00:01
<Hixie>
othermaciej: should i just go ahead and do it? (i dunno if it's too late to be adding entire untested APIs...)
00:02
<othermaciej>
Hixie: probably should get more opinions than mine, if you think it might be worth doing
00:02
<Hixie>
well it's gonna happen eventually anyway, i was just wondering from an htmlwg perspective
00:02
<othermaciej>
on the one hand, it is late in the game, on the other hand, there is an opportunity to both improve accessibility and improve the API at the same time, so that's cool
00:03
<AryehGregor>
I haven't tried drag-and-drop support, but by all accounts (particular ppk's) it's horrifyingly terrible.
00:04
<TabAtkins>
Hixie: Something like - <p draggable ondragstart="function(e) { e.dataTransfer("datatype", "my data") }">foo</p><div dropzone="datatype">Drop stuff here</div>
00:04
<TabAtkins>
Hixie: And, since the page noticed a @dropzone with "datatype", whenever something is being dragged with that datatype, it *only* fires events at the elements with @dropzone="datatype".
00:04
<TabAtkins>
@dropzone would accept a space-separated list.
00:05
<Hixie>
TabAtkins: i don't understand why we'd only fire events at the declared dropzones that support that type, but i guess that depends on which events you mean
00:06
<TabAtkins>
Actually, now that I read dnd some more, i see that you can just ignore the events on things that aren't dropzones. You have to cancel them for the dropzones.
00:06
<Hixie>
TabAtkins: space-separated list of what? How do you distinguish a binary file of type application/x-foo and a text string of type application/x-foo?
00:06
<TabAtkins>
So, never mind.
00:06
<Hixie>
k
00:06
<Hixie>
i agree with the approach you're describing, i just don't know what the attribute's vlaue should be
00:06
<Hixie>
that's where i got stuck
00:06
<Hixie>
if we can revamp the API at the same time, that makes life easier
00:07
<Hixie>
cos we can just map the new API to the new values
00:07
<Hixie>
e.g. File/Blob-only values can be prefixed with an @
00:07
<Hixie>
or some such
00:17
<TabAtkins>
Hixie: So, what exactly is the "easier" we're looking for here? Less events? Some declarative broadcasting to the user? What kind?
00:29
<TabAtkins>
Hixie: So, I'm thinking my idea was sorta sound, though not with the effect I stated.
00:29
<TabAtkins>
The current drag model is unintuitive, but isn't as bad as I thought in terms of having to deal with stuff (I had thought you had to cancel the events on everything that *wasn't* a drop target).
00:32
<TabAtkins>
So, have @dropzone be an attribute on drop targets. It takes a space-separated list of datatypes (the DataTransfer.types things). Whenever you start a drag, any elements with a @dropzone naming a datatype in the dataTransfer's type list gets a pseudoclass.
00:32
<TabAtkins>
Maybe other stuff?
00:32
<TabAtkins>
Anyway, that makes, for example, highlighting drop zones trivial.
00:33
<TabAtkins>
You could also then query the DOM for the valid dropzones, and expose them via accessibility apis.
00:33
<TabAtkins>
Which I think answers at least one of the requests in the bug.
00:40
<AryehGregor>
gsnedders, did you get my new crash report now?
00:43
<AryehGregor>
I don't really get why Shelley seems to want to ax such large swathes of the spec. She seems to want to just remove as much stuff as possible on principle.
00:45
<TabAtkins>
...yes?
00:46
<TabAtkins>
If you can reverse time such that HTML5 becomes HTML4, maybe you can revive XHTML2 as well.
00:46
<AryehGregor>
Was she a big XHTML2 fan?
00:46
<TabAtkins>
I don't know if she was "big" or not, but I know she loves her some XML.
00:46
<AryehGregor>
I thought pretty much everyone agrees that XHTML2 was a terrible idea.
00:46
<AryehGregor>
Barring the occasional Slashdot idiot.
00:46
<Hixie>
TabAtkins: seems reasonable, yeah. That's the conclusion I'd reached too, though I'm not sure exactly what to do with the attribute's value.
00:46
AryehGregor
shrugs
00:47
<TabAtkins>
AryehGregor: That's about right.
