now, can't look at the web :-) [14:39:02.0000] Lachy_: wii game [14:39:03.0000] Lachy_: i'm playing with allan [14:40:00.0000] webben_: I would say that information is useful for sighted users as well [14:40:01.0000] the abbr attribute provides a short form of the headers in a data table [14:41:00.0000] it would be useful in flexible layouts, as the short form could be disabled if the columns got too narrow for the long form [14:41:01.0000] takkaria: So you reckon sighted people would need that explanation above the table (say) in order to understand it? [14:42:00.0000] it's also intended to reduce the pain of long headers being repeated in the aural rendering of data tables [14:43:00.0000] only 1 table in my survey used it [14:44:00.0000] people who don't care enough about users to write short table header text probably don't care enough to write short table header text in a special attribute as well as their long table header text [14:44:01.0000] I used abbr to try and cut down the verbosity of some of the tables in the Yahoo! UK Funds Centre: http://uk.biz.yahoo.com/16/sector.html [14:44:02.0000] e.g. "Quartile Rank over 1 Year" => "Rank over 1 Year" [14:45:00.0000] it's not much, but if it helps a little it's worthwhile [14:47:00.0000] webben_: I don't they need it, no, but it's a useful summary [14:47:01.0000] I don't see it doing any harm to sighted people [14:48:00.0000] the quick summary of the service times would be useful, for example [14:48:01.0000] It turns out that trying to figure what makes sense when integratating HTML5 and ARIA is non-obvious [14:49:00.0000] it would be much easier to leave it unspecified and make it Someone Else's Problem [14:49:01.0000] hsivonen: if you take a baseline assumption of "ARIA always wins over native markup role, on a property-by-property basis", what else needs to be specified? [14:49:02.0000] (honest question - I am not sure) [14:50:00.0000] that's a terrible design [14:51:00.0000] othermaciej: I am assuming that it is a bad idea to let document conformance allow override of strong native semantics [14:51:01.0000] Hixie: that is how the ARIA guys say it should work - it certainly has the advantage of being easy to implement [14:51:02.0000] othermaciej: more importantly, there are some native semantics that could be augmented with states and properties [14:52:00.0000] hsivonen: it does seem to me like a more careful integration design would not let use use anything but an element for a header, or let you make a checkbox appear to AT to be a radio button [14:52:01.0000] othermaciej: it's also a horrific abuse of semantics [14:52:02.0000] othermaciej: and requiring a role in that case would just be silly [14:52:03.0000] othermaciej: for example, I think

should allow aria-sort without having a role [14:53:00.0000] hsivonen: aaron's statement on this (not sure it should be taken as official, since it was just an offhand remark on a list) is that it should work property by property, even if no explicit role is set [14:53:01.0000] and that in fact, on elements that would have "native" state like a checkbox, it should show through even if you set the role to something else [14:53:02.0000] hsivonen, you have an informal list of ARIA stuff which can be inferred for HTML5 elements? maybe create a document listing all the ARIA stuff which can be inferred by default [14:53:03.0000] as an implementor I am pleased, as an architect I find this design distasteful [14:53:04.0000] BenMillard: a list like that would be great [14:54:00.0000] BenMillard: I'm writing the list right now [14:54:01.0000] BenMillard: ARIA guys did not seem to think having one was very important [14:54:02.0000] takkaria: Yes. That's a reasonable argument for removing some of that text from SUMMARY (although there's a risk of swamping the page in explanatory text). Whether designers would be persuaded that enough sighted users would be helped if it were included is a different matter; persuading designers to include some text that doesn't mess with their design is likely to be easier. (Of course, if sighted users are /equally/ helped then there's no argument that it sh [14:54:03.0000] FWIW we are seriously considering implementing ARIA in WebKit soon, the only thing that makes me hesitate is HTML integration issues (not just HTML5 but even just HTML4 stuff) [14:55:00.0000] yay, a useful contribution. with that I'll depart for dinner...cya! [14:56:00.0000] othermaciej: the ARIA spec is written to allow certain states and properties with certain roles [14:56:01.0000] othermaciej: so if Validator.nu only allowed those, it would proclaim as non-conforming [14:56:02.0000] hsivonen: I see [14:57:00.0000] and since I don't want to allow just any random stuff, this requires thinking [14:57:01.0000] hsivonen: so even if the implementation is as I said, perhaps it would make sense to change conformance to take into account implied roles, and to perhaps in some cases make override of native role non-conforming [14:57:02.0000] and the questions that thinking about this turns up are don't all have obvious answers [14:58:00.0000] At least some of the WAI folks were very sympathetic to the argument that ARIA should not supersede or discourage proper semantic markup where that would work for the intended use [14:58:01.0000] othermaciej: right. and someone has to figure that out [14:58:02.0000] othermaciej: their best practice doc doesn't follow that policy [15:00:00.0000] hsivonen: I wouldn't say all agree with it, and I don't think any of them has tried to fully think through the implications of such a policy [15:20:00.0000] /me is puzzled with the assertion that people think 0 is positive [15:21:00.0000] wtf is aria-sort=ascending and why doesn't it just apply generally? [15:22:00.0000] as in, why is it an "aria" thing [15:22:01.0000] Hixie: it makes a claim to the AT that a table column is sorted in an ascending order [15:22:02.0000] it doesn't actually sort it [15:22:03.0000] I'd assume because it's easy to tell at a glance whether something is sorted alphabetically if you can see it. [15:22:04.0000] hsivonen: ok so why don't we just have a "sort-order" attribute? [15:23:00.0000] why don't you? [15:23:01.0000] Hixie: because the whole ARIA thing assumes that whoever defines HTML is unresponsive and they have to do their thing in their own syntactic partition [15:23:02.0000] that's not going to fly well if they want it integrated with html5 [15:23:03.0000] Hixie: also, ARIA assumes that the host language space is wider than HTML+SVG [15:24:00.0000] ah, overgeneralisation syndrome [15:24:01.0000] typical of w3c [15:24:02.0000] yeah [15:24:03.0000] in an ideal world, the language space would be clamped to HTML and SVG and the HTML WG and the SVG WG would be responsive [15:25:00.0000] i'm still baffled by this concept of wanting to annotate parts of svg with aria roles [15:25:01.0000] well, people want to build controls out of SVG [15:25:02.0000] so it's much the same as building controls out of divs and annotating those, I should think [15:25:03.0000] controls is one thing [15:26:00.0000] marking things as tables with a sort order is something else [15:26:01.0000] If you're going to build controls out of SVG, why not tables too? [15:26:02.0000] because that's what is for? [15:27:00.0000] Hixie: also, one of the tacit assumptions of ARIA is that the a significant proportion of browsers is uncooperative [15:27:01.0000] seems to me that the underlying assumptions of the aria work are out of date [15:27:02.0000] which doesn't hold for ARIA anymore but still holds for XBL2 [15:28:00.0000] how so? [15:28:01.0000] Hixie: Dunno. It's like people building text in Flash I'd suppose. [15:29:00.0000] Hixie: there's no indication of Microsoft wanting to implement XBL2, as far as I can tell [15:29:01.0000] Hixie: or Canvas [15:29:02.0000] webben_: yes, it's exactly like that. and people building text out of flash (or tables out of svg) have already given up on accessibility, so why is providing aria attributes going to help? [15:29:03.0000] it's like people who want new features in spec to work around bugs in browsers [15:30:00.0000] hsivonen: microsoft never indicate anything. [15:30:01.0000] hsivonen: (of this kind of stuff) [15:30:02.0000] Hixie: I'm not sure they have given up on accessibility. I think they're more likely to be privileging something else over accessibility. [15:30:03.0000] building tables out of SVG seems like a weird concept [15:30:04.0000] Hixie: right, so ARIA takes a strategy that works regardless of what MS does [15:30:05.0000] Hixie: but then in this case it turns out that they implement it as well [15:30:06.0000] on the other hand, at least some SVG-related thinking is based on the premise that your whole UI is SVG [15:30:07.0000] Hixie: but that wasn't necessary for ARIA to be useful elsewhere [15:31:00.0000] hsivonen: i don't see how any of the dumb things in aria have anything to do with whether ms implements aria or not [15:31:01.0000] othermaciej: exactly [15:31:02.0000] I would rather stab myself in the eye with a fork than implement table layout using SVG [15:32:00.0000] Hixie: the general idea of putting AT integration away in attributes instead of requiring new elements like and works without Microsoft's participation [15:32:01.0000] othermaciej: already done I think: http://thomas.tanreisoftware.com/?p=42 [15:32:02.0000] Hixie: the specifics of ARIA are specifics [15:32:03.0000] (the implementing, not the eye-fork-stabbing ;) ) [15:32:04.0000] Hixie: there are a number of specifics I quite dislike [15:32:05.0000] hsivonen: dude. isn't new. [15:32:06.0000] webben_: I read the phrase "SVG flow layout engine" and closed the window [15:32:07.0000] hehe [15:32:08.0000] Hixie: that's one part I seriously dislike [15:33:00.0000] Hixie: to me, it makes no sense to have that there [15:33:01.0000] same with the header stuff [15:34:00.0000] certainly some roles seem redundant in HTML [15:34:01.0000] HTML4, not just 5 [15:34:02.0000] I can see why, if you are a crazy AJAX guy, you might want to pretend a
is a checkbox [15:34:03.0000] but not why you would want to pretend it is a table or header [15:34:04.0000] Hixie: I already complained about the header "best practice" stuff [15:34:05.0000] you can already customize the presentation of the real elements for those as much as you want [15:34:06.0000] Hixie: tomorrow, I'm going to complain about the table stuff [15:35:00.0000] othermaciej: not really [15:35:01.0000] Hixie: and about templateid [15:35:02.0000] not in current browsers anyway [15:35:03.0000] there are more bugs with styling table elements than divs [15:35:04.0000] webben_: h1 can be customized just as much as div [15:35:05.0000] othermaciej: h1 probably can yes [15:35:06.0000] table does imply row/cell structure [15:35:07.0000] Hixie: I also want to complain about aria-owns, but I predict it is too late to fix that one in this browser release cycle [15:36:00.0000] aria-owns scares me [15:36:01.0000] but if you have something that does not follow that row/cell structure, how is it a table? [15:36:02.0000] what is aria-owns? [15:36:03.0000] othermaciej: aria-owns overrides the parent-child relationship induced by tree containment [15:36:04.0000] does it replace it or add to it? [15:36:05.0000] othermaciej: the scary part is that since it overrides what children a parent has, it can make cycles [15:37:00.0000] letting the child point to a single new parent would be safer [15:37:01.0000] othermaciej: I think it is supposed to replace the normal containment when computing what to tell AT [15:37:02.0000] that's kind of a pain [15:38:00.0000] my work would be much simpler without aria-owns [15:38:01.0000] since to report children to AT, you have to know about all aria-owns attributes everywhere in the document [15:39:00.0000] besides, I think the problem aria-owns is supposed to solve could be solved in a less disruptive way [15:39:01.0000] the stated problem is that table has a flat list of rows [15:39:02.0000] and aria-owns makes it possible to make some rows children of other rows [15:40:00.0000] I think aria-level could be tweaked to solve that [15:40:01.0000] so you can fake an outline view with a table? [15:40:02.0000] right [15:40:03.0000] like [16:22:00.0000] the links in http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2006-June/006498.html suggest that maybe presentational mathml is not an accessibility problem after all [16:30:00.0000] Hixie: also http://firevox.clcworld.net/features.html [16:34:00.0000] I'd guess http://www.nfbnet.org/mailman/listinfo/blindmath would be good folks to ask though 2008-03-24 [21:37:00.0000] Hixie, i'd use sequence rather than sequence
