00:02
<AryehGregor>
Hixie, I think WeBMartians meant that I made an ad hominem attack on Microsoft, by saying they care about developers more than users.
00:41
<heycam>
jamesr, hi (though busy)
00:41
<jamesr_>
heycam: so it seems the web perf WG has decided that requestanimationframe should be in their charter. the FX TF has also decided that requestanimationframe is in their scope
00:42
<jamesr_>
i'm happy to help get it standardized in some form with either group
00:42
<zewt>
what does that have to do with performance? isn't that basically a vsync feature or something?
00:42
<jamesr_>
zewt: that is somewhat mysterious to me as well
00:42
<zewt>
web perf is profiling stuff, right?
00:43
<jamesr_>
previously, yeah
00:43
<jamesr_>
they've decided to expand their charter
00:43
<zewt>
not too familiar with it, but sounds ... sort of contrived, heh
00:43
<heycam>
i think it's ok to go with web perf
00:44
heycam
must return to svg wg f2f stuff
00:44
<jamesr_>
ok
00:44
<jamesr_>
my main concern with web perf is that nobody from apple is there (i assume that you can keep roc / boris z up to date)
00:45
<roc>
I think it's OK, heycam can represent us there :-)
00:46
<jamesr_>
i guess i could harass smfr into joining the web perf WG
00:50
<AryehGregor>
Google people: are there internal Google aliases that reject all external mail or something? Some people have been sending me stuff CCd to webkit-editing⊙gc, but when I reply all, it bounces.
00:50
<jamesr_>
yes, most internal mailing lists bounce external mail
00:50
<jamesr_>
you shouldn't be CC'd on such lists in general
00:51
<jamesr_>
tell whoever is adding you to the CC list that they are doing it wrong ;)
00:52
<jamesr_>
and should probably be having the whole thread on the appropriate external list
00:53
<AryehGregor>
I told them to repost feedback to whatwg, which they by and large did, so all is well.
00:53
<AryehGregor>
Just wondering.
05:42
<Hixie>
is there a mouse capturing api in the platform?
05:45
<Hixie>
i'm doing stuff using dragging around a canvas, and i can't work out how to not have one of these bugs:
05:45
<Hixie>
1) dragging out, releasing the mouse button, and going back in makes the ui drag things around without a button
05:52
<Hixie>
2) dragging out and back in without releeasing the button loses the drag
07:59
<zcorpan>
Hixie: i think annevk worked around those issues in his canvas paintr
07:59
<zcorpan>
http://annevankesteren.nl/2006/08-paintr21
08:17
<zcorpan>
lol "If your web application fails in browsers with scripting disabled, Jakob Nielsen’s dog will come to your house and shit on your carpet." http://diveintohtml5.org/history.html
08:24
<zcorpan>
boo for using sync xhr
08:43
<zcorpan>
hsivonen, MikeSmith: <meta http-equiv="Content-type" content="text/html; charset=utf-8"><meta charset="utf-8"> validates
08:43
<MikeSmith>
yeah
08:44
<MikeSmith>
validator is currently not doing any check for that at all
08:44
<zcorpan>
known?
08:44
<zcorpan>
oh?
08:44
<MikeSmith>
yeah, I filed a spec bug on this
08:44
<zcorpan>
ok
08:45
<MikeSmith>
I think the spec should just say, "A document must not have more than one character-encoding declaration."
08:45
<MikeSmith>
and leave it at that
08:45
<MikeSmith>
the current spec language doesn't even prohibit multiple <meta http-equiv="Content-type"… instances
08:47
<MikeSmith>
zcorpan: what I think we should implement is: report an error for any character-encoding declaration except the first one (in document order)
08:48
<MikeSmith>
and possibly, also report the location of the first one
08:49
<MikeSmith>
a problem with that is, this part of the code currently has no means for emitting Info-level messages
08:49
<MikeSmith>
it can only emit errors and warnings
08:50
<MikeSmith>
zcorpan: anyway, this is bug : http://www.w3.org/Bugs/Public/show_bug.cgi?id=12054
08:50
<MikeSmith>
of course feel free to comment there if you think anything I said there is wrong, or if you have anything to add
08:52
<zcorpan>
seems fine
08:53
<MikeSmith>
ok
08:53
<karlcow>
Does it send a warning when there are conflicting?
