00:16
<Hixie>
maybe instead of .data['foo']
00:16
<Hixie>
we should have .getData('foo')
00:16
<Hixie>
but then it's not much better than .getAttribute('data-foo') anyway
00:19
<annevk>
.dataset.foo
00:20
<Hixie>
the advantage of .data.foo is that we could provide .data.foo.has('foo')
00:20
<Hixie>
or .data.foo.asList(',')[0] if it's a comma separated list or something
00:20
<Hixie>
i guess that'd just be .data.foo.split(',')
00:21
<Hixie>
i should look at the dojo source
01:06
<eseidel>
othermaciej, Hixie: continueing from #webkit..
01:06
<eseidel>
so you do you have an opinion on postMessage sync behavior?
01:07
<eseidel>
as I was saying in #webkit, I think sync is broken/bad for browsers
01:07
<Hixie>
i don't have any strong opinions
01:07
<Hixie>
i don't recall why we did it synchronous
01:07
<eseidel>
Hixie: my issue is in the scrollback of #webkit. but basically, postMessage being sync seems like it could cause issues for FF or Safari, once they land squirrel fish
01:08
<othermaciej>
I don't think postMessage being sync creates implementation-specific issues
01:08
<eseidel>
since with suspendable js, tabs can be running js in parallel
01:08
<Hixie>
wtf is squirrel fish
01:08
<eseidel>
Hixie: yet-another-re-write of the js interpreter
01:08
<othermaciej>
I'm also not sure we'd ever support different tabs running JS in parallel if it can be observed
01:09
<othermaciej>
it's our silly code name for a JS refactoring branch
01:09
<eseidel>
well, back to the basic issue. tab opens another tab. first tab posts to the second, second tab hangs
01:09
<othermaciej>
async feels right and works better if it is likely postMessage sequences will involve lots of replies back and forth
01:09
<othermaciej>
I can't really argue for it more than that
01:09
<eseidel>
assuming JS isn't run in teh UI thread
01:09
<eseidel>
which was always my point with bytecode work
01:10
<Hixie>
well i'm certainly ok with changing it
01:10
<eseidel>
I wanted an interruptable interpreter
01:10
<Hixie>
but i guess it depends on the browsers
01:10
<othermaciej>
keep in mind, if two windows are same domain they can call each other directly
01:10
<Hixie>
postMessage() has change a dozen times already
01:10
<othermaciej>
without postMessage
01:10
<Hixie>
the mozilla guys will kill me if it changes again
01:10
<othermaciej>
in Safari we turned it off for 3.1 due to spec changes
01:10
<eseidel>
othermaciej: so maybe i'm missunderstanding it. maybe it's when two tabs are opened independantly
01:10
<othermaciej>
I would not be against making it async
01:10
<othermaciej>
I don't feel strongly
01:11
<othermaciej>
eseidel: to use postMessage you still need a reference to the other window object
01:11
<othermaciej>
the only new thing it opens up is that this can be cross-domain
01:13
<dglazkov>
the async spec is what Hixie was working on?
01:13
<Hixie>
yeah there is an async message thing i'm working on too
01:13
<Hixie>
but that's separate really
01:13
<Hixie>
window.postMessage could be sync or async independent of this
01:14
<dglazkov>
http://hixie.ch/specs/dom/messages/0.9
01:14
<Hixie>
right
01:14
<aroben>
eseidel: what issues do you see that are unique to postMessage?
01:17
<eseidel>
you reply to a postMessage w/ another yes?
01:17
<eseidel>
if that's the case, than any time you reply, you're likely to get hung
01:18
<eseidel>
I guess we could add a timeout
01:18
<eseidel>
but postMessage feels very UDP to me
01:18
<othermaciej>
I see, that's essentially passing file descriptors over sockets
01:18
<othermaciej>
(in unix systems programming terms)
01:18
<othermaciej>
eseidel: you don't have to reply synchronously, though it does offer the API to do so
01:18
<eseidel>
you don't wait for your UDP response packet (since it returns void) :)
01:19
<othermaciej>
eseidel: you can store the source window instead and message it later at a time of your choosing
01:19
<eseidel>
honestly, the UDP comparison is what sells it for me
01:19
<eseidel>
if postMessage had a return value, I'd understand
01:19
<eseidel>
(and that would force it to be sync)
01:19
<eseidel>
but it's a UDP send call, that's all
01:20
<eseidel>
ok, time to mail the list
01:20
eseidel
just sees postMessage as causing all sorts of deadlocks as-is
01:20
<eseidel>
list-posting post salsa-class however...