with no attributes or or for table headers [09:37:00.0000] in general, inline-level markup can be anything you want and it will adapt well enough [09:38:00.0000] /me tries to work out how to create a "succint" alt attribute for a complex chart or diagram. [09:39:00.0000] providing sensible alt for graphical links and buttons is something you can't really get around...unless we require such images to have readable filenames :P [09:41:00.0000] hmm, over 600 e-mails [09:43:00.0000] /me has been sending email mostly to http://lists.w3.org/Archives/Public/public-pfwg-comments/2008JanMar/ instead of public-html... [09:43:01.0000] 80 marked as spam [09:44:00.0000] http://lists.w3.org/Archives/Public/public-pfwg-comments/2008JanMar/author.html#msg54 - hsivonen's messages [09:46:00.0000] ah, my e-mail was not just since yesterday though :) [10:12:00.0000] hsivonen, http://hsivonen.iki.fi/aria-html5/ times out for me (tried 3 times) [10:19:00.0000] works now [10:19:01.0000] yep [10:29:00.0000] hsivonen:
? Fun fun fun. [10:30:00.0000] "In many cases the user agent may be able to calculate the level of an item if it can be determined correctly from the document structure. This property may be used to provide an explicit indication of the level if that is not possible from the document structure. User agent automatic support for automatic calculation of level may vary; it is recommended that authors test with user agents and assistive technologies to determine whether this property [10:30:01.0000] is needed." [10:31:00.0000] What kind of spec text is that? [10:32:00.0000] "You might want to consider using this property in some undefined set of circumstances when some undefined other mechanisms for achieving the same thing fail in some undefined set of software you have access to" [10:35:00.0000] Possible steps to fixing this: [10:35:01.0000] perhaps: "The aria-level property must not be used on elements which are specified to indicate a level natively." [10:35:02.0000] a) identify the set of circimstances in which aria-level is needed [10:36:00.0000] b) try to reduce that set to 0 by adding features to HTML [10:36:01.0000] dunno if that will work though [10:36:02.0000] followed by: "If it is used on such elements, it is ignored." [10:37:00.0000] jgraham: Note that if support for those new features trails support for aria-level, web developers may be forced to use both to support older UAs. [10:37:01.0000] webben: Sure [10:39:00.0000] ARIA is itself a new feature which older UAs don't support, though? [10:39:01.0000] (for some definition of "forced") [10:39:02.0000] BenMillard: Well it looks like ARIA will be supported faster than e.g. [10:39:03.0000] BenMillard: older UAs being the current crop (IE8, Fx 2-3, Opera 9.5) [10:40:00.0000] + the current crop of AT of course [10:41:00.0000] webben: The nice thing is that browser upgrade cycles are much faster than AT upgrade cycles. [10:41:01.0000] And exposing e.g. is a browser problem [10:42:00.0000] jgraham: That doesn't help if the browser upgrades require AT upgrades (as IE7 did, for example) [10:42:01.0000] That sucks. I assumed they all used these standard APIs. [10:43:00.0000] if all that changed between releases was the browser recognized more native HTML5 elements and treated them the same as the aria attributes when exposing to accessibility frameworks, all would be well. But that's not all that changes. [10:43:01.0000] jgraham: Well part of the underlying problem is MSAA is old and crusty; and UI Automation isn't widely used. [10:44:00.0000] jgraham: And with Apple, you need to upgrade your entire OS to get the full benefits of accessibility improvements in WebKit. [10:44:01.0000] In any case, what BenMillard said is a more viable short-term solution (i.e. change the wording to forbid using aria-level where it isn't appropriate and make native semantics take precedence) [10:44:02.0000] the most stable system is perhaps GNOME AT-SPI, oddly enough. [10:44:03.0000] (Firefox3+Orca basically) [10:45:00.0000] jgraham: But how do you define "where it isn't appropriate" in a way that doesn't exclude using aria attributes for backwards compatibility? [10:45:01.0000] jgraham, yay a useful contribution. :) [10:45:02.0000] webben: The most stable system is a dead one :-) [10:45:03.0000] the first part of that (conformance criteria) should be easy. The second part (browser behavior) goes against current thinking [10:46:00.0000] Philip`: well, Apple would be much the same if the upgrades were free. [10:46:01.0000] but the unfreeness creates a longer upgrade cycle [10:47:00.0000] jgraham: I can't see the point of forbidding web developers from doing stuff that doesn't break newer browsers and makes old browsers work. It's a dead letter conformance requirement. [10:47:01.0000] webben: What sort of back-compat scenario requires html-headings (for example) to be labelled with explicit aria-levels [10:48:00.0000] jgraham: Well, the spec could list a specific set of /currently/ widely supported elements not to use it on. [10:48:01.0000] that's a bit messy though [10:49:00.0000] jgraham: Oh actually. There might be one. [10:49:01.0000] webben: The problem is that you want a consistent view of the document regardless of whether you view it through AT or through a normal browser [10:49:02.0000] jgraham: HTML5 changes how headings should be ordered (thanks to section). [10:49:03.0000] Browsers and AT don't support that yet. [10:49:04.0000] AFAIK. [10:49:05.0000] we should allow text/html with root-
and
. the exact number is less important [10:50:01.0000] (inclusive) [10:50:02.0000] BenMillard: It's /less/ important. It's not irrelevant. [10:51:00.0000] But don't take my word for it: http://video.yahoo.com/video/play?vid=514676 (on importance of heading levels in understanding structure of page) [10:51:01.0000] I saw that months ago, thanks :) [10:52:00.0000] considering how rare it is for content to use *any* heading element, we should not forget what's important [10:52:01.0000] I see no reason as a web developer to make content less accessible in order to satisfy arbitrary conformance criteria. [10:52:02.0000] me neither. do you have any examples where that is happening? [10:53:00.0000] BenMillard: I just posited one. [10:53:01.0000] webben: The conformance criteria aren't arbitrary. It's needed for logical consistency. [10:54:00.0000] /me doesn't really see how. [10:54:01.0000] I'm more worried about the logical inconsistency between heading levels in different versions of HTML, myself. [10:55:00.0000] Sounds like aria-level would provide a way to work around that. [10:55:01.0000] I suppose theoretically you could make the criteria "aria-level used on heading elements must produce an outline structure that is identical to that produced by the html5 algorithm" [10:55:02.0000] but hsivonen might kill you ;) [10:56:00.0000] I think we're in different countries, so I'm probably safe for the moment. ;) [10:56:01.0000] webben: There is no logical inconsistency between HTML versions [10:56:02.0000] HTML4 had no structure [10:56:03.0000] except one that you inferred the authors probably meant [10:56:04.0000] It didn't have "no structure". [10:57:00.0000] Certainly not once you take WCAG and actual implementations into account. [10:57:01.0000] webben, if you want backward-compatible headings (as do I) you just choose the correct number for each heading (whatever your feeling of "correct number" is). there is no need for aria-level in this respect, afaict. [10:57:02.0000] webben: There is no text anywhere in HTML 4 that says what heading structure
, [10:57:03.0000] erm.. [10:57:04.0000]
corresponds to, let alone
[10:57:05.0000] s/~// [10:58:00.0000] BenMillard: Yeah, I'd believe that more if HTML5 authorial conformance criteria required Hx headings to be used that way. [10:59:00.0000] jgraham: "There are six levels of headings in HTML with H1 as the most important and H6 as the least." (http://www.w3.org/TR/html401/struct/global.html#edef-H1) [10:59:01.0000] define "important" [10:59:02.0000] many web authors think headings in side bars are "less important" [10:59:03.0000] define "structure" [10:59:04.0000] eh? [10:59:05.0000] Headings in side bars may well be less important. [11:00:00.0000] And to that extent, putting them at a lower level is fine. [11:00:01.0000] I guess I should have said "outline structure" [11:01:00.0000] BenMillard: "considering how rare it is for content to use *any* heading element" - I see
on over a fifth of pages, so it doesn't seem that rare [11:01:01.0000] Does HTML5 say that h1s in different positions of the "outline structure" have the same "importance"? Are the notions of "importance" and "outline structure" incompatible anyways? [11:01:02.0000] Which I define as a tree-like view of the document in which only the text of headings is visible and the headings of subsections are children of their parent headings (and probably some more stuff to deal with siblings) [11:02:00.0000] Philip`, that means it isn't used on somewhere nera 80% of pages, then. :) [11:02:01.0000] webben: The problem with having "less important" headings in sidebars is that it breaks this outline view [11:02:02.0000] So putting them at a lower level isn't fine [11:02:03.0000] jgraham: That's debatable. [11:03:00.0000] It's not ideal, sure. [11:03:01.0000] BenMillard: Compared to most other HTML features, that's hugely popular :-) [11:03:02.0000] But I don't think it breaks the view. [11:03:03.0000] webben: That's not what I found [11:04:00.0000] what was your test for breakage? [11:04:01.0000] Philip`, consider it's popularity from a user's perspective: 80% of the time they can't get around the page easily [11:04:02.0000] fixing that 80% so they use heading elements *at all* is the priority, as I see it [11:05:00.0000] BenMillard: That doesn't follow necessarily. Maybe they just stick with webpages that are coded better. [11:05:01.0000] hsivonen: ping [11:05:02.0000] webben, the research I want to do into use of headings on the web would throw light on this. haven't quite secured sponsorship for it, yet [11:06:00.0000] despite taking this whole month off work, I might add :( [11:06:01.0000] BenMillard: Well, good luck getting sponsorship from whomever you're trying to get it from :) [11:08:00.0000] webben: My test for breakage was whether it made sense in my Firefox outliner extension [11:08:01.0000] jgraham: I think I more meant your test for "making sense" then ;) [11:08:02.0000] BenMillard: Hmm, how much harder does a lack of headings make it to use a page? (I'm wondering if that's more/less of a problem than ~50% of images not using alt) [11:08:03.0000] Philip`: That very much depends on the complexity of the page doesn't it? [11:09:00.0000] (and what /else/ is on the page to navigate by) [11:09:01.0000] (oh and what else the user knows (s)he can navigate by) [11:09:02.0000] webben: Well the harshest test is whether there was ever a "subsection" in the outline that clearly wasn't a logical child of the parent [11:09:03.0000] Philip` and webben, absence of headings is similarly problematic to absence of sensible alt [11:09:04.0000] webben: That sounds sensible; I've got no idea how that works out in practice, though [11:10:00.0000] That often happened when the sidebar came out as a child of the last "content" section [11:11:00.0000] jgraham: Was the sidebar only a sidebar to the last content section? [11:11:01.0000] If not, maybe the choice of level didn't really reflect its importance [11:11:02.0000] No, to the whole page [11:11:03.0000] Also, things around the page header often broke [11:12:00.0000] with subtitles often causing confusion [11:12:01.0000] http://projectcerbera.com/web/articles/heading-numbers - this article I wrote was well received by website developers on Accessify Forum. [11:12:02.0000] Basically, authors interpret the vague specs differently. So I think a strict, tree-based outline algorithm is doomed to failure for web content, sadly. [11:12:03.0000] Philip`: Well, in the simplest case, if you only have one heading on a gallery page, alt is going to be more crucial than