08:54
<karlcow>
s/there/they/
08:54
karlcow
is going to tes
08:54
<karlcow>
test
08:54
karlcow
is in typos day
08:56
<MikeSmith>
karlcow: it doesn't do anything currently
08:57
<karlcow>
the direct input does funny things
09:07
<annevk>
email is down
09:07
<annevk>
bah
09:26
<karlcow>
rha bad network here.
09:27
<karlcow>
MikeSmith: interesting thing with the meta and the validator
09:27
<karlcow>
http://validator.w3.org/check?uri=http%3A%2F%2Fwww.la-grange.net%2F2011%2F03%2F03%2Ftest%2Fvalid-charset-02&charset=%28detect+automatically%29&doctype=Inline&group=0&user-agent=W3C_Validator%2F1.2
09:27
<karlcow>
http://validator.w3.org/check?uri=http%3A%2F%2Fwww.la-grange.net%2F2011%2F03%2F03%2Ftest%2Fvalid-charset&charset=%28detect+automatically%29&doctype=Inline&group=0&user-agent=W3C_Validator%2F1.2
09:28
<karlcow>
the order of the meta is different but it always warn about one being wrong.
09:28
<karlcow>
which is the case
09:28
<karlcow>
but it doesn't warn there are two :) indeed
09:29
<karlcow>
I wonder if browsers pick the first one or the last one in the pile
09:29
<zcorpan>
karlcow: http://www.la-grange.net/2011/03/03/test/valid-charset has charset in http content-type
09:29
<zcorpan>
karlcow: which wins over meta
09:30
<karlcow>
ah interesting making a 3rd test
09:30
<zcorpan>
charset detection has already been researched and specified and implemented
09:31
<karlcow>
zcorpan: I would have thought that the meta/charset would win over meta/http
09:33
<karlcow>
http://validator.w3.org/check?uri=http%3A%2F%2Fwww.la-grange.net%2F2011%2F03%2F03%2Ftest%2Fvalid-charset-04&charset=%28detect+automatically%29&doctype=Inline&group=0&user-agent=W3C_Validator%2F1.2
09:33
<karlcow>
here it says there are two
09:33
<karlcow>
"The character encoding specified in the HTTP header (utf-8) is different from the value in the <meta> element (iso-8859-1, utf-8). I will use the value from the HTTP header (utf-8) for this validation."
09:33
<karlcow>
or more exactly it sees there are two
09:34
<zcorpan>
it doesn't seem to be validator.nu that emits that message
09:41
<karlcow>
nope, probably the additional layer of W3C validator for messages
09:41
<hsivonen>
Is my reading comprehension correct that the Chairs "accepted" http://www.w3.org/html/wg/wiki/ChangeProposals/ARIAinHTML5#Details but excluded more than 50% of its proposed edits?
09:41
<hsivonen>
thus effectively formulating a middle ground that wasn't fully represented by any one CP
09:42
<hsivonen>
might have been easier to document which edits were upheld than to say which edits were excluded
09:42
<annevk>
I should have just stayed asleep
09:42
<annevk>
no email no news no nothing
09:43
<annevk>
although I guess I can concentrate more on this blog post on Web Notifications
09:44
<karlcow>
hsivonen: there's a thread in between sam and steve about it http://lists.w3.org/Archives/Public/public-html/2011Mar/thread.html#msg5
09:44
<hsivonen>
karlcow: yeah, so it seems. thanks
09:45
<zcorpan>
annevk: you mean there's nothing that can procrastinate real work?
09:46
<karlcow>
arf http://www.opera.com/press/releases/2011/03/03/
09:46
<karlcow>
>"I'm very concerned," says Standal. "Seventeen is very young, and I am not sure if, at that age, people are ready to use such an application. It's very fast, you know, and it has a lot of features. I think the download requirement should be at least 18."
09:47
<zcorpan>
heh
09:47
<annevk>
zcorpan, well, trying to clean up my inbox is pretty much my job
09:47
<annevk>
it contains bug reports with specs
09:48
<annevk>
but when I do that I'd like to have up to date information
09:48
hsivonen
wonders how Apple came up with 17
09:49
<hsivonen>
aren't 15, 16, 18 and 21 more common special ages?