00:47
<variable>
Has there been any discussion on the altlink attribute for <a>?
00:49
<AryehGregor>
altlink?
00:49
<TabAtkins>
AryehGregor: I think it specifies alternative links to use if @href gives 404 or similar.
00:49
<TabAtkins>
Like a client-generated redirect.
00:50
<variable>
that one
00:50
<variable>
it might even provide a list of alternate links
00:50
<variable>
in case of slashdotting, 404, etc
00:51
<jcranmer>
/. the altlink :-)
00:51
<variable>
heh - I'm curious if there has been any discussion about it
00:51
<AryehGregor>
I've never heard of it.
01:14
AryehGregor
discovers that Larry Masinter is on the TAG
01:28
<TabAtkins>
Hixie: I wrote up my suggestion in the @dropzone bug.
01:28
<Hixie>
for the attribute value's syntax?
01:28
<TabAtkins>
Yeah. Loosely, but it's clear.
01:29
<Hixie>
thanks
01:29
<TabAtkins>
I guess to make it work, you'd need to restrict setData so that the format argument had no spaces.
01:31
<Hixie>
we can't limit the types at all, since they can come from the OS, as i understand it
01:31
<Hixie>
re the pseudos, that's something we should do in the CSSWG
01:31
<TabAtkins>
Do they have spaces in them when they come from the OS?
01:31
<TabAtkins>
Yeah, CSSWG should define the pseudo in UI.
01:32
<Hixie>
hyatt and i talked about it years ago and figured we'd need three pseudos, :drop-target, ::drop-target-before, and ::drop-target-after, which would match respectively the current target, and anonymous boxes before and after that element
01:32
<TabAtkins>
Um? Why aren't ::before and ::after suitable?
01:32
<Hixie>
(so that you can make the element slide out of the way and leave room to drop something in)
01:32
<Hixie>
i guess ::before or ::after could work
01:33
<Hixie>
but they're defined as being inside the element
01:33
<Hixie>
so you'd need extra <div>s
01:33
<Hixie>
in many cases
01:33
<TabAtkins>
Oh, you want something outside. Interesting.
01:33
<TabAtkins>
Unfortunate that we don't have ::outside::before.
01:33
<TabAtkins>
I'm not sure what you mean by "make the element slide out of the way".
01:34
<TabAtkins>
Like, just visually? Like when you drag a list element around and the elements slide so there's droppable space between them?
01:34
<Hixie>
presumably in your suggestion the events for the drop zone would get canceled automatically if the attribute was present? (or rather, the default action would only run if the attribute was absent or didn't specify a matching type?)
01:34
<Hixie>
TabAtkins: yes
01:35
<TabAtkins>
Hm. What information is available at the time of the dragenter to tell if you want to accept it? Just the datatype?
01:35
<Hixie>
that and any global information
01:36
<TabAtkins>
Okay. Then yeah, it should auto-cancel or equivalent.
01:36
<Hixie>
i'll look at this more closely in a bit
01:36
<TabAtkins>
Or, hm. I'd still want a way to keep it from dropping, perhaps based on global info, besides just temporarily changing the @dropzone attribute.
01:36
<Hixie>
gotta go right now
01:37
<Hixie>
yeah dunno how to do that
01:37
<Hixie>
that was the other problem i ran into
01:37
<Hixie>
bbiab
01:45
<AryehGregor>
I'm glad to see that a large chunk of the upmodded comments on the latest Slashdot HTML5 discussion are either me or BZ. Maybe there's some hope of educating the masses after all.
01:45
<TabAtkins>
Hixie: Why do you need to distinguish between text and file types?
03:41
<Hixie>
TabAtkins: they use entirely different APIs to get them out
04:11
<Hixie>
you have to love how the TAG's e-mail on ISSUE-41 cites exactly zero technical arguments
04:11
<Hixie>
(and thus by definition will just be ignored by the chairs)
04:59
<TabAtkins_>
Hrm. Someone wanna help me with EventSource in Chrome?
05:00
<TabAtkins_>
I'm using this code: http://www.xanthir.com/test.txt
05:00
<TabAtkins_>
But when I load the page, I get a "Failed to load resource" error.