[11:13:00.0000] Yeah, I don't think strict, tree-based algorithms need to be 100% successful for the lousy semantics established by HTML4 and common practice to be useful [11:14:00.0000] I think we should let authors use heading numbers however they want, because at least they're actually using heading elements that way [11:15:00.0000] If using heading numbers in some sort of meaningful logic helps, I don't see any reason to do that. [11:15:01.0000] in practice, authors are going to make mistakes anyways [11:15:02.0000] encouraging them to be consistent throughout the website is wortwhile, as WCAG 1.0 requires [11:16:00.0000] In principle HTML 5 solves all these problems with
and
[11:16:01.0000] you trust authors to get that right? I wouldn't :) [11:17:00.0000] BenMillard: No, but at least the results when they get it wrong will be well defined [11:18:00.0000] So it will be wrong in the same way for everyone - easier for people to complain, easier for a conformance checker to provide an outline view [11:18:01.0000] (or an accessibility checker, for that matter) [11:18:02.0000] BenMillard: If Joe Bloggs gets it wrong on a minor website, it's less crucial than Google, Yahoo!, MSN, the New York Times, the BBC etc getting it right on theirs. [11:19:00.0000] because people spend more time on big websites than small ones [11:20:00.0000] more people use one big website than use one small website. the time each person spends on each website is random [11:20:01.0000] I hardly spend any time on big websites, for example [11:20:02.0000] BenMillard: I don't think that's typical. [11:20:03.0000] Don't know of any market research to back that up off-hand though. [11:21:00.0000] webben, are you saying I'm weird? [11:21:01.0000] actually don't answer that... [11:21:02.0000] BenMillard: No! :) [11:21:03.0000] just a different "demographic" ;) [11:22:00.0000] somebody must have done some actual research on this though [11:23:00.0000] the goal is to make *every* website accessible as soon as possible? [11:23:01.0000] http://blog.compete.com/2007/01/25/top-20-websites-ranked-by-time-spent/ [11:24:00.0000] /me postulates that facebook has risen somewhat [11:24:01.0000] jgraham: Hmm... is that collective time. [11:24:02.0000] Because that's slightly different. [11:25:00.0000] webben: It's hard to see how to get meaningful data without averaging large groups [11:25:01.0000] "Read as: Of all time spent online by all US Internet users" [11:25:02.0000] jgraham: sure, but that's still measuring something very different. [11:25:03.0000] Unless you start trying to subdivide the data based on real/imagined patterns [11:25:04.0000] what one would want to do is establish which sites are big, then average how much time is actually spent on those by each person. [11:26:00.0000] so let's say myspace is big [11:26:01.0000] but how much time does the average person spend on Myspace (as opposed to everywhere else) [11:26:02.0000] How do you define "big" in an independent way? [11:26:03.0000] hang on, we are trying to make *everything* accessible. not just the big, time-consuming sites? [11:26:04.0000] jgraham: It can be arbitrarily defined, e.g. top 100 websites by traffic [11:26:05.0000] jgraham: might get better if you subgrouped more [11:27:00.0000] e.g. maybe everyone spend's most of their time on like 10 websites. [11:27:01.0000] BenMillard: Yes. But that doesn't mean the accessibility of every site is equally important. [11:28:00.0000] So one definition which would fit handily with the data that I happened to turn up is that we define "bigness" by time spent and define "big" sites to be those which are in the top 20 by bigness. If we do that we find that 39% of all time is spent on big sites [11:28:01.0000] ;) [11:28:02.0000] BenMillard: So one wouldn't needn't get too depressed by figures suggesting inaccessibility is widespread in absolute terms. [11:29:00.0000] BenMillard: But would need to get very depressed by figures showing inaccessibility is common on highly popular sites (which of course it is). [11:29:01.0000] in practice, site owners decide if accessibility is important enough to them to pay for it [11:30:00.0000] Well sure, but only if the developers charge them extra for it. [11:30:01.0000] Rather than just doing it right. [11:30:02.0000] FWIW, sdesign1, the company I work for, doesn't charge extra for accessibility [11:30:03.0000] webben: Arguably technologies like aria encourage that because doing it right is considerably more effort than just doing it [11:31:00.0000] prioritising the important things, like using heading elements at all, is what site owners need to know about [11:31:01.0000] (There are edge cases where that's not the case, where site owners need to stump up for accessible content though) [11:31:02.0000] jgraham: ARIA was mainly designed to build accessibility into div-based widgets systems. [11:31:03.0000] Once built into those systems (like Dojo), the cost of using an ARIA-fied widget is the same as using a non-ARIA-fied widget. [11:32:00.0000] webben: I know. Your argument works iff everyone is using off-the-shelf widgets [11:32:01.0000] people who actually build widgets are already a very select grouo [11:32:02.0000] *group [11:33:00.0000] I've used widgets. they are built terribly, on the whole [11:33:01.0000] Does e.g. the new bbc design use entirely off the shelf widgets? [11:33:02.0000] I wouldn't /expect/ the BBC design to use entirely off-the shelf widgets. [11:33:03.0000] I'd expect Joe Bloggs building his intranet app in a hurry to use them etc [11:35:00.0000] I'd expect the BBC to be the sort of place which would hire web developers who could build custom widget sets. [11:35:01.0000] I'm sure the bbc, in particular, is since they have a very strong commitment to such things [11:36:00.0000] expectations are irrelevant. resistance is futile. you will be laid out in a [11:36:01.0000] if HTML5 makes accessibility seem hard, that will just put people off [11:36:02.0000] we need basic accessibility to be dead simple [11:36:03.0000] refined accessibility can be harder, if necessary [11:36:04.0000] There's no point in making accessibility seem simpler than it is. [11:37:00.0000] I do see your point about quick wins. [11:37:01.0000] But I don't think the place to encourage quick wins is in the conformance criteria. [11:37:02.0000] Most people don't care about conformance criteria. [11:38:00.0000] /me suggests there is a high correlation between caring about conformance criteria and caring about accessibility [11:38:01.0000] sure. [11:38:02.0000] caring about either is rare [11:38:03.0000] Not that I have evidence to back that up or anthing [11:38:04.0000] but people who really care about accessibility aren't going to be put off because it needs thinking [11:38:05.0000] we need accessibility to be more accessible :) [11:38:06.0000] Hmm maybe. [11:39:00.0000] webben: Not true. People might want to be accessible but fail because it's too hard [11:39:01.0000] The biggest obstacles to making stuff accessible tend to be further down the chain though. [11:39:02.0000] e.g. "What? Blind people can use computers? But how?" etc [11:39:03.0000] which way is down? [11:39:04.0000] s/down the chain/up the chain, imho [11:39:05.0000] i.e. managers rather than developers [11:40:00.0000] jgraham: Down is managers/commissioners of websites. Down is a million miles away from the spec and conformance criteria and the HTML. [11:40:01.0000] and up is users? [11:40:02.0000] no up is implementors, developers, testers [11:40:03.0000] and the spec and the code [11:41:00.0000] where are users in this chain [11:41:01.0000] Where is AT? [11:41:02.0000] they're not. [11:41:03.0000] Browsers? [11:41:04.0000] (users aren't anyhow) [11:41:05.0000] /me envisioned a chain connecting content to users [11:41:06.0000] I'd say for AT and browsers there's a similar chain. [11:41:07.0000] ah, I think webben is talking about the process of building websites, rather than the process of using websites? [11:41:08.0000] and that these are 2 different chains? [11:41:09.0000] Sorry, yeah, I thought we all were. [11:42:00.0000] my approach to building websites involves users, which is what threw me [11:42:01.0000] /me suggests a chain is too 1D [11:42:02.0000] jgraham: If you're talking about, the end-product (users being able to access content/functionality) then yes. [11:43:00.0000] users/disability organizations/governments/AT/browsers/operating systems - all sorts of factors [11:44:00.0000] Also, down is in people's level of interest. A lot of sitepoint comments on the target case were from apparent webdevs surprised/outraged that blind people can use the internet. [11:45:00.0000] *down is people's level of knowledge rather. [11:45:01.0000] remembering that a lot of people developing websites aren't very interested in the front-end [11:47:00.0000] making accessibility seem like a grand and difficult struggle puts devs off, in my experience (I am often contracted by other companies) [11:48:00.0000] showing a code sample where is replaced by
