00:01
<TabAtkins>
jdaggett: Mind if I clean up the Fonts event stuff to hook into DOM Core terminology and whatnot? No significant changes from your perspective, just synchronization and tightening of some definitions.
00:02
<Hixie>
TabAtkins: (plus scoped fonts for <style scoped>, though i dunno how to even do that given the cascade)
00:03
<TabAtkins>
I don't think impls accept that <style scoped> scopes resource loads like that.
00:03
<Hixie>
TabAtkins: right now the spec doesn't have a conceptual list at all, as far as i can tell, it's all implied that hte fonts are reset on page load, etc
00:04
<TabAtkins>
Hixie: Yeah, there's no concept of it yet. Just saying that it's an easy thing to define.
00:05
<Hixie>
agreed
00:07
<Hixie>
(<style scoped> has to have some solution for font scoping, otherwise two different scoped blocks defining the same font differently would conflict)
00:07
<TabAtkins>
Yes. I think that problem is what impls plan to stick themselves with.
00:08
<Hixie>
it's not a problem for impls, it's a problem for syndicators using scopes styles.
00:08
<TabAtkins>
Sure. But it's impls that decide whether they'll do anything about it. ^_^
00:09
<Hixie>
it would be sad if we didn't find a solution, but yeah
00:09
<Hixie>
wouldn't be the first time we screwed authors, certainly :-)
00:09
<TabAtkins>
Same thing with all the name-defining at-rules, like @counter-style.
00:11
<Hixie>
yeah
00:11
<Hixie>
we really should find a solution
00:11
<Hixie>
maybe it's as "simple" as automatically munging names to be prefixed with the block's name or something?
00:15
<Hixie>
in other news, rewriting the entire canvas section to compatible with workers is about as mind-numbing as rewriting a library to be thread-safe
00:19
<Hixie>
oh.... bummer.
00:20
<TabAtkins>
"The block's name"? There is no such thing.
00:20
<Hixie>
addHitRegion is going to be exciting
00:20
<Hixie>
TabAtkins: an anonymous one
00:20
<Hixie>
TabAtkins: not one the author can see
00:20
<TabAtkins>
Oh, and then rewrite all references inside the stylesheet as well?
00:20
<TabAtkins>
Contextually?
00:20
<Hixie>
(s/anonymous/internal/)
00:20
<Hixie>
something like that
00:20
<TabAtkins>
That sounds unfun.
00:21
<Hixie>
yeah
00:21
<Hixie>
hrm
00:21
<Hixie>
really no idea how to do addHitRegion()
00:22
<Hixie>
oh i guess it's no so bad
00:27
<jamesr>
Hixie, i don't think you could call addHitRegion() from the worker anyway, since that's about interactivity with the DOM and you can't really manipulate interactive things from a worker
00:27
<Hixie>
you can, it just ends up pushing the data to the main thread
00:27
<Hixie>
the only thing you can't do is use the 'control' thing
00:29
<Hixie>
TabAtkins: can you think of any way in which changing the directionality would affect the metrics measured in a canvas TextMetrics object?
00:29
<Hixie>
bounding box, positions of baselines, advance width...
00:29
<Hixie>
i guess if a ( and a ) have different dimensions
00:29
<Hixie>
hmm
00:30
<zewt>
kerning differences?
00:31
<Hixie>
hm yeah I guess "+-" could kern differently than "-+"
00:31
<zewt>
oh, i was thinking horizontal vs. vertical, but i guess that applies to rtl too
00:32
<Hixie>
oh i'm not talking about the writing direction, just the directionality
00:32
<zewt>
yeah
00:32
<Hixie>
canvas doesn't do vertical text yet
00:53
<jdaggett>
TabAtkins: can you send me a patch with whatever changes you want to make?
00:54
<jdaggett>
TabAtkins: then I'll move it back into the src
08:00
<zcorpan>
hsivonen: https://twitter.com/builtwith/status/268951926085910528 https://twitter.com/W3Techs/status/268835927500148737 we'll see if they publish correct reports next
08:15
<hsivonen>
zcorpan: thanks
08:17
<zcorpan>
hsivonen: do you have any ideas for how to approach http://bugzilla.validator.nu/show_bug.cgi?id=945#c7 ?