09:49
<karlcow>
yes very strange
09:50
<hsivonen>
I wonder how the iUniverse works for young people and people without credit cards
09:50
<jgraham>
floor((15+16+18+21)/4) maybe
09:50
<hsivonen>
I started using Apple products as a minor and without a credit card
09:51
<karlcow>
well I'm an adult (well maybe) and I do not use Mac AppStore specifically because of this credit card policy
09:52
<karlcow>
no shiny new tweetie for me
09:52
<hsivonen>
I was a hassle to get my dad pay for software using his credit cards back when vendors wanted stuff faxed to them
09:52
<hsivonen>
karlcow: don't you have a credit card or do you oppose to registering yours in order to download free stuff?
09:53
<hsivonen>
s/I /It
09:53
<karlcow>
I have credit cards but indeed you are forced to register to download free stuff
09:57
<hsivonen>
I haven't installed any apps from the Mac App Store, either
09:58
<hsivonen>
I'd probably have more credit card activity with Apple if they got rid of the territorial restrictions on movie rentals
09:58
<karlcow>
to their credits, I think the territorial stupidity is coming from the movie industry.
09:59
<hsivonen>
karlcow: that's one possibility.
09:59
<hsivonen>
karlcow: it could also be coming from the laws the Finnish copyright lobby has gotten set up to favor Finnish copyright holders
09:59
<annevk>
yeah, I think it's that each country has their own silly laws
09:59
<hsivonen>
karlcow: or it could come from Apple (incorrectly) thinking that they have to have the movies translated to the local language before they can open the store
10:01
<karlcow>
so many missed opportunities for the movie industry. Which reminds me that I should add that on my next talk on the topic.
10:01
<hsivonen>
but since Apple already sells music in Finland, they have to have had some kind of contact with the local copyright lobby anyway
10:01
<hsivonen>
the book industry is super-sad, too
10:01
<karlcow>
yep
10:02
<hsivonen>
can't get a single e-Ink-based reader device that'd make sense for both English-language and Finnish-language books
10:02
<karlcow>
many people do not understand. At least there are new publishers riding on their own fully digital.
10:02
<hsivonen>
because the broadest selection for English uses Amazon's DRM and the broadest selection for Finnish uses Adobe's DRM
10:02
<zcorpan>
"CDATA sections must start with the character sequence U+003C U+0021 U+005B U+0043 U+0044 U+0041 U+0054 U+0041 U+005B (<![CDATA[)." is easier to read than:
10:02
<zcorpan>
"CDATA sections must start with the character sequence U+003C LESS-THAN SIGN, U+0021 EXCLAMATION MARK, U+005B LEFT SQUARE BRACKET, U+0043 LATIN CAPITAL LETTER C, U+0044 LATIN CAPITAL LETTER D, U+0041 LATIN CAPITAL LETTER A, U+0054 LATIN CAPITAL LETTER T, U+0041 LATIN CAPITAL LETTER A, U+005B LEFT SQUARE BRACKET (<![CDATA[)."
10:03
<hsivonen>
zcorpan: that bit of the spec is putting specifiers ahead of implementors anyway. :-)
10:03
<hsivonen>
in other ways that is
10:04
<karlcow>
"CDATA sections must start with the character sequence "<![CDATA[" (U+003C U+0021 U+005B U+0043 U+0044 U+0041 U+0054 U+0041 U+005B)"
10:04
<karlcow>
even easier to read?
10:04
<zcorpan>
karlcow: yeah maybe
10:05
<hsivonen>
are Julian's CPs about slashes in charset all about the content-type sniffing algorithm?
10:05
<hsivonen>
or does he have CPs that affect HTML parsing?
10:13
<hsivonen>
where can I locate the different proposals that are referred to by proposal numbers in the script execution thread?
10:18
<hsivonen>
oh. there's http://www.w3.org/2002/09/wbs/40318/issue-125-objection-poll/
10:18
<hsivonen>
I guess I need to take the time to object to it.