05:05
<TabAtkins_>
I'm responding to the eventsource url correctly, with a 200 OK and text/event-stream content-type, so basically I'm confused.
05:21
<zcorpan>
hsivonen: in ie6 (and probably ie7 and ie8), <p><hr> end up with the hr inside p, but <p>foo<hr> the hr ends up as a sibling
06:25
<hsivonen>
zcorpan: interesting. not something I'd jump to cloning, though
06:28
<zcorpan>
Hixie: ftr, i'm not using the zcorpan⊙hc bugzilla account. apparently it's not possible to remove it
06:29
<zcorpan>
Hixie: i use the simonp⊙oc bugzilla account
06:48
<wycats>
Hixie: wtf is up with the W3C comments about HTML5 today?
06:49
<wycats>
(I need to add this room to my auto-join again -- got a new laptop and totally forgot :/ )
06:50
<zcorpan>
so, the tag position is (1) "we like distributed extensibility" and (2) "we'd like html5 to do it "like SVG""
06:51
<zcorpan>
without saying why
06:51
<othermaciej>
wycats: which comments do you have in mind?
06:52
<wycats>
othermaciej: "HTML5 is not ready"
08:20
<asmodai>
heh
08:21
<asmodai>
so Firefox prompts me to install 3.6.11 - after that you get redirected to the whatsnew page for .11 and it shows you:
08:21
<asmodai>
You’re running an unknown version of Firefox.
08:38
<annevk>
XHTML2 WG still going strong? http://www.w3.org/News/2010.html#entry-8922
08:38
<annevk>
(I guess they'll just publish whatever is left, but still kind of weird given how long ago they were supposed to be closed down...)
08:40
<asmodai>
annevk: maybe one rebel who keeps up truckin'
08:57
<annevk>
wow that TAG email is meaningless
08:58
<annevk>
might as well have said +1 to "like SVG"
08:59
<othermaciej>
wycats: I hope PLH was giving his own opinions there (and perhaps was selectively quoted) rather than presenting an official W3C position
09:00
<othermaciej>
since W3C management has been pushing me and the other co-chairs to get HTML5 to Last Call as soon as possible, and it would be weird in light of that to present it as highly unready and totally lacking in interoperability
09:00
<othermaciej>
annevk: I questioned their last publication request when it went by on the chairs list
09:00
<othermaciej>
Ian Jacobs said that they'd been given a number of extensions for their existence, which unfortunately were not publicly announced
09:01
<othermaciej>
for those with W3C Member access: http://lists.w3.org/Archives/Member/chairs/2010JulSep/0050.html
09:01
<annevk>
interesting
09:39
<zcorpan>
othermaciej: the difference between personal opinion from a w3c rep and "the w3c" is non-existent in article headlines
10:08
<jgraham>
We should write a spec for the IE selectNodes and selectSingleNode XPath APIs. As far as I can tell they are orders of magnitude better than document.evaluate (not hard)
10:08
jgraham
dons flameproof suit for mentioning XPath
10:11
<jgraham>
Although the way of dealing with namespaces look odd
10:11
jgraham
wonders what Opera does here
10:12
<jgraham>
(odd but less insane than the "pass an arbitary function in" stuff from DOM3 XPath)
10:13
<zcorpan>
jgraham: go ahead (and write the spec)
10:14
<jgraham>
Yeah, that's the problem
10:15
<annevk>
W3C twittered they encourage people to use HTML5 again btw
10:22
<jgraham>
So now we will get headlines like "W3C Confused About HTML5 Readiness"
10:26
<zcorpan>
"Confused about HTML5? Well you should be!"
10:41
<zcorpan>
annevk: yay for domcore fpwd :) now let's hope the attr idea works out
10:43
<annevk>
heh yeah; sicking seems keen on experimenting post Firefox 4
10:44
<annevk>
I now realize what I forgot in my blog post
10:44
<annevk>
history + acknowledgments
10:44
<annevk>
guess i'll write that up separately
11:05
<jarib>
jgraham: wait, IE has native XPath?