has a more positive reaction [11:48:01.0000] I don't think accessibility has to be made grand and difficult. [11:48:02.0000] Actually, I'm not sure accessibility at the HTML level is /that/ difficult. [11:49:00.0000] it gets a lot trickier with scripting and CSS. [11:49:01.0000] and there's only a limited amount HTML5 can do about either. [11:50:00.0000] We need CSS5 and ECMAScript5. [11:50:01.0000] gsnedders: Well, a lot of accessibility problems produced through CSS derive from implementations. [11:51:00.0000] /me closes implied sarcasm element [11:52:00.0000] Hixie's e-mail "several messages about tables and related subjects" says: "I also looked in detail at the information that Ben and James provided, which was incredibly useful." yay us! [11:52:01.0000] :) [11:52:02.0000] scripting ... well maybe if HTML5 comes up with a nice solution for exposing shortcuts to functionality or something. [12:07:00.0000] hsivonen, http://hsivonen.iki.fi/rdf/ s/is assumes/assumes/ [12:18:00.0000] http://www.openplans.org/projects/opencore/blog/2008/03/24/accessibility-on-our-site/ does he just have something wrong with his NVDA installation? [12:19:00.0000] s/just/"just"/ [12:30:00.0000] Hixie? [12:39:00.0000] jgraham: he helpfully doesn't say what version he's using [12:40:00.0000] webben: I would presume he just got 0.5 [12:45:00.0000] a-ja: pong [12:45:01.0000] hi henri [12:46:00.0000] hsivonen: re: your "ARIA in HTML5" strawman "Strong Native Role—No Override" section [12:46:01.0000] thing like article only allowing region, e.g. [12:47:00.0000] should those also be allowed the "landmark" roles? [12:47:01.0000] like main or secondary? [12:47:02.0000] andersca: here [12:47:03.0000] BenMillard: i don't believe i've ever looked at so many tables before :-) [12:48:00.0000] BenMillard: i ended up covering two whiteboards with little table diagrams [12:48:01.0000] Hixie: correct me if I'm wrong, but [12:48:02.0000] Hixie: manifest lines can't have leading whitespace, right? [12:49:00.0000] hixie, I can well imagine! [12:49:01.0000] a-ja: I don't know. [12:49:02.0000] possibly secondary for aside, e.g. vs just note possible? [12:49:03.0000] a-ja: could make sense. I'm not sure. [12:49:04.0000] hixie, I think I looked at thousands overall because I wanted to see how consistent each site was [12:50:00.0000] Hixie: brb [12:50:01.0000] a-ja: in any case, I think eventually a good outcome is that if the landmarks are really important, they have native HTML5 elements [12:50:02.0000] BenMillard: thanks for that, it made my life much easier :-) [12:50:03.0000] a-ja: those would be good questions to raise on the public-html and public-pfwg-comments lists [12:51:00.0000] hixie, I'm happy to be useful. not bad for a noob :P [12:51:01.0000] a-ja: the problem is that the landmarks and native HTML5 elements don't have a one-to-one mapping [12:51:02.0000] a-ja: I gather landmarks are more controversial than ARIA widgets. I wouldn't be surprised if landmarks changed still. [12:52:00.0000] yeah,,,know what you mean [12:52:01.0000] BenMillard: :-) [12:53:00.0000] hsivonen: was it decided one role allowed or multiple? been a while [12:54:00.0000] a-ja: (I can't associate your IRC nick with an email name) [12:54:01.0000] andersca: 4.6.3.2. Parsing cache manifests [12:54:02.0000] a-ja: It's only one role per element in ARIA 1.0 [12:54:03.0000] andersca: steps 13 and 15 [12:54:04.0000] a-ja: multiple values are for later extensions [12:54:05.0000] ajabanyon at gmail [12:54:06.0000] ah [12:55:00.0000] Is there a reasonable mailing list to subscribe to to keep track of aria? [12:55:01.0000] Hixie: ah [12:55:02.0000] hixie, what specifically did you find useful about it? was it just having links to lots of genuine data tables in one place? [12:55:03.0000] jgraham: public-pfwg-comments [12:55:04.0000] jgraham: the main list in Member-only [12:56:00.0000] hsivonen: Can you actually subscribe to that? Maybe I've ben misled by the archve page which only had unsubscribe links [12:56:01.0000] s/b en/been/ [12:56:02.0000] jgraham: I don't know. [12:57:00.0000] jgraham: I'm not subscribed to any ARIA lists. [12:57:01.0000] jgraham: I go with CCs and the Web archive interface [12:57:02.0000] The problem with the web archive interface is that it's hard to guess when there will be interesting content [12:57:03.0000] I think you can subscribe [12:58:00.0000] Hixie: thanks [12:59:00.0000] /me tries subscribing [13:00:00.0000] hsivonen: don't think header should necessarily always have banner role....since aisdes/sections/articles can all have header [13:01:00.0000] s/aisdes/asides/ [13:01:01.0000] Well that appears o have worked [13:02:00.0000] BenMillard: that was a big part of it [13:02:01.0000] BenMillard: but also your analyses were useful to clue me in to what was important and useful [13:02:02.0000] a-ja: what use case does role='header' fulfill that HTML native h1...h6 don't? [13:03:00.0000] BenMillard: also your stats were very useful as a way to quantify the qualitative experience of the many tables and the anecdotal impression of the distribution of the types of tables [13:03:01.0000] jgraham, sure? [13:03:02.0000] doesn't look like it [13:04:00.0000] well....as container for h1 title, h2 subtitle, p slogan sorta thing [13:04:01.0000] annevk: No, I'm not sure but I got a "reply to this to confirm" type message [13:04:02.0000] hixie, that's good to know. if/when I do more research, I'll concentrate the resulting document(s) on these useful aspects [13:05:00.0000] BenMillard: awesome [13:07:00.0000] Hixie: haha, I had implemented the parser according to your steps and it already handled leading whitespace :) [13:08:00.0000] jgraham, ah [13:08:01.0000] andersca: nice [13:09:00.0000] Hixie: can I give you an acid3 bug report? or should I send mail? [13:09:01.0000] whats the bug report? [13:11:00.0000] Hixie: test 79, two of the advances are wrong, it expects there to be an advance before the first code point of a surrogate pair and the second [13:11:01.0000] dude, get with the program [13:11:02.0000] Hixie: the advance should be after the second [13:11:03.0000] i fixed that ages ago [13:11:04.0000] :-P [13:11:05.0000] oh [13:11:06.0000] I wasn't sure if Cc-ing you would be enough [13:11:07.0000] (i commented on the bug and everything) [13:11:08.0000] thanks [13:12:00.0000] actually so long as the bug says "acid" somewhere in the bugmail, it ends up in my inbox [13:12:01.0000] that test revealed 6 webkit bugs (well, missing features in some cases) [13:12:02.0000] bugmail that doesn't say "acid" anywhere gets read much later [13:12:03.0000] Hixie: thanks for the fix [13:12:04.0000] np [13:13:00.0000] thank _you_ for finding it! [13:13:01.0000] that test case is pretty evil [13:13:02.0000] I mean, not to someone who actually knows anything about SVG or text layout [13:13:03.0000] but I know very little about either [13:14:00.0000] heh [13:14:01.0000] i didn't write it [13:14:02.0000] though i did introduce the aforementioned error [13:14:03.0000] Yeah, I gather heycam did originally, I wrongly blamed him for the error [13:17:00.0000] /me decides to use a wiki page to note down the problems raised in the namespace folder [13:19:00.0000] Hixie: Some super-secret Hixie wiki or...? [13:19:01.0000] whatwg.org [13:19:02.0000] a-ja: that use case works with HTML5
but does it work with role=header? [13:21:00.0000] BenMillard: apparently my server had a maintenance break earlier [13:24:00.0000] hsivonen: i'd think yes for header, no for heading....OIW not all headers are banners, but all banners are headers [13:24:01.0000] annevk: typo noted. thanks. (I'll fix when connectivity to the server comes back) [13:27:00.0000] http://wiki.whatwg.org/wiki/New_Vocabularies [13:27:01.0000] a-ja: yeah, the implicit role mappings aren't perfect matches [13:28:00.0000] hsivonen: i'm no ARIA experts by any means though...just a site dev trying to make sense of landmarking mapping to html5 elements [13:29:00.0000] Actually, I don't like the introduction of the landmark stuff when implementing the HTML5 new landmark containers would be relatively low-hanging fruit [13:29:01.0000] Hixie: I think typography comes before maintainability when it comes to math markup [13:29:02.0000] Hixie: maintainability is abstracted behing itex2mml [13:30:00.0000] behind [13:30:01.0000] it wasn't an ordered list [13:30:02.0000] :-) [13:31:00.0000] and to be honest, i wouldn't necessarily agree that quality comes first [13:31:01.0000] one of the main pieces of feedback i got from the video discussion was that in reality, contrary to common belief, the priority was sustainable frame rate, not video quality [13:31:02.0000] /me wave to BenMillard [13:31:03.0000] i imagine the same applies here [13:31:04.0000] I think ease-of-creation is paramount and that ease of creation means LaTeX syntax [13:32:00.0000] hsivonen: e.g. been taking hyml4 sites with landmarking incorporated in link rels and trying to shoehorn them to html5 elements with roles instead [13:32:01.0000] (because that's what authors already know) [13:32:02.0000] jgraham: i am continuously amazed that nobody seems to be able to say anything in these discussions without linking the problem with the solution :-) [13:32:03.0000] Hixie: I know. [13:32:04.0000] :-p [13:33:00.0000] OK, let me rephrase. The problem is that the primary body of content authors has no interest in learning a new authoring format to create content. [13:33:01.0000] a-ja: do you mean using an HTML5 element *and* a role attribute on it? [13:34:00.0000] I don't see how to fully explain that problem without reference to the extant widely known formats [13:34:01.0000] Hixie: I'm linking the solution to it, because I think one of the problems is minimizing the implementation delta from here to there in certain Web engines [13:35:00.0000] hsivonen: yeah, like role=main on article [13:35:01.0000] instead of link rel=main pointing to a div [13:35:02.0000] hsivonen, jgraham: wiki page updated [13:35:03.0000] In particular ease of authoring doesn't cover the problem I referred to because a new easy format would still not leverage the bias toward the old format [13:35:04.0000] a-ja: OK. my sense of language design aesthetics revolts at that, but practically speaking we should probably allow explicit declaration of the implicit role [13:36:00.0000] Hixie: I suppose the parenthetical covers it [13:36:01.0000] jgraham: if a new format is easier to author than the known format, including cost of switching, then it'll most likely win [13:37:00.0000] Hixie: Tell that to all the scientists who insist on using Fortran for everything [13:37:01.0000] jgraham: the cost of switching is probably high for them [13:39:00.0000] Indeed. But given the large timescales involved (decades) and the possibility of integrating old and new solutions (by calling old fortran from new whatever) there has to be an element of unwillingness [13:39:01.0000] beyond the purely rational [13:40:00.0000] hsivonen: maybe an outer article needs to automagically have main role, and inner articles secondary, or somesuch...less generic than region [13:41:00.0000] biab [13:41:01.0000] hsivonen, you say "[...] we should probably allow explicit declaration of the implicit role" but that is likely to make authors think they must always supply it just in case their markup would be too vague if they omit it [13:42:00.0000] BenMillard: that's why I don't like this landmark stuff at all [13:44:00.0000] /me waves back to davidb, somewhat belated! [13:45:00.0000] I have an urgent appointment with a cottage pie. bye all! [13:46:00.0000] /me blogs on the whatwg blog [13:47:00.0000] Opera on S60 is already pretty good at guessing what the "main" content is without role=main [13:47:01.0000] Hixie: typo: tablular [13:48:00.0000] fixed thanks [13:50:00.0000] hsivonen: just curious...screenie of the S60 UI for that, perchance? [13:51:00.0000] a-ja: I don't know how to take screenshots on S60. the UI is a one-pixel horizontal line in the scrollbar and a keybinding that jumps to that point, to the end and to the start [13:52:00.0000] sounds nice & simple [13:57:00.0000] well, gotta go, too. hope your draft gets feedback from the aria gurus, so discussion to nail some/most of it down can occur in reasonable timeframe [13:57:01.0000] ttfn [13:57:02.0000] see you [14:02:00.0000] jgraham__: are you going to send feedback about aria-level? [14:18:00.0000] it seems to me that ARIA treegrid may kill as a native element [14:19:00.0000] how so? does aria treegrid have any effect on non-screen-reader software? [14:19:01.0000] I predict the API will resurface as a JS library written on top of HTML