01:20
<annevk>
now in general dispatching events is synchronous
01:21
<annevk>
so making it async here doesn't make much sense to me
01:21
<othermaciej>
postMessage is not like UDP
01:21
<othermaciej>
it is not intended to be an unreliable transport
01:21
<othermaciej>
I do think that "message passing" often implies asynchrony
01:23
<annevk>
http://www.w3.org/mid/083D18C6B9B71F4CBCA7B76D97B7483102C681BD21⊙Nwwnmc ... I hope he follows that e-mail with one that addresses the concerns raised.
01:24
<othermaciej>
I'm still waiting for specifics on the security claims
01:24
<Hixie>
you won't get any
01:24
<eseidel>
othermaciej, annevk: so what happens when you message to a frame who is spinning/locked/whatever?
01:24
<othermaciej>
eseidel: in Safari at least, frames do not have separate threads of control
01:24
<othermaciej>
so it is not a meaningful question
01:26
<eseidel>
othermaciej: I guess that's true. gmail runs in the background via settimeout, refresh, etc. but when it hangs, it has already brought down the frame which was trying to talk to it
01:27
<eseidel>
anyway, I've got to run
01:28
eseidel
wonders what other pieces of HTML5 will strike him as broken
01:28
<eseidel>
;)
02:09
<Hixie>
man, why does <object> use data="" instead of src=""
02:09
<Hixie>
that's so dumb
02:11
<othermaciej>
it is
02:12
<othermaciej>
one of the dumber things about it
02:18
<Hixie>
i wonder if we can just get away with not having data-* on <object>
02:18
<Hixie>
or if we should just use another DOM name
02:18
<Hixie>
like anne's "dataset" suggestion
02:22
<othermaciej>
that's not bad
02:22
<othermaciej>
or maybe Data Access for Rich Internet Applications - daria
04:02
<BenMillard>
just to add another syntax suggestion to the custom data brew, how about meta="foo" with meta="bar:baz quux:quuz" for multiple pairs and replace ".data" with ".meta" for the DOM suggestions?
04:11
<BenMillard>
a single custom attribute for all custom data seems neater in some respects than an unrestricted number of custom attributes
04:12
<BenMillard>
the "meta" attribute name sort of complements the <meta> element name and they have a similar purpose
04:13
<BenMillard>
it's slightly fewer keystrokes for authors and slightly fewer bytes to transfer
04:15
<BenMillard>
meta="bar:baz" would be for a value whose key couldn't be determined from the element or context around that element
04:28
<BenMillard>
authors concerned about the accessibility and usability of data in title attributes have discussed solutions on Accessify Forums previously, so I've linked them to the recent IRC logs and am now linking you to them: http://www.accessifyforum.com/viewtopic.php?p=59246#59246
06:45
<Hixie>
i "transclution" a word?
06:45
<Hixie>
the fourth hit on google for that word is an e-mail i wrote
06:45
<Hixie>
and there are less than 100 hits
06:45
<Pavlov>
m-w.com?
06:45
<Hixie>
which leads me to suspect it's not
06:45
<Hixie>
but i don't know what the word would be
06:46
<Hixie>
Pavlov: m-w.com doesn't have it
06:47
<jruderman>
i think the word you are looking for is "transclusion"
06:48
<Hixie>
yes, so it seems
06:48
<Hixie>
thanks
06:48
<Hixie>
doesn't come up in m-w.com either, of course
06:53
<Hixie>
http://wiki.whatwg.org/wiki/New_Vocabularies has some solutions on it now
07:05
<mitsuhiko>
data is a good word
07:12
<mitsuhiko>
i really hope that ie8 is a joke
07:13
<mitsuhiko>
interestinly ie 6 renders better than ie8 when it comes to what we do
08:16
<hsivonen>
Does anyone have experience on how much time it takes to get things running in Amazon EC2?
09:26
<virtuelv>
Hixie: am I missing something, or is an explanation of what the dirty* arguments to putImageData are missing?
09:57
<hsivonen>
http://www.w3.org/TR/2008/WD-wai-aria-primer-20080204/#alookahead
10:00
<zcorpan>
hsivonen: that section confuses me
10:01
<hsivonen>
there's no good reason why certain bits of HTML5 could not be implemented in the same schedule as some bits of ARIA
10:02
<zcorpan>
but the group seems to care more about getting to LC than getting implementations...
10:03
<zcorpan>
(at least that's the impression i get)
10:03
<hsivonen>
do PFWG specs need two interoperable impls to advance to REC?
10:04
<zcorpan>
no idea
10:33
<hsivonen>
aaargh. annoying. The XPath attribute thingy requires quotes where CSS attribute selectors work without quotes
10:44
<hsivonen>
hmm. ../../ doesn't appear to actually work...