11:06
<jgraham>
As part of the MSXml stuff. I don't think you can run it against HTML documents or anything useful like that
11:06
<jgraham>
But I could be wrong
11:06
<jarib>
right, phew
11:07
<jgraham>
Opera has the same APIs that afaik work for anything with a DOM
11:07
<jgraham>
As well as the sort-of-standard-but-not really DOM3 Xpath crap
11:08
<jarib>
webdriver is injecting a JS XPath lib to get document.evaluate in IE, so I'd be surprised if there were alternatives we didn't know about
12:10
<annevk>
http://twitter.com/komputist/status/26736239613 -- I give up
12:12
<jgraham>
I gave up a long time ago
12:13
<annevk>
I'm not yet old and wise
12:13
<jgraham>
Nor am I
12:14
<jgraham>
Just old
12:19
<jgraham>
http://www.whatwg.org/specs/web-apps/current-work/complete.html#overview-of-the-parsing-model the example there should say "script end tag token", right?
12:20
<zcorpan>
yes
12:22
<jgraham>
Bugged
14:45
<sideshow>
so do other people observe that in Chrome, WBS pages seem to be getting some kind of misguided over-aggressive caching that causes them to appear to never change?
14:46
<sideshow>
e.g., if I look at the issue-41 poll results in Chrome, it shows that Aryeh is the only person who's responded
14:50
<aja>
anyone from CSS WG here?
14:51
<aja>
just noticed that Planet CSS is borken....it's giving 403 errors
14:51
<aja>
http://www.w3.org/Style/CSS/Planet/
14:52
<MikeSmith>
hmm, I guess shift+command+R is the way to cache bust in Chrome
14:54
<MikeSmith>
http://www.w3.org/Style/CSS/Planet/ indeed does seem to be broken
15:06
<hsivonen>
MikeSmith: the over-aggressive caching happens in Firefox/Minefield, too
15:07
hsivonen
uses Midori to read wbs response pages
15:08
<MikeSmith>
hsivonen: guess I must have already monkeyed with my caching in Mozilla to disable it
15:09
<MikeSmith>
I know I did that it in WebKit nightly already
15:09
<MikeSmith>
turned caching off completely
15:09
<MikeSmith>
anyway, systems team really need to fix that
15:10
<MikeSmith>
but I don't see the problem in Opera
15:11
<MikeSmith>
maybe I changed something tehre too
15:11
<MikeSmith>
if I did, I can't remember
16:58
<TabAtkins>
Hixie: Right, but do you need to distinguish them at accept time? You can detect differences and decide how to extract it once you've accepted the drop.
16:59
<TabAtkins>
Hixie: Is there a good reason to accept a text type with some particular datatype, but not a file type with the same datatype?
17:13
<jgraham>
So external scripts inserted via appendChild are racy, per spec, right? Or am I missing something?
18:51
Ms2ger
just noticed that Web DOM Core was published along with XHTML 1.1
18:53
<gsnedders>
Thoughts on hsivonen's email, as asked at the end? That's a long email.
19:38
<Hixie>
TabAtkins: i think it's more likely the other way around... you don't want to accept image/png if it's not a File, for instance; another site shouldn't be able to set some string to type image/png and have you crash when you assume it's available as a file
19:40
<TabAtkins>
Hixie: But that'll happen *without* dropzone too, if you're not writing your code robustly. If I don't sniff the .types for "Files", then I'll dumbly accept a random string that has a datatype of "image/png".
19:40
<Hixie>
well right now, you won't get image/png if the user dragged a file
19:40
<Hixie>
you'll only get Files
19:40
<Hixie>
that's one of the things that gets fixed by that API i was talking to maciej about
19:41
<TabAtkins>
That doesn't seem to match the spec - doesn't it say that files should get a platform-specified datatype, converted to a mimetype by the UA?
19:44
<Hixie>
does it?
19:44
TabAtkins
goes to check his memory.
19:45
<Ms2ger>
Hixie, is the file you feed to anolis for the W3C ED available?
19:45
<TabAtkins>
user agents must similarly use platform-specific conventions when deciding on what drag feedback to use.
19:45
<TabAtkins>
All the format strings must be converted to ASCII lowercase. If the platform conventions do not use MIME types to label the dragged data, the user agent must map the types to MIME types first.