[14:19:02.0000] Hixie: It doesn't but a JS library changing a table would have [14:19:03.0000] i think you've been breathing the accessibility koolaid for too long and have come to the mistaken assumption that most people care about accessibility [14:20:00.0000] and even consider it a priority in any sense [14:20:01.0000] I think they care about compat with browsers that don't have native [14:20:02.0000] breathing kool-aid, hah [14:20:03.0000] othermaciej: these days it's distributed in aerosol form [14:20:04.0000] my impression is that the average content author does not care much about accessibility, *but* [14:21:00.0000] some people have a business need to have their apps in some way "certified" for accessibility [14:21:01.0000] hsivonen: i'd hope we'd see libraries implement , yes, just like i hope to see the same for WF2 [14:21:02.0000] for instance for government use [14:22:00.0000] and toolkit developers therefore want their toolkits to be able to enable compliance [14:23:00.0000] sure [14:23:01.0000] and browser developers want JS library developers to like them [14:23:02.0000] sorry about spotty connectivity [14:23:03.0000] sure [14:23:04.0000] but the point is that it's not aria that would drive library development [14:23:05.0000] it might be added on afterwards as an afterthought [14:23:06.0000] that's basically what drives ARIA, the dependency chain of gov't contract requirements [14:23:07.0000] but that's wholly different [14:23:08.0000] I don't think anyone would code ARIA-first, no [14:24:00.0000] but toolkits like Dojo seem hot on adding it [14:24:01.0000] sure [14:24:02.0000] i'm just disagreeing with hsivonen's original statement [14:24:03.0000] what's a "pivot table"? [14:25:00.0000] "Note: If the expanded property is provided, it applies only to the individual cell. It is not a proxy for the container row, which also can be expanded. The main use case for providing this property on a cell is pivot table type behavior." [14:26:00.0000] oh. it's on wikipedia [14:30:00.0000] hsivonen: Sure I'll email about aria-level [14:30:01.0000] jgraham: great [14:42:00.0000] Is aria using RFC 2119 terminology? Section 6.2 seems to be using it but it's not clear if the rest of the document is or not [14:45:00.0000] jgraham: that would merit another comment :-) [14:46:00.0000] hsivonen: Yeah, I'm starting o accrue a stack of comments [14:47:00.0000] (reading the introduction to try and work out what terminology was being used reminded me of several minor issues I thought merited attention there) [14:48:00.0000] argh, pushed the send buton by mistake :( [14:48:01.0000] Press "undo" quickly [14:49:00.0000] so [14:49:01.0000] many [14:49:02.0000] e-mails [14:50:00.0000] Hixie: mind if I send you a message, broken into one character per email? [14:50:01.0000] yes [14:50:02.0000] but these e-mails are long [14:50:03.0000] that's the real problem [14:51:00.0000] "What is the most widely known language for authoring mathematics?" — Can I take a guess at LaTeX? [14:51:01.0000] no guesses [14:52:00.0000] gsnedders: In principle it could be MS-Word or ascii art of something [14:53:00.0000] s/of/or/ [14:53:01.0000] the answer might depend on what kind of math [14:53:02.0000] pen/paper? [14:53:03.0000] jgraham: Issue one: What is a language? [14:53:04.0000] I expect Word to be used less and LaTeX more as the math gets more complex [14:53:05.0000] We're lacking a good enough definition of what the question actually means to be able to answer it. [14:53:06.0000] I guess there's also a lot of math typeset in high end publishing programmes [14:53:07.0000] 42 [14:54:00.0000] I mean, is pen/paper well enough known? :P [14:54:01.0000] s/programmes/programs/ [14:54:02.0000] jgraham: does InDesign have a decent math typesetter? [14:54:03.0000] zcorpan_: that's definitely the correct answer. [14:54:04.0000] /me removed the word language from the wiki a few minutes ago [14:54:05.0000] or is it all TeX and Frame? [14:54:06.0000] gsnedders: yeah. but what's the question? :) [14:54:07.0000] what about Quark [14:54:08.0000] zcorpan_: 6 times 9 [14:55:00.0000] /me isn't sure what the "high end publishing programme" landscape looks like these days [14:55:01.0000] hsivonen: I dunno. All I know is that with Blackwells publishing all one's carefully typeset mathematics gets redone in some way [14:55:02.0000] (usually introducing errors in the process) [14:55:03.0000] Processing namespace e-mail... [## ] 10% [14:55:04.0000] jgraham: do you mean you send in LaTeX and they do something else? [14:56:00.0000] /me though TeX was the high end when it comes to math typesetting [14:56:01.0000] hsivonen: Yep. Although I don't know what "something else" is [14:56:02.0000] thought even [14:56:03.0000] hsivonen: it is [14:56:04.0000] (the final thing still says 'this document was prepared from a manuscript typeset by the author' or somesuch) [15:00:00.0000] jgraham said "gsnedders: In principle it could be MS-Word or ascii art of something" and that possibility is somewhat supported by me finding ASCII-art data tables for astronomy [15:01:00.0000] /me notes that's jgraham's subject [15:03:00.0000] I'm tempted to take the ostrich approach to aria-owns [15:06:00.0000] /me notes that the common forms for expressing maths are undoubtedly different to those for authoring it [15:06:01.0000] on the web [15:09:00.0000] Processing namespace e-mail... [#### ] 20% [15:09:01.0000] /me goes to get a drink [15:11:00.0000] /me considers complaining that the aria spec expressing its conformance requirements in RDF makes it inaccessible to people who don't think that RDF is human-readable [15:14:00.0000] jgraham: I'm reading the tables. I'm not reading RDF. [15:15:00.0000] I figured it wouldn't be cost-effective to try to generate stuff from the RDF [15:16:00.0000] hsivonen: I'm reading the tables too. But section 6.1 is clear that documents must conform to the RDF but doesn't seem to mention the tables [15:17:00.0000] As far as I've seen, computer science seems to be almost exclusively LaTeX, but areas like biology use Word a lot more, so you can draw very different conclusions depending on what subset of math-writers you look at [15:18:00.0000] Philip`: Undoubtedly there are far more biologists than physical+computer scientists but they probably generate fewer equations each, on average [15:37:00.0000] Hixie, fwiw, using the wiki more would be great [15:37:01.0000] (as you're doing now) [15:38:00.0000] it seems that people are missing some kind of entry point to find arguments for and against certain ideas that have already been discussed and "resolved"; having that documented somewhere other than in piles of e-mail is nice :) [16:03:00.0000] oh, about author mathematics [16:03:01.0000] *authoring [16:03:02.0000] at the HTMLWG meeting in Boston I just remembered talking to a couple of MathML guys [16:04:00.0000] when I asked them what people used before MathML, they immediately said "LaTex" [16:04:01.0000] one of them was working on version 3 of it, IIRC [16:05:00.0000] BenMillard: LaTeX3 or Mathml 3? [16:06:00.0000] LaTeX 3 [16:06:01.0000] http://www.latex-project.org/latex3.html - I think one of them was David Carlisle, the other names don't ring a bell [16:07:00.0000] Yeah, David Carlisle is in the MathML-WG [16:07:01.0000] i wish the person who posted http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2006-June/006518.html would have finished their sentence in para5... what does it mean it is important for us to do??? [16:13:00.0000] /me giggles at http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2006-June/006711.html [16:14:00.0000] according to my calculations there, a tag name costs $31,790 [16:16:00.0000] $10 per hour is about £5 per hour which I think is below minum wage in the UK [16:18:00.0000] yeah the numbers are a bit dubious [16:18:01.0000] It's right to an order of magnitude, though [16:18:02.0000] which is more accurate than many of the other numbers [16:20:00.0000] that's the cost of a tag name to who? [16:20:01.0000] the industry in general, afaict [16:20:02.0000] othermaciej: humanity [16:20:03.0000] hahahahahaha [16:21:00.0000] oh, there's also the time delay of the extra bytes being transferred over the wire :) [16:21:01.0000] if you're on company time that would count [16:23:00.0000] If the feature adds maybe a kilobyte to the size of the browser download, and there's a billion web users, then that's about an extra hundred dollars of bandwidth charges [16:23:01.0000] /me recalls this entire thread on mathematics becoming less than pleasant [16:24:00.0000] I meant the bytes for the tags in the document, but that's interesting as well [16:25:00.0000] the cost can only be in terms of time and non-renewable resources [16:25:01.0000] money that goes around the circle doesn't affect the total cost [16:27:00.0000] Time doesn't seem like a non-renewable resource - you can always just employ an extra person, and it will not make the world poorer [16:28:00.0000] it's time that that person would not spend doing what they want to do, one presumes [16:31:00.0000] anyway, it's just a bit of fun :) [16:42:00.0000] so far, out of 133 e-mails of mathematics discussion, a grand total of 2 lines, in just 2 separate e-mails, actually discussed what problem was being solved. [16:42:01.0000] that's an even worse ratio than the tables discussions 2008-03-25 [17:22:00.0000] 2 lines out of 160 e-mails so far [17:22:01.0000] sigh [17:23:00.0000] Does "I want latex2html to give less ugly output" count as a problem? [17:23:01.0000] it's closer than most of these e-mails [17:24:00.0000] "I have content in LaTeX and want to convert it to HTML and have it retain a high degree of typographical quality" is a problem [17:24:01.0000] though you should probably include an explanation of _why_ you want to convert it to HTML [17:24:02.0000] Because Adobe Reader is horribly slow, so I don't want to use PDF, and there's no other option [17:25:00.0000] well HTML isn't an option either now [17:25:01.0000] (Actually I'd just use PDF anyway, because it's people's own faults if they don't have a nicer PDF reader) [17:25:02.0000] so really your problem is "I have content in LaTeX and want to view it in a good viewer while retaining a high degree of typographical quality" [17:26:00.0000] at which point, HTML really doesn't seem like it has a good place in any solution for that problem [17:26:01.0000] It needs to be a widely-distributed reader too, so that most people are able to access my content [17:26:02.0000] and it needs to be indexed by search engines, so that people can find my stuff [17:27:00.0000] PDFs do get indexed by search engines? [17:27:01.0000] these all seem like LaTeX problems, not HTML problems :-) [17:28:00.0000] BenMillard: They do by at least Google, which is all that counts [17:28:01.0000] Hixie: They're still problems, and problems that could be solved by HTML [17:28:02.0000] how about "I have content in LaTeX and want to include it on my web page." [17:28:03.0000] to be honest i don't really see the problem here [17:28:04.0000] BenMillard: yes, that's a different problem entirely [17:29:00.0000] Philip`: if you have content in LaTeX, and you publish it to PDF, you already have a widely available viewer, good typopraphy, and searchability [17:29:01.0000] Philip`: seems like a solved problem [17:29:02.0000] BenMillard: and a valid problem [17:29:03.0000] PDF is its own world [17:30:00.0000] BenMillard: though not one that anyone on the mailing list seems to have mentioned so far :-) (in the 163 mails i've read) [17:30:01.0000] it seems nice in principle to be able to do math decently without escaping to a different basic technology for your whole document [17:30:02.0000] I have to agree that MathML syntax seems vile, now that I have looked more closely [17:31:00.0000] Hixie: Many people use latex2html now, despite it being slightly ugly, so there must be reasons why they do that, and I'm just not entirely sure what those reasons are :-) [17:31:01.0000] hixie, perhaps you are being to picky about their wording? [17:31:02.0000] Philip`: yeah, me either [17:31:03.0000] but that doesn't make the reasons not exist [17:32:00.0000] BenMillard: no, i'm trying to pull problems out of these e-mails, even if they haven't mentioned them intentionally. Sadly most of these e-mails are arguing about which syntax to use, on the basis of the syntax's own characteristics. [17:32:01.0000] not suitability to solving any particular problem [17:32:02.0000] hixie, ah [17:32:03.0000] Philip`: oh i agree. but if i don't know what the problems are, i can't solve them, or evaluate proposed solutions for them. [17:41:00.0000] man, i'm going to be skipping a lot of these e-mails in my eventual reply [17:41:01.0000] so much arguing back and forth about nothing [17:42:00.0000] the inexpressiveness of plain text can bring out the worst in people (including me) [17:42:01.0000] oh i've seen these discussions happen in person too [17:43:00.0000] the only difference is that those discussions don't end up in my inbox two years later :-P [17:43:01.0000] win! [17:44:00.0000] Elliotte Harold wrote: [17:44:01.0000] > The specific syntax is important because there's a huge, useful toolchain for processing XML and there's essentially zilch for [17:44:02.0000] > processing this strange HTML 5 thing. [17:44:03.0000] yay for all of you guys implementing the html5 tools over the past 2 years [17:56:00.0000] wohay, e-mail 250/368 is henri's e-mail earlier this month [17:56:01.0000] 118 e-mails in a few weeks [17:56:02.0000] hopefully _some_ will have problem descriptions [17:56:03.0000] because the first 249 e-mails had nothing beyond what's on the wiki now [18:04:00.0000] ok henri's e-mail more than doubled the size of that wiki page [18:15:00.0000] would the top level of list items be more useful as headings on that page? [18:16:00.0000] /me tries [18:16:01.0000] nah, that looks more ugly i think [18:17:00.0000] http://wiki.whatwg.org/index.php?title=New_Vocabularies&oldid=2894 is what it looked like [18:17:01.0000] i've reverted it [18:19:00.0000] i wonder if mathml would take off more if it were available in text/html, but IE continued to not support it [18:21:00.0000] are there any useful tools for producing mathml? [18:21:01.0000] hm, http://www.w3.org/mid/ABFEED64-8A5F-43C9-981B-416C6F65EA50⊙if is a very good point [18:22:00.0000] vlad_: dunno [18:22:01.0000] and potentially some way to easily put in normal-looking ascii math (perhaps with some script)? [18:22:02.0000] yeah, i really don't want to require that people write out raw mathml [18:22:03.0000] that's just insane stuff [18:22:04.0000] like, being able to type x^2/y^2 and have it do the right thing would be nice [18:22:05.0000] even if it's something like x^2/y^2 and then there's some JS that translates [18:23:00.0000] JS translating latex math expressions to mathml would probably be doable [18:23:01.0000] i'd rather not rely on JS here [18:23:02.0000] vlad_: thats a font feature [18:23:03.0000] but that's another issue [18:26:00.0000] vlad_: The OO.o equation editor produces MathML and could be considered useful [18:27:00.0000] hixie, the headings version seems a more conventional wiki page, to me. [18:27:01.0000] BenMillard: i'm not a conventional kind of guy :-) [18:28:00.0000] Pavlov_: e.g. http://www.zvon.org/HowTo/Output/MathMLtestsuite_abs1.php [18:29:00.0000] is a pretty crappy way to write |H/K| (granted, you also want italic H and K, but still) [18:29:01.0000] well [18:29:02.0000] its not great [18:30:00.0000] maybe some simple stuff would be OK [18:31:00.0000] feedback on HTML5 from a friend I knew when I did GTA modding some years ago: "Damn you guys are fuckin smart" [18:32:00.0000] I guess a poor first impression of the HTML5 effort is not a universal experience [18:32:01.0000] heh [18:33:00.0000] /me also happened to do GTA modding some years ago, where 'some' = 9 [18:33:01.0000] ooh, I modded the early ones too :) [18:34:00.0000] speaking of GTA, is GTA4 out yet? [18:34:01.0000] /me has some cash earmarked for a PS3 and GTA4 and is just waiting [18:37:00.0000] http://www.rockstargames.com/IV/events/ - 29th April 2008 [18:37:01.0000] /me calendars it [18:38:00.0000] http://web.archive.org/web/20020121215533/http://gta2.games.ultrastore.com/ - yay, my site sort of still exists [18:39:00.0000] OMG I REMEMBER THAT WEBSITE [18:41:00.0000] Philip`, this is a muliplayer level I made for GTA2: http://projectcerbera.com/gta2/multislayer/ [18:41:01.0000] Unfortunately I barely knew what I was doing, and didn't know how to write GUI programs in anything other than Visual Basic [18:42:00.0000] I still don't know how to write GUI programs in anything other than VB :$ [18:42:01.0000] Later I learnt how to write GUI programs in Perl, which wasn't really any better [18:43:00.0000] heh [18:44:00.0000] (The only ones I've liked were in C++ / wxWidgets) [18:44:01.0000] delphi baby [18:45:00.0000] I like to write GUI programs in HTML+CSS+JS [18:46:00.0000] everything else seems annoying [18:46:01.0000] (though ObjC+Cocoa is not so bad as such things go either) [18:46:02.0000] othermaciej: seriously.. though HTML could use some love [18:46:03.0000] BenMillard: The world looks oddly familiar there - the textures and shapes are all familiar, but I got bored and moved away before the map editor was released so the perspective is odd [18:46:04.0000] (cough xul box model cough) [18:46:05.0000] vlad_: we got your xul box model [18:47:00.0000] I wish the CSS WG would show it some love [18:47:01.0000] yeah [18:47:02.0000] instead of figuring how to full-justify arabic text in the presence of chinese vertical writing that uses double-width characters [18:47:03.0000] all it would take is an editor [18:48:00.0000] othermaciej: we should really fedex those people a nice pen and some parchment [18:48:01.0000] othermaciej: HTML+CSS+JS doesn't work so well when you're trying to decode and render graphics from complex binary files on the user's computer, and trying to do it fast [18:48:02.0000] and they can have at it [18:48:03.0000] Philip`: oh, well, I wouldn't use it for the non-GUI part [18:50:00.0000] othermaciej: Do you mean you'd have a non-GUI part in a proper language (C++ or Python or whatever) and expose a web interface to it for all interaction? [18:51:00.0000] if I were called on to write a GUI app I would probably do a desktop/web hybrid with most of the UI in HTML and compute-intensive or system-integrating code in a lower level language [18:52:00.0000] Philip`, yeah seeing proper perspective in a 2.5D sprite-based game is pretty odd :) [18:53:00.0000] Philip`, I remeber using M1win for GTA1, which had no draw distance limit. that was cool [18:54:00.0000] Philip`, the textures for MultiSlayer are actually from the Vice City level in GTA1 [19:02:00.0000] hey enough reminiscing :) I have a question about the state of the HTML5 spec [19:03:00.0000] would you say that HTML5 is an "upcoming spec"? [19:03:01.0000] /me is looking for shiny words [19:04:00.0000] upcoming? [19:05:00.0000] yes, as in "of the upcoming HTML5 specification" [19:06:00.0000] would "HTML5 specification draft" be better? [19:06:01.0000] if it's not already "upcome", for lack of a better term, i wonder what it will take for it to do so :-) [19:06:02.0000] what's the whole sentence? [19:06:03.0000] i would just say "HTML5" [19:06:04.0000] but that's just me [19:18:00.0000] /me thinks the peeps here are awesome [19:27:00.0000] oh crikey it's 2:30am [19:45:00.0000] Marco Zehe blogged about labelledby and describedby. I chatted to him in Mozilla's #accessibility channel and blogged an HTML4 alternative: http://projectcerbera.com/blog/2008/03#day24 [21:24:00.0000] hsivonen: yt? [21:24:01.0000] i don't understand your use case in http://www.w3.org/mid/3733150B-0B02-496D-8517-A9A6D040B3E8⊙if [21:24:02.0000] can you give an example of a page doing something like that? [21:26:00.0000] in particular, i don't understand why isn't the best solution for that (even in xhtml today) [21:34:00.0000] hm, people want to be able to extract svg snippets from html documents and edit them in svg editors [22:23:00.0000] i am seriously considering not individually replying to all these e-mails [22:23:01.0000] even the ones sent to whatwg [22:23:02.0000] there is so much bikeshedding in these threads it's insane [22:23:03.0000] especially since what we're actually talking about here is the nuclear plant [22:53:00.0000] holy crap [22:53:01.0000] Philip` provides the first really detailed problem description: http://www.w3.org/mid/47DC70FB.1060205⊙cau [22:56:00.0000] ok i've read every e-mail [22:56:01.0000] http://wiki.whatwg.org/wiki/New_Vocabularies summarises what i found [01:42:00.0000] Hixie: http://lists.w3.org/Archives/Member/w3c-wai-pf/2008JanMar/0507.html [03:48:00.0000] re: typing
latex
http://www.math.union.edu/~dpvc/jsmath/ [03:50:00.0000] someone who knows HTML5 drag&drop might want to look into the ARIA drag&drop stuff [07:10:00.0000] Hixie: the icon case is basically Sam's blog [07:11:00.0000] Hixie: reasons for not using : 1) Fallback behavior. 2) manageability as part of the markup text in a CMS. [09:50:00.0000] Hixie? [10:32:00.0000] hsivonen: i honestly think that sam's blog would be better off using [10:32:01.0000] andersca? [10:33:00.0000] Hixie: hey [10:33:01.0000] Hixie: about the cache update process [10:33:02.0000] Hixie: is the order of the files in the file list important [10:33:03.0000] Hixie: (the file list described in step 13) [10:38:00.0000] um [10:39:00.0000] it just says "For each URI in file list" on step 20, without specifying order, so i guess it doesn't matter [10:41:00.0000] the only time order matters is during parsing [10:41:01.0000] because it affects duplicate elimination in the fallback mode [10:41:02.0000] yeah [10:56:00.0000] /me tries to decode "Content-reinforcing contextual progressive-enhancement eye-candy" [10:57:00.0000] i really have no idea what that means [10:57:01.0000] good luck with that ;) [10:57:02.0000] maybe hsivonen is just again referring to sam's icons [11:04:00.0000] Hixie, yeah, Sam's icons would be a good example of that. I think it means content that is related to the surrounding content, but doesn't provide any additional information, and is done using progressive enhancement techniques [11:04:01.0000] yeah so i really don't see why you wouldn't want to use for that [11:05:00.0000] duplicating the icon every time it is used seems like an anti-pattern [11:05:01.0000] bad for caching, bad for maintainability, etc [11:08:00.0000] yeah, I don't know why wouldn't be better either, but I don't know why Sam does it the way he does [11:08:01.0000] maybe SVG in isn't as widely supported yet [11:09:00.0000] that's obviously not a relevant argument when discussing new features in a spec with an ETA of more than 10 years from now :-) [11:09:01.0000] SVG in only works in Opera 9.5, so it's useless in practice [11:10:00.0000] i thought i heard it worked in webkit now too [11:10:01.0000] Oh, okay, that could be true [11:10:02.0000] but if it doesn't work in Firefox, it's still useless in practice [11:11:00.0000] well sure [11:11:01.0000] but like i said [11:11:02.0000] not a problem when we're discussing future technologies [11:12:00.0000] yeah, I know. It's not an argument, merely a possible explanation [11:18:00.0000] http://wiki.whatwg.org/wiki/New_Vocabularies updated [11:18:01.0000] lots more on there now [11:23:00.0000] how the hell do we solve the custom-data-in-html issue [11:26:00.0000] Namespaces! [11:27:00.0000] Or make any element name and attribute name conformant [11:31:00.0000] "namespaces" would just lead to people taking divitis to the next level [12:13:00.0000] hmm. even http://www.dreamhoststatus.com/ is down, that probably means dreamhost is having serious problems right now [12:13:01.0000] that sucks, cause I even log into my email right now [12:15:00.0000] allowing any element or attribute name to be conforming would make the language less useful and prevent future standardised extensions [12:15:01.0000] Lachy: dreamhoststatus is offsite [12:17:00.0000] should be noted that there is a conflict between Cogent and TeliaSonera that may affect connectivity for you [12:19:00.0000] who are Cogent and TeliaSonera? [12:19:01.0000] but the problem is apparently with dreamhost, they wrote: [12:19:02.0000] "Sorry about the downtime. A lot of you have already been receiving blingy status messages. Blingy is your cluster of machines for web, mysql, and email. The main file server for this cluster is having serious problems." [12:24:00.0000] dreamhost is fine for me [12:24:01.0000] i'm editing through it right now [12:24:02.0000] the status page does seem down though [12:26:00.0000] Lachy: backbone providers in the US and scandinavia, respectively [12:30:00.0000] my site is still up, it's just my email having problems [12:31:00.0000] I'm probably also on a different server from you [14:27:00.0000] othermaciej: that use case is the first one people have presented that actually would require just "raw xml"-equivalent support in text/html [14:29:00.0000] Hixie: is that because XBL2 inherently requires raw XML? [14:29:01.0000] not so much raw XML, but it allows you to do things like have [17:03:03.0000] jgraham__: we're talking about author extensions, not vendor extensions [17:04:00.0000] With scripting the difference between authors and vendors isn't so large [17:04:01.0000] large enough for me [17:04:02.0000] :-) [17:05:00.0000] I can't say I particularly disagree, but I won't be surprised if others do [17:05:01.0000] sure [17:06:00.0000] having both would work [17:06:01.0000] custom elements makes sense for XBL [17:07:00.0000] why? [17:07:01.0000] you mean in the shadow tree? [17:07:02.0000] /me -> bed [17:07:03.0000] or for binding [17:07:04.0000] for xbl, and for aria for that matter, and libraries, you want to "bind" to the most appropriate html element, so that the document still mostly works without them [17:07:05.0000] for the bound element [17:08:00.0000] [17:08:01.0000] [17:09:00.0000] (i'm not sure why