08:30
<hsivonen>
zcorpan: the datatype back end can report which character was the problem. I don’t recall how well (if at all) the front end handles that data
08:30
<hsivonen>
zcorpan: but I expect it to not handle line breaks in the attribute value very well
08:35
<zcorpan>
hsivonen: ok
08:45
<hsivonen>
zcorpan: yeah, the front end doesn’t handle the error index given by the back end: http://html5.validator.nu/?doc=data%3Atext%2Fhtml%3Bcharset%3Dutf-8%2C%3C!DOCTYPE+html%3E%3Ctitle%3E%3C%2Ftitle%3E%3Cinput+type%3Drange+step%3D0.k5%3E
09:18
<MikeSmith>
zcorpan: yeah I think this is a general limitation for all the attribute-microsyntax checking
09:18
<zcorpan>
yeah :-(
09:18
<MikeSmith>
maybe can try to figure out some clever way to get past it
09:19
<zcorpan>
can the front end be fixed to use the position that the back end reports?
09:20
<MikeSmith>
possibly
09:21
<MikeSmith>
though even if we do that it's not clear to me how it's going to handle line breaks in attribute values
09:22
<MikeSmith>
anyway I'll mess around with it a bit
09:51
<MikeSmith>
zcorpan: one thing I guess we could have it do it is, emit the entire current command expression it was parsing when it found the error
09:52
<MikeSmith>
just one command along with all its arguments
09:53
<MikeSmith>
do you think would that be useful enough that I should bother to try adding it?
09:54
<zcorpan>
MikeSmith: giving a small context around the error seems good enough to be able to locate the error
09:55
<MikeSmith>
OK
09:56
<zcorpan>
MikeSmith: the message doesn't trim the attribute value, but the source extract does. that seems like a bug
09:57
<zcorpan>
MikeSmith: maybe the message should omit the value completely if it's longer than some value
09:57
<MikeSmith>
yeah
09:58
<MikeSmith>
that would not be too hard to change
09:58
<MikeSmith>
so I wonder how long is too long
10:01
<MikeSmith>
300 characters?
10:01
<MikeSmith>
200?
10:08
<MikeSmith>
zcorpan: btw the attribute value doesn't get trimmed in the source extract either, does it?
10:09
<zcorpan>
my test case in the bug seems to get trimmed in the source extract
10:10
<MikeSmith>
strange
10:10
<MikeSmith>
it doesn't for me
10:10
<MikeSmith>
can you make a screen shot?
10:11
<MikeSmith>
the extract I'm seeing has the whole element emitted and highlighted in red
10:11
Ms2ger
mumbles something about CSS specs
10:11
<zcorpan>
um. now it doesn't
10:13
<MikeSmith>
zcorpan: ok
10:14
<zcorpan>
MikeSmith: if i use <img src="aaaaaaaa b"> where there are many "a"s, i get trimmed source extract.
10:14
<zcorpan>
d>↩<body>↩<img src="aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa…aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa b">↩</bod
10:14
<MikeSmith>
oh
10:14
<zcorpan>
i thought i saw that for d="" previously as well
10:15
<MikeSmith>
I'll take a look at that part of the message code
10:16
<MikeSmith>
for the riduculously-long case I can have it truncate the value or omit it completely
10:16
<MikeSmith>
in both places
10:17
<zcorpan>
i think i'd prefer omit in the message and truncate in the source extract
10:17
<MikeSmith>
OK
10:18
<MikeSmith>
that makes sense
10:18
<zcorpan>
as for too long, let's go with 200, since that's about as long as the truncated highlighted source extract is
10:19
<MikeSmith>
yeah that's what I was thinking too
10:19
<MikeSmith>
300 is too long
10:35
<kmbarnhart>
Can someone answer a question I have about the WebVTT specification
10:35
<kmbarnhart>
Sepecifically, a number of times the spec states "one or more characters other than" and say line feed and ampersand
10:37
<kmbarnhart>
It seems to me that that means line feed and ampersand are not allowed at all
10:38
<kmbarnhart>
Which would mean that in the cue text cannot have newline or an & without &amp;
10:39
<zcorpan>
kmbarnhart: right, line feed and ampersand are not allowed at all for that particular part of the syntax it's discussing, because they have special meaning and are allowed in other parts of the syntax
10:39
<kmbarnhart>
ok thanks
10:40
<zcorpan>
kmbarnhart: e.g. with the cue data "foo &amp; bar", there is text "foo ", entity "&amp;" and text " bar"
10:40
<zcorpan>
kmbarnhart: ampersand is not allowed in the text parts
10:40
<zcorpan>
does that make sense?