19:46
<TabAtkins>
http://www.whatwg.org/specs/web-apps/current-work/complete/dnd.html#when-the-drag-and-drop-operation-starts-or-ends-in-another-application
19:46
<Hixie>
Ms2ger: it's briefly available each time i generate the spec
19:46
<Hixie>
Ms2ger: do you need a copy? what's up?
19:47
<Hixie>
TabAtkins: that's referring to raw data, not files
19:47
<Hixie>
i'll be the first to admit that section's a mess, btw
19:48
<Ms2ger>
Hmm, do you hardcode that it's an editor's draft?
19:49
<TabAtkins>
So, I'm wondering what the problem is. If files don't currently broadcast their datatype, then there's no risk of expecting a file and getting a string, right?
19:49
<gsnedders>
Ms2ger: Yeah, [Status] isn't used at all in HTML5
19:49
<Ms2ger>
Oh
19:50
<Hixie>
Ms2ger: it's one of the inputs i give to my script
19:51
<Hixie>
TabAtkins: as part of this attribute, i want to provide a new API that fixes all these problems... one of the parts of that is that files' types get exposes
19:51
<Hixie>
exposed
19:51
<TabAtkins>
Okay, I didn't realize there was more being talked about here than just what was in the spec.
19:51
<TabAtkins>
Then you probably just want some way to say "this datatype is for a file" or "this datatype is for a string". Like prefixing with an @ or something for files.
19:52
<TabAtkins>
dropzone="@image/png" would only accept files of type image/png.
19:52
<Hixie>
yeah, agreed. that's what i was suggesting yesterday.
19:53
<TabAtkins>
Ah, okay. Then yay for synchrony!
19:53
<Hixie>
i wonder how many of these "kinds" we'll have
19:54
<TabAtkins>
Surely not more than the number of useful non-alphanum characters we have at our disposal.
19:55
<TabAtkins>
Alternately - use a "magic type" to switch the entire dropzone to only accepting files or whatever - if "Files" is present in the dropzone list, all the other datatypes are assumed to be file types, not string types.
19:55
<TabAtkins>
I suspect that may be less useful, though. I dunno.
19:57
<Hixie>
yeah, i dunno
19:58
<Hixie>
there's also the weird situation of the types not necessarily being MIME types... not sure what we're doing with that currently, i'd have to check the spec
19:58
<Hixie>
but some platforms just allow arbitrary strings (including spaces?) as the type names
19:58
<Hixie>
which makes it difficult to do a space-separated or @-prefixed list
19:58
<Hixie>
i'll have to study this more carefully
20:01
<TabAtkins>
>_<
20:01
<TabAtkins>
Convert spaces to _. Done.
20:01
<TabAtkins>
Alternately... Escape?
20:01
<Hixie>
theoretically, how would you distinguish "text_" and "text "? (hopefully it'll never matter of course)
20:02
<Hixie>
yeah, escaping is the obvious solution if it's a real problem
20:02
<Hixie>
but that's a huge pain
20:50
jgraham
reads hsivonen's email (asking about the same thing before was coincidence)
20:58
<jgraham>
Doesn't it make more sense to make preserving order the default
20:58
<jgraham>
and opt-in to racy behaviour?
20:59
jgraham
reads again
21:03
<jgraham>
Specifically the script-inserted exteral scripts blocking parser-inserted scripts part
21:03
<jgraham>
Although I suppose the percieved problem with the safe solution is that it will not be accepted by WebKit/IE
21:05
gsnedders
would rather have non-racy behaviour
21:07
<othermaciej>
jgraham: I suspect in most cases you don't actually want order dependencies
21:08
<othermaciej>
sometimes you do have a specific order dependency and want it respected, but in other cases this just causes a needless perf hit
21:20
<jgraham>
othermaciej: Right, but in general it seems safer to have lower-performance safe behaviour and let people opt in to higher performance less safe behaviour
21:20
<jgraham>
Rather than the reverse
21:20
<hober>
agreed
21:20
<hober>
(declaim (optimize (speed 3) (safety 0))) :)
21:21
gsnedders
has seen sites depend upon the exact perf of browsers before
21:44
AryehGregor
raises an eyebrow at WHATWG
21:45
<othermaciej>
jgraham: I'm not sure the tradeoffs are quite so completely obvious
21:45
<othermaciej>
you could take it to an extreme and say processing of *all* resources is constrained by order of load requests
21:46
<othermaciej>
even, say, images, or stylesheets
21:46
<AryehGregor>
I've adopted a new policy to combat procrastination: for all the stuff I check every day (like e-mail and feeds), I will only check it once, except for my priority inbox.