would be better apart from maybe being slightly faster with styling given cached class names) [17:09:01.0000] well ideally it wouldn't be
[17:12:00.0000] man, svg is worthless [17:12:01.0000] wtf are the ua conf requirements for ? [17:12:02.0000] u1 rather [17:12:03.0000] and u2 [17:15:00.0000] I've no idea [17:27:00.0000] annevk: yt? [17:27:01.0000] annevk: can you check that the build that got 100/100 did go down to 99/100? [17:28:00.0000] no, but i'm passing on the change madea [17:28:01.0000] please do get someone to confirm that the score went down [17:29:00.0000] otherwise there's another bug in the test [17:29:01.0000] kk, i guess i'll hear it at some point [17:29:02.0000] thanks [17:30:00.0000] not exactly day time in Norway :) [17:30:01.0000] / Sweden [17:30:02.0000] yeah [17:38:00.0000] maybe better here [17:38:01.0000] Hixie: you have to give the EE glyph a glyph-name="EE" too [17:40:00.0000] oh [17:40:01.0000] so how did opera pass this? [17:40:02.0000] that makes no sense to me [17:40:03.0000] where do i add this glyph-name thing... hmm... [17:40:04.0000] wait, it already has unicode=EE [17:40:05.0000] that's not enough? [17:41:00.0000] oh that's the kerned version? [17:41:01.0000] wait [17:41:02.0000] i'm VERY confused now [17:41:03.0000] Hixie, no it needs an explicit name [17:41:04.0000] the EE glyph spec should be this: e('glyph', { 'unicode': 'EE', 'd': 'M100,0 h100 v-100 h-100 z', 'horiz-adv-x': '1300', 'glyph-name': 'EE'}) [17:41:05.0000] the kerning pair references it by name now, not by unicode value [17:41:06.0000] those are separate [17:41:07.0000] (insanely enough) [17:41:08.0000] so that's the unicode=EE thing?? [17:41:09.0000] what [17:41:10.0000] rather [17:41:11.0000] unicode gives the actual characters that map to the glyph [17:41:12.0000] unicode=EE is the chars the glyph mathces [17:41:13.0000] the name could be anything [17:41:14.0000] but kerning pairs match differently than glyphs [17:42:00.0000] (insanely enough) [17:42:01.0000] glyph-name="theEEGlyph" e.g. [17:42:02.0000] ok fixed [17:42:03.0000] so why is u2=EE wrong, given that unicode=EE is ok? [17:42:04.0000] u2 is defined in terms of unicode, no? [17:42:05.0000] so confused [17:43:00.0000] no, u2 is a list of characters (and possible ranges) -- it could be "E,U+1234-U+1300" [17:43:01.0000] iirc [17:44:00.0000] oh, mistake in the spec [17:44:01.0000] s///g in the u2="" definition [17:44:02.0000] omg [17:44:03.0000] Hixie: unicode is a string of characters, u2 is a comma-separated list [17:44:04.0000] /me suggests dropping all SVG tests [17:45:00.0000] ok fuck this, we're not adding svg to html5 [17:46:00.0000] lets add svg5 [17:46:01.0000] svg fonts are one of the more insane parts [18:02:00.0000] gah this custom data thing is dhard [18:02:01.0000] it's actually hard to come up with real examples [18:03:00.0000] er, the test is changing after it was published? [18:04:00.0000] to fix bugs that have been found, yes [18:04:01.0000] thats kind of confusing my brain [18:04:02.0000] but OK [18:05:00.0000] one advantage of not trying to pass the test straight away is that you don't have to worry about such changes :-) [18:05:01.0000] it's just like point releases for software [18:05:02.0000] for security fixes [18:05:03.0000] and stuff [18:05:04.0000] except it isn't [18:05:05.0000] Acid3 SP1 [18:05:06.0000] sure [18:05:07.0000] i'd rev the version [18:06:00.0000] its just confusing to target one thing and then have it change [18:06:01.0000] the target is the spec [18:06:02.0000] not the test [18:06:03.0000] targetting the test is missing the point [18:07:00.0000] so everyone has missed the point? [18:07:01.0000] that's a common problem [18:07:02.0000] the webkit guys haven't been targetting the test, that's why they keep finding the bugs in the test :-) [18:07:03.0000] adding an idl file to pass the test isn't targetting the test? [18:07:04.0000] they've been prioritising passing the test [18:08:00.0000] that's not the same as just changing the code to pass the test [18:08:01.0000] (which i don't think anyone would actually do, reddit crowd notwithstanding) [18:12:00.0000] Philip`: ironically, doing getAttributeNS() is exactly what you say you do (and want to have still work in text/html) in your e-mail http://lists.w3.org/Archives/Public/public-html/2008Mar/0156.html [18:24:00.0000] http://wiki.whatwg.org/wiki/CustomData [19:08:00.0000] Hixie: what was the acid3 "fix"? [19:09:00.0000] is there a revision log somewhere with changes? [19:17:00.0000] does Opera fully pass Acid3 now? [19:18:00.0000] they did for about 2 hours today [19:19:00.0000] i thought there were still some rendering issues. [19:19:01.0000] "for about 2 hours"? [19:20:00.0000] acid tests should be scored as pH [19:20:01.0000] acid3 was updated?!? [19:21:00.0000] again [19:21:01.0000] thats what wiki is saying... [19:53:00.0000] http://webkit.org/blog/173/webkit-achieves-acid3-100100-in-public-build/ [19:58:00.0000] so how we doing on acid5? [20:54:00.0000] /me pre-orders the "Acid5 > Acid2" t-shirt [20:55:00.0000] haha [21:40:00.0000] Hixie: I think we have a rendering pass now [21:40:01.0000] Hixie: but I can't really tell on the animation smoothness [21:40:02.0000] it looks smooth to me in a release build on my MacBook Pro, but I don't exactly have the best calibrated eyeballs [00:23:00.0000] Hixie: umm. the colon is bad exactly for the same reasons it was bad for ARIA [00:28:00.0000] i didn't say it was good [00:28:01.0000] i haven't found a good solution yet [00:30:00.0000] Philip`'s custom data example was in XML or was it even SVG [00:30:01.0000] the Inkscape cruft and similar happen in SVG trees [00:30:02.0000] where the legacy considerations are differnt from HTML [00:32:00.0000] /me wonders if hendry is around [00:44:00.0000] hmm. Adobe dilutes the Photoshop brand with a Flash app that is less photoshop-like than competing flash apps... [01:21:00.0000] is it known whether http://www.w3.org/MarkUp/2008/ED-xhtml-role-20080318/ arose of stated use cases? [01:37:00.0000] i wish the people who keep saying acid3 doesn't test useful stuff would be more specific about what's not useful [01:37:01.0000] sigh [01:40:00.0000] it is edge casey but in many cases that forced us to implement non-edge cases of things too [01:41:00.0000] I do think testing DOM 2 Events is pretty useful, if way belated [01:41:01.0000] i think i probably overstepped usefulness with the dom traversal tests [01:42:00.0000] but not sure what else to test with it [05:27:00.0000] custom no-namespace attributes starting with underscore or so would be convenient, i think [05:28:00.0000] [05:30:00.0000] MikeSmith: yes? [05:31:00.0000] hendry - wanted to ask if you knew much about details of JSR 248 [05:32:00.0000] in particular, does it include the location API [05:32:01.0000] (I forgot what the JSR number for that is) [05:32:02.0000] MikeSmith: JSR179 [05:33:00.0000] i don't know the details off hand [05:33:01.0000] OK [05:33:02.0000] /me is sick at home with flu :/ bit useless today [05:33:03.0000] The JCP annoyingly only provides specs in PDF form [05:34:00.0000] I'll quit being lazy and I'll go ahead and download the PDF and see [05:37:00.0000] hmm [05:38:00.0000] hendry - it says that JSR 179 is "conditionally mandatory" [05:38:01.0000] whatever tf that means [05:39:00.0000] [[ [05:39:01.0000] JSR 179 MUST be implemented if the target device meets at least one of the following [05:39:02.0000] conditions: [05:39:03.0000] • The device has a GPS receiver that is able to deliver the geographical coordinates [05:39:04.0000] within the device [05:39:05.0000] • The device supports a location method that is capable of delivering the geographical [05:39:06.0000] coordinates and is used to deliver the coordinates to downloadable applications (in [05:39:07.0000] Java ME or other runtime platforms) [05:39:08.0000] • The device supports an accessory device that can be used to obtain geographical [05:39:09.0000] coordinates, and which is used to deliver the location to downloadable applications [05:39:10.0000] (in Java ME or other runtime platforms) [05:39:11.0000] ]] [05:39:12.0000] oops [05:39:13.0000] sorry [05:39:14.0000] that was a bit more than I intended to paste.. [05:41:00.0000] so seems, in a nutshell, if any device supports has some kind of location-sensing capability and it claims to comply with JSR 248, then it must support JSR 179 [05:46:00.0000] and my understanding is that JSR 248 is supposed to be sort of the baseline/"lowest common denominator" set of JSRs that Java ME environments should support [06:29:00.0000] hmm. I wonder if the correct way to mark up the term "a priori" is with [06:44:00.0000] takkaria: a priori [06:45:00.0000] Philip`, that doesn't degrade as nicely as would when author CSS is not used [06:46:00.0000] both seem fine to me [06:47:00.0000] Philip`: hm, interesting [06:47:01.0000] I should have thought of lang [06:47:02.0000] Philip`: would you expect a proper Latin pronunciation or an English approximation? [06:49:00.0000] BenMillard: But the italic styling is just a presentational effect that is conventional in high-res visual contexts with suitable text rendering, and isn't universally applicable, whereas any UA could process lang=la in a suitable way for its particular medium [06:50:00.0000] Philip`: my point is: is proper Latin pronunciation really the suitable thing for a user whose listening comprehension is calibrated for English? [06:51:00.0000] Philip`, does not preclude UAs doing their own thing: user stylesheets can override the italic with something more approrpriate, although I've no idea what that would be [06:52:00.0000] also, for the languages where is conventional, you don't need "high-res visual contexts which suitable text rendering". for example, Lynx running on a terminal will apply colour rather than italicising [06:52:01.0000] s/which/with [06:53:00.0000] BenMillard: doesn't allow UAs to distinguish the Latin italics from the e.g. ship name italics [06:53:01.0000] ...though I suppose you should use then [06:54:00.0000] yes, my suggestion was rather than . I could have made that clearer :) [06:54:01.0000] hsivonen: Hmm, I suppose that's a problem [06:55:00.0000] Maybe it should be lang="en-la" or something [06:55:01.0000] (though that's probably incorrect usage) [06:55:02.0000] English