10:41
<kmbarnhart>
yes. I was a bit unclear because ok, however the newline issues is contrary to the non-normative example and is allowed by the parser
10:42
<zcorpan>
can you explain what the newilne issues are?
10:42
<zcorpan>
newline, even
10:42
<kmbarnhart>
The part I'm concerned with is here http://dev.w3.org/html5/webvtt/#webvtt-cue-text-span
10:43
<kmbarnhart>
The example here: http://dev.w3.org/html5/webvtt/#cues-with-multiple-lines
10:43
<zcorpan>
see http://dev.w3.org/html5/webvtt/#webvtt-cue-text
10:44
<zcorpan>
each line is a cue text span
10:45
<kmbarnhart>
ok ic
10:45
<kmbarnhart>
I missed that, thank you
10:45
<zcorpan>
np
10:46
<zcorpan>
(each line in the example in the spec, i mean. a line can have several webvtt components, too.)
14:45
<annevk>
Oh Mozilla killed XML Events. Well Opera, when will you?
14:48
<smaug____>
Does Opera have XML Events?
14:48
<smaug____>
for what?
14:48
<smaug____>
XML Events require special kinds of event listeners
14:49
<smaug____>
xforms elements or such
14:49
<smaug____>
Don't remember how that all worked in X+V
14:52
<darobin>
I didn't know anyone supported XML Events in the first place
14:53
<jgraham>
If you supported XML events you would keep quiet about it too
14:53
<darobin>
yeah you would, nasty buggers to use they were
15:01
<smaug____>
XML Events was the first larger contribution I made to Gecko (2004, I think) . And now I removed it :)
15:28
<annevk>
Pretty sure Opera has it for SVG
15:29
<annevk>
Opera also still has xml:id :-(
16:51
<dglazkov>
good morning, Whatwg!
17:41
<tantek>
good morning dglazkov!
18:18
<TabAtkins>
jdaggett: Looks like Ms2ger handled the rewrite for me. His suggested patch looks good on its face.
18:19
<Ms2ger>
TabAtkins, not to me, half of it needs to be filled in still
18:20
<TabAtkins>
Ms2ger: The part you've filled in looks fine with me skimming it, though.
19:30
<Hixie>
should scrollPathIntoView() in a worker do anything?
19:31
<Hixie>
(i could, in theory, have it remember what it was told to scroll to, and when you call commit(), have it tell teh main thread to actually do that scroll)
19:33
<TabAtkins>
Interesting.
19:33
<TabAtkins>
That sounds useful, actually.
19:34
<TabAtkins>
It would probably be good, as much as possible, to make the deferred canvas act like the original canvas would, just delayed by the commit().
19:35
<Hixie>
ok
19:38
<Hixie>
so when you're doing this worker-to-main-thread thing, i push the bits out when the main thread is in a stable state (between tasks)
19:38
<Hixie>
(though i suppose i could queue a task)
19:39
<Hixie>
anyway. if, for whatever reason, the "worker" is actually script running on the same thread, should commit() switch to being synchronous? i'm thinking not, it should still queue...
19:54
<Hixie>
context.fillRect(0,0,10,10); context.drawImage(context, 20,0); -- how many rectangles?
19:54
<Hixie>
i was planning on making it one rectangle in workers and two for the case where you getContext()
19:55
<Hixie>
(because in the former case you have to commit() and the latter you don't)
19:55
<Hixie>
but i'm thinking maybe it should always be two