21:46
<AryehGregor>
Now I have to find something to actually do for the next hour.
21:46
<othermaciej>
but that would probably be too extreme, given the risk caused by the "raciness" vs. the huge performance cost
21:47
<gsnedders>
AryehGregor: I decreased frequency of checking emails to once per hour, and noticed a larger perf increase. And nowadays I mostly ignore specs email.
21:47
<variable>
AryehGregor, smart idea. too bad I'll probably put off implementing it ;-)
21:48
<variable>
seriously - I should implement the same policy :-\
21:53
<jgraham>
AryehGregor: You should spend it on irc instead
21:54
<TabAtkins>
hober: Hey, don't knock DECLAIM.
21:54
<TabAtkins>
Saves me necessary execution time when doing Project Euler.
21:54
<jgraham>
othermaciej: Indeed. But it at least seems somewhat more likely that scripts will depend on other scripts since there are already hacks in the platform to wait for stylesheets and so on
21:55
<annevk>
this two-day break from normal work was great
21:56
<gsnedders>
annevk: meaning you spend the entire time doing site compat on porn sites?
21:56
<gsnedders>
*spent
21:57
jgraham
wonders where that came from
21:57
<gsnedders>
An IM conversation I'm in the middle of.
21:58
<othermaciej>
jgraham: most dynamically inserted scripts are either a bad way of loading JSON (and so likely to have no dependencies on anything) or are library modules that don't do anything substantive other than declare some functions (and so also are unlikely to have order dependencies)
21:58
<gsnedders>
I guess that's not useful to reference.
21:58
<annevk>
that is what I was taking a break from
21:58
<annevk>
;p
21:58
<gsnedders>
:P
21:58
annevk
was at a web developer conference
21:58
<annevk>
see da twitters
21:58
<gsnedders>
annevk: I saw. :) Håkon still sitting down? You haven't said he's stood up yet!
21:59
<annevk>
we had to carry him out to the airport
22:03
<jgraham>
othermaciej: It's not clear that "most" is the problem. The problem is if *some* scripts react badly to out-of-order loading
22:04
<AryehGregor>
othermaciej, could you look at http://www.w3.org/Bugs/Public/show_bug.cgi?id=10355 ? In particular, is it true that WebKit only supports two reflected IDL attributes on marquee; and if so, would it be interested in supporting more to match IE/Gecko/Opera?
22:05
<annevk>
would be more interesting to support less :)
22:05
<othermaciej>
jgraham: in practice this only seems to arise in cases where libraries are used to try to explcitly support it
22:05
<othermaciej>
i.e. the current de facto model on the Web is to opt in to ordering
22:30
<hober>
TabAtkins: I'm not knocking DECLAIM; I'm a big fan
22:31
<TabAtkins>
Excellent.
23:03
<TabAtkins>
Oh god. AryehGregor, I apologize unofficially for the UI for <input type=date>. How did someone honestly think this was an improvement over no UI?
23:10
<TabAtkins>
Dang, I didn't realize <progress> and <meter> landed in Chrome 6. They've got pretty decent UI!
23:15
<hober>
yeah, looking nice
23:30
<smaug____>
huh, progress looks pretty terrible
23:30
<TabAtkins>
Really? I think it's just fine.
23:30
<TabAtkins>
indeterminate especially
23:31
<smaug____>
though Chrome 6 doesn't really use right theming on Linux anyway
23:31
<smaug____>
(does it use any native theming on linux)
23:39
<TabAtkins>
smaug____: Oh, wait, never mind. It does suck on Linux. It's pretty on Windows, though.