is a melting pot of lots of languages; I think Latin phrases like which have been in use as long as they have are basically English now [06:56:00.0000] BenMillard: There are lots of words that nobody even notices are based on Latin, but there are still words and phrases that people tend to write in italics to indicate that they're foreign [06:56:01.0000] indeed, there are basically 3 categories as I understand it: [06:57:00.0000] 1. English words [06:57:01.0000] 2. words from other languages commonly used in English and pronounced as English (such as these Latin phrases) [06:57:02.0000] 3. foreign words [06:57:03.0000] 2. could use foo [06:58:00.0000] 3. could use foo? [06:58:01.0000] the italicising of a priori (et al) seem to me to be equivalent to using scare quotes round a word [06:58:02.0000] "Our 'a priori' knowledge of the world..." vs. "Our /a priori/ knowledge of the world..." [06:59:00.0000] not necessarily an indication of the source language [07:00:00.0000] I think italicising a priori makes it much easier to read, because otherwise I start to read as if it was talking about a priory and then have to backtrack and switch to a pseudo-Latin mode and carry on [07:00:01.0000] s/italicising/visually distinguishing in some way/ [07:01:00.0000] It would be much easier if England hadn't got itself invaded so much [07:05:00.0000] we've avoided that for nearly 1,000 years, IIRC. the source of our diversity is a very long history [07:54:00.0000] Philip`: we did our fair share of invading back :) [07:58:00.0000] camaban, yeah and brought even more new words back with us :P [07:59:00.0000] and curry [07:59:01.0000] mmmm [08:35:00.0000] /me sees http://canvex.lazyilluminati.com/misc/cgi/issues.cgi/folder_expand/global-custom pointing to http://www.quirksmode.org/blog/archives/2007/04/html_5.html pointing to http://www.alistapart.com/articles/scripttriggers/ suggesting to use a custom DTD so that custom attributes are valid, which doesn't sound like a good idea [08:36:00.0000] Also that last page's custom attributes seem to conflict with WF2, which doesn't sound good either [08:56:00.0000] By the way, why does A List Apart say "Was it good for you, too?" at the end of the articles? That really doesn't fit with the style of the site [10:10:00.0000] html5 parsing of ... seems to be incompatible with real pages [10:27:00.0000] MikeSmith: You're right there [11:50:00.0000] Hixie: i think the problem is that Acid3 divided web developers and useres into ie|firefox and webkit|opera ;) now the mozilla guys have to come up with explanations why they fail the test, and obviously the first one is that acid3 doesn't matter [11:51:00.0000] beside that are they correct. acid3 doesn't matter *as long* as firefox and IE are failing it as they have the lion share of the browser market currently [11:56:00.0000] Acid3 only matters if people think it matters - otherwise it's just like any other list of browser bugs that might hang around for years with nobody caring enough to fix them [11:57:00.0000] same goes for standards in general [12:01:00.0000] major browser vendors saying it doesn't matter does create something of a self-fulfilling prophecy perhaps [12:01:01.0000] still, I think Acid2 ended up mattering, despite being initially downplayed in some quarters [12:19:00.0000] zcorpan_: don't want the custom attributes to be _too_ convenient... just more convenient than abusing html and clashing with future extensions [12:20:00.0000] oh, that sounds like the tail end of a conversation I wish I'd been here for. *runs off to the IRC logs* [12:22:00.0000] there needs to be enough encouragement for people to use them [12:22:01.0000] custom attributes are causing pain for Web Forms 2 already [12:22:02.0000] (I didn't see what zcorpan_ said btw) [12:24:00.0000] he said to use _foo="" [12:24:01.0000] wfm, but I wonder if authors would do that correctly [12:25:00.0000] maybe x- [12:27:00.0000] nah, dismiss that [12:27:01.0000] x is experimental, not custom [12:27:02.0000] c-foo? [12:28:00.0000] less characters the better [12:28:01.0000] nearer to other HTML attributes the better [12:28:02.0000] custom attribute technique [12:28:03.0000] cat- [12:29:00.0000] i was also thinking pua-, custom-, private- [12:29:01.0000] but all this is moot when you consider that the use case Philip` pointd out is already abuse [12:30:00.0000] class="ropacity25" [12:30:01.0000] I remember chatting about this briefly in a taxi with hsivonen during the HTMLWG F2F in Boston 2007 [12:31:00.0000] /me is still reading through the log for today [12:36:00.0000] the thing I'm most interested by is removing data from the title attribute and putting it somewhere out of the way [12:37:00.0000] Microformats put various types of data in title which are not human-readable [12:38:00.0000] fwiw, if there's a lot of deployed content that'd be hard to change [12:43:00.0000] annevk, the Microformats authors generally seem responsive to developments in the formats. so perhaps not impossible [12:44:00.0000] I mean, authors who use Microformats [12:53:00.0000] k [12:53:01.0000] _foo is ugly but WFM [12:54:00.0000] -x apparently does not work for XML [12:54:01.0000] I think human-unfriendly data in title is an antipattern [12:55:00.0000] Hixie: Dojo has lots of custom attribute use cases, though I don't expect you to like them :-) [12:55:01.0000] yeah, i know [12:56:00.0000] custom-foo= or private-foo= seem like the most plausible solutions [12:57:00.0000] "custom-" and "private-" are longish [12:58:00.0000] yes [12:59:00.0000] annevk: why not :? [13:00:00.0000] the problem with custom- and private- prefixes will be getting authors to use them, which will be difficult given their length [13:01:00.0000] the problem with _foo is that it's ugly and doesn't convey that it's not really ok [13:01:01.0000] (as a python programmer _foo means don't touch it) [13:02:00.0000] I think conveying that you don't think custom attrs are OK by making them suck is not a good policy [13:02:01.0000] why not? [13:02:02.0000] how else would we do it? [13:02:03.0000] and it's not like custom- and private- suck _that_ much [13:02:04.0000] no more than _, imho [13:02:05.0000] they are readable, as a pro [13:03:00.0000] Hixie: what's the harm with custom attributes vs. loading script data over, say, XHR? [13:03:01.0000] hsivonen: nothing, that's a valid use. the problem is when people start using them for things that should be in html markup [13:03:02.0000] mitsuhiko, mostly because it clashes with XML [13:03:03.0000] e.g.
[13:03:04.0000] mitsuhiko, just like starting with - [13:03:05.0000] [13:04:00.0000] ? [13:04:01.0000] something IE could do to sneak past validators [13:04:02.0000] annevk: xml doesn't disallow : per se, but xmlns aware parsers would probably irk [13:04:03.0000] mitsuhiko, oh, that may be true, but the world is based on namespace aware parsers :) [13:05:00.0000] touche ;) [13:05:01.0000] annevk: i did too much wordpress lately [13:05:02.0000] they are parsing with regular expressions.. [13:05:03.0000] and yes, letting vendors invent new values that end up passing through here is another problem [13:06:00.0000] hmm. I think "Required Child Elements" in ARIA means "Only kind of permitted child elements" [13:07:00.0000] annevk, what's the difference between experimental and custom? [13:07:01.0000] experimental is not conforming [13:08:00.0000] ah, thanks [13:08:01.0000] Experimental is deployed by multiple parties forever. Custom is deployed by a single party. [13:08:02.0000] true, experimental is generally a bad idea [13:08:03.0000] imagine having to standardize on [13:09:00.0000] experimental features which prove useful enough are then standardised as standard features, without experimental syntax, and the people who participated in the experiment update their content? [13:10:00.0000] haha [13:10:01.0000] is there existing content with treegrids built out of SVG? [13:10:02.0000] BenMillard: precisely. except for the bit where the name changes. :-) [13:10:03.0000] actually that's not really fair [13:10:04.0000] css has good experience getting names changed after the experimental phase [13:11:00.0000] will be fun to see if MS actually changes any NS URIs in OOXML... [13:11:01.0000] wow: http://www.paciellogroup.com/blog/?p=50 [13:11:02.0000] the reason why I ask about treegrids is that I'm considering sending a comment about optimizing grid and treegrid for
and making them
-only [13:14:00.0000] hixie, CSS property names and values were examples I was going to give [13:15:00.0000] annevk, what do you find surprising about that piece? [13:15:01.0000] In Steve's blog post, what is a line? [13:15:02.0000] with CSS it sort of works because of the "error handling" is vastly different [13:15:03.0000] BenMillard, the way he tries to bash Hixie [13:16:00.0000] without really good reason, because the case Hixie was talking about doesn't work well by default and the Jaws people actually acted on bug reports done by Hixie [13:16:01.0000] Because if I understand correctly, Jaws sounds, naively, broken by design [13:17:00.0000] (as in, my naive reaction to its behaviour is "that's broken") [13:18:00.0000] yes, I get the feeling they have such a hard time squeezing any amount of accessibility from all the desktop apps and formats their customers what to use that basic "fit and polish" gets a bit neglected [13:19:00.0000] s/what/want [13:19:01.0000] /me seriously hopes the open source screen readers shake up the market [13:20:00.0000] time for dinner, bye all [13:20:01.0000] hm, henri once suggested just using script-private="..." [13:30:00.0000] hmm. does VoiceOver have a command "read from this point onwards until I press a key"? [13:41:00.0000] Lachy: The only people who will use a specified custom-attribute syntax will be people who care about conformance, because people who don't care can just use anything they like and it'll work just as well; and for the people who do care, they'll do the easiest thing the conformance checker lets them get away with, and 'custom-foo' is easier than adding attributes through script or writing a custom DTD [13:45:00.0000] Philip`, it's the ones that don't care about validation that are the problem. [13:46:00.0000] Lachy: There's nothing we can do to those people [13:46:01.0000] (unless they use libraries written by people who do care) [13:47:00.0000] We're not adding any feature, we're just changing conformance, so people who don't care about conformance won't be affected at all [13:53:00.0000] Philip`, one of the problems that custom attributes is attempting to solve is reducing the possibility of clashes, especially with future standards. [13:55:00.0000] css has solved that neatly [13:55:01.0000] (without namespace indirection) [13:55:02.0000] finding a solution that is as easy or easier to use than not using it increases the chance that people who don't care so much about validation will use it [13:56:00.0000] zcorpan_: Custom extensions to CSS are useless because they don't do anything at all, which is why nobody does that and there's no problem [13:56:01.0000] Philip`: there's UA-specific extensions [13:58:00.0000] zcorpan_: UA-specific extensions are done by people who know what they're doing and tend to follow guidelines, which is totally different from custom attributes in HTML content [13:59:00.0000] Philip`: yes, but the same syntax can be used for html nevertheless [14:01:00.0000] I don't know if UA-specific extensions are always done by people who know what they are doing [14:01:01.0000] I mean, look at
or whatever)) [14:18:00.0000] Oh, you could just use .data['var'] all the time instead, to avoid the special cases [14:18:01.0000] yeah [14:20:00.0000] ok well i'm gonna go to work, but unless someone finds a better solution, i'm going with data-*="" and .data[*] (the latter not being available on