12:39
<Lachy>
Hixie, yt?
12:41
<Lachy>
Hixie, can you zip up a copy of acid3 and all support files for me? The last copy you sent me seems to be a bit out of date and some tests aren't working properly.
13:03
<hsivonen>
hmm. I have now sent 44% of all-time messages to public-pfwg-comments
13:07
<zcorpan>
hsivonen: interesting
13:09
<hsivonen>
which makes me wonder where all the comments from accessibility professionals are
13:09
<Philip`>
Maybe some are on non-public lists?
13:14
<hsivonen>
Philip`: surely all the pros aren't members of the secret list. (then there's wai-xtech, but I'm not seeing much spec review in the archives)
13:16
<zcorpan>
i haven't seen much spec review at all
13:16
<zcorpan>
it's also not so easy to review the spec, since it's impossible to do diffs
13:16
hsivonen
is just getting more grumpy when diving deeper into HTML5/ARIA integration issues
13:16
<zcorpan>
(unless you're an editor or in w3c staff or something)
13:17
<zcorpan>
i'll work on writing test cases for aria combined with native html semantics soonish
13:18
<zcorpan>
(i have no idea about pass conditions yet so they'll be demos initially)
13:18
<zcorpan>
i also don't know how to automate aria tests
13:18
<zcorpan>
i've heard it should be possible somehow
13:18
<hsivonen>
I think I'm just going to deploy an iteration of Validator.nu with part of the conformance reqs pulled out of my sleeve
13:19
<zcorpan>
that would be cool
13:19
<zcorpan>
hsivonen: btw, the conformance stuff is updated in the Member-only draft
13:20
<hsivonen>
zcorpan: I had a quick look, and it didn't seem to address major issues. I keep reading the public draft otherwise in order to avoid leaking secrets in the Validator.nu SVN repo
13:21
<hsivonen>
It would be nice if the PFWG made their Editor's Drafts public
13:21
<zcorpan>
i agree
13:21
<zcorpan>
i don't get the impression that anything is intended to be secret
13:22
<zcorpan>
it's just behind closed doors due to cargo cult
13:22
<zcorpan>
or at least that's what i want to believe
13:42
<virtuelv>
for what it's worth, http://labs.opera.com/news/2008/03/28/
13:43
<virtuelv>
100/100, pixel perfect
13:44
Philip`
thinks Acid3's definition of "pass" is too subtle
13:44
<Philip`>
since 100/100 isn't enough to pass, and being pixel-perfect to the reference isn't either
13:45
<virtuelv>
Philip`: thinking of the performance stuff?
13:46
<Philip`>
virtuelv: Yes
13:47
<hsivonen>
Safari's perf side wasn't smooth on my Mac
13:47
<virtuelv>
Philip`: clicking the "A" reveals it, but I agree -- it's too subtle
13:47
<hsivonen>
but it was a lot smoother on reload (but still not smooth)
13:47
<virtuelv>
which, I think is why we added that extra instruction to the build notice
13:48
<Philip`>
Particularly with people competing to be the first to 'pass', it's really not helpful if passing is partly subjective or depends on hardware and network conditions
13:50
<hsivonen>
does Opera on Linux and Windows use the system TTF rasterizer?
13:57
<virtuelv>
hsivonen: I'll have to admit to being a bit unsure about what we do on every platform
13:58
<hsivonen>
virtuelv: ok. (the OS X builds appear to use the same TrueType rendering code path as Firefox 3)
14:42
<BenMillard>
zcorpan, section 6 confuses me as well
14:42
<BenMillard>
the "WCAG 1.0 Style Accessibility" could use a nested list to convey depth
14:43
<BenMillard>
the alt on the image for the + and - symbols could say "Expanded" or "Closed"
14:43
<BenMillard>
position would also be handled by using a nested list, AFAICT
14:45
<BenMillard>
upon expanding part of the tree, the number of items in that branch would become apparent by counting the <li>
14:46
<BenMillard>
you could use title="4 items" when a branch is closed
14:47
<hsivonen>
BenMillard: a screen reader saying "item two of fifteen" is chrome but title='' is content
14:49
<BenMillard>
according to that document: "Description/Help: Limited to HTML 4.0 - Alt Text, title text"
14:55
<Lachy>
othermaciej, does apple have any selectors api test cases to contribute to the test suite, which I need to start creating very soon?
15:00
<BenMillard>
hsivonen, I agree that title="Item x of y" on every item is unnecessary.