10:18
<hsivonen>
I'd appreciate it if I could fix bugs instead of spending time to object to these
10:23
<zcorpan>
opera has now removed uts22 charset alias matching, fwiw
10:24
<annevk>
in public builds?
10:25
<zcorpan>
i thought i saw it in a changelog of a snapshot
10:25
<zcorpan>
don't remember which and it seems i can't access the desktop team site right now
10:25
<karlcow>
annevk: http://my.opera.com/desktopteam/blog/2011/02/17/a-first-glimpse-at-barracuda
10:25
<karlcow>
"CORE-21407 remove support for UTS22 §1.4 charset alias matching"
10:26
<annevk>
nice
10:27
<karlcow>
the changelog of Barracuda is quite impressive
10:30
<hsivonen>
responded to http://www.w3.org/2002/09/wbs/40318/issue-125-objection-poll/results
10:39
<karlcow>
http://www.ecyrd.com/timeismoney/
10:39
<karlcow>
for people who loves meetings
10:40
<zcorpan>
annevk: i would imagine showing notification in-window first and having a button on the notification to opt-in to os-level notification for the origin
10:40
<zcorpan>
annevk: and an easy way to opt-out again
10:42
<annevk>
yeah, Hixie proposed that
10:42
<annevk>
well, that was his design
10:42
<annevk>
but apparently people want to know things upfront and such
10:42
<annevk>
but maybe we should just not cater to them?
10:43
<zcorpan>
show a notification upfront? :)
10:43
<annevk>
yeah, but they didn't quite like that either
10:43
<jgraham>
Umm, that sounds quite complex
10:43
<annevk>
I forgot where this was discussed
10:44
<annevk>
and I rather have this be someone else their problem
10:44
<jgraham>
as you can't just hand off to an OS notification mechanism
10:46
<annevk>
rel=pingback unused?
10:46
<annevk>
has this person not heard of WordPress?
11:17
<zcorpan>
apparently
11:20
zcorpan
filed that bug btw
11:29
<hsivonen>
hmm. looks like @opera has "space astronauts" instead of architecture astronauts working on network operations
11:31
<zcorpan>
hmm?
11:31
<benschwarz>
adactio Whatup!
11:32
<adactio>
benschwarz: Nice work on HTML5 for developers — lovely stuff.
11:32
<benschwarz>
adactio—Thanks. I enjoyed seeing your tweet
11:57
<annevk>
hsivonen, what do you mean?
11:59
<hsivonen>
annevk: https://twitter.com/#!/opera/status/43254428245032961
11:59
<annevk>
:)
12:01
<adactio>
I have a niggling doubt about the definition of the the figure element that I'd like to get clarified http://www.whatwg.org/specs/web-apps/current-work/multipage/grouping-content.html#the-figure-element
12:01
<adactio>
"The figure element http://www.whatwg.org/specs/web-apps/current-work/multipage/rendering.html#represents some http://www.whatwg.org/specs/web-apps/current-work/multipage/content-models.html#flow-content, optionally with a caption, that is self-contained and is typically referenced as a single unit from the main flow of the document."
12:02
<adactio>
Is it indeed "the document" that we're talking about or is it actually "containing sectioning content"?
12:02
<adactio>
(I suspect it is indeed the whole document but I just want to make sure)
12:07
<karlcow>
I wonder if a space astronauts is like vocal singer
12:07
<annevk>
adactio, it's the whole document
12:08
<adactio>
annevk: Thanks. That's what I thought.
12:08
<annevk>
adactio, e.g. you might have a bunch of graphs in an appendix referenced from the main text
12:11
<hsivonen>
regarding http://dev.w3.org/html5/status/formal-objection-status.html : Have the FOs to adopting HTML5 as the WGs "basis for review" been withdrawn?
12:12
<annevk>
oh right
12:12
<annevk>
I need to submit a CCP today
12:13
<hsivonen>
annevk: which ISSUE?
12:21
<annevk>
140
12:36
<hsivonen>
I can't work out from Noah's CP if a "Conforming HTML5 Document" is allowed to have stuff in svg or math subtrees
12:37
<hsivonen>
also, I can't work out from Anne's current CCP if a "Conforming Document" is allowed to have stuff in svg or math subtrees
12:37
<hsivonen>
and with both CPs: which stuff?
12