15:01
<BenMillard>
I think I've misunderstood their example. I thought aria-setsize="4" meant how many items were contained by the current item, but it actually seems to be the number of items in that level of the tree
15:01
<BenMillard>
so you don't need title for that
15:03
<hsivonen>
BenMillard: my point is that "item x of y" should probably be read in the chrome voice in screen readers that distinguish between content and chrome (like VoiceOver)
15:04
<hsivonen>
OTOH, untrusted content shouldn't be able to feed arbitrary text to the chrome voice for security reasons
15:05
<BenMillard>
if it were marked up as a nested list, announcements like that can be figured out by the screen reader without authors adding additional information
15:05
<BenMillard>
(or marked up as a new HTML form control)
15:06
<BenMillard>
aria-posinset and aria=setsize are made redundant by the structure of the markup if sensible elements are used as the basis, afaict
15:08
<hsivonen>
BenMillard: to me, aria-posinset and aria-setsize seem mostly redundant with native semantics, yes
15:08
<hsivonen>
BenMillard: if you are browsing a huge webmail inbox, it could make sense to allow a grid widget declare its starting offset, though
15:18
<BenMillard>
hsivonen, that makes sense. information like that can help everyone, so it might even be visible: "Messages 200-250 of 1,000"
15:19
<hsivonen>
BenMillard: I suggest posting your points about the setsize/posinset stuff to public-pfwg-comments
15:20
<BenMillard>
okies, I'll do that now
16:10
<othermaciej>
Lachy: I'm sure we have at least a couple, I'll ask around
16:13
<Lachy>
othermaciej, thanks
17:03
<BenMillard>
I've sent feedback: http://lists.w3.org/Archives/Public/public-pfwg-comments/2008JanMar/0100.html
23:02
<Hixie>
i've updated http://wiki.whatwg.org/wiki/New_Vocabularies with some obvious ideas
23:02
<Hixie>
like embedding MathML and SVG in HTML using extensions to the parser
23:03
<Hixie>
and have marked which requirements get met by those solutions
23:10
<hsivonen>
what's the deal with the XHTML2/Forms WG CSS namespace objection?
23:10
<Hixie>
hm?
23:12
<Hixie>
i wonder if i can get the "Maintainability" requirement for equations by defining enough clever parsing rules, or if that really just makes it worse
23:14
<hsivonen>
http://lists.w3.org/Archives/Public/www-style/2008Mar/0417.html Why are the XHTML2 and Forms WGs suddenly interested in *removing* CSS namespace syntax?
23:15
<hsivonen>
Hixie: my guess is on makes it worse
23:16
<Hixie>
yeah i figure worse too
23:16
<Hixie>
can't see how else to make mathml maintainable, though, and latex in html just has so many other problems...
23:17
<Hixie>
hahaha
23:17
<Hixie>
they're using process issues on the css group
23:17
<Hixie>
that's funny
23:17
<annevk2>
it's annoying
23:17
<annevk>
but not that bad
23:17
<Hixie>
it's easy to ignore :-)
23:17
<hsivonen>
but why are the XHTML2 and Forms WGs suddenly so interested in backwards compat axioms?
23:18
<hsivonen>
in a CSS feature that was added for the kinds of languages those WGs are specifying
23:18
<Hixie>
because they got the wrong lesson from html5
23:19
<Hixie>
same way that the css group thinks that being transparent is the lesson one should learn from html5
23:19
<hsivonen>
Hixie: surely being transparent is *one* of the lessons?
23:20
<Hixie>
being transparent is a side-effect of soliciting feedback from all sources and attempting to address everyone's feedback
23:38
<jgraham>
Hixie: Re: maths I think this is one case where "the tools will save us" might be the most convincing argument (since *TeX in HTML doesn't make sense and MathML is so impossible to hand author)
23:38
<jgraham>
But it would be interesting to know what the MathWG think
23:38
<hsivonen>
Markdown+iTeX2MML
23:39
<Hixie>
the math wg sent their input to the whatwg list
23:39
<Hixie>
er
23:39
<Hixie>
the public-html list
23:39
<jgraham>
I have vague memories that they had plans for a human-editable MathML syntax
23:39
<jgraham>
Hixie: In more detail than that post
23:40
<jgraham>
(which I thought was very positive as they are willing to have text/html compatible parsing rules for MathML without a fight)
23:40
<Hixie>
http://lists.w3.org/Archives/Public/public-html/2008Mar/0255.html
23:40
<Hixie>
(for those who care)
23:40
<Hixie>
jgraham: yeah
23:40
<Hixie>
jgraham: dunno
23:40
<nickshanks>
does webkit support calc() in CSS?
23:41
<nickshanks>
oops, wrong tab
23:48
<hsivonen>
bed time. nn
23:49
<jgraham>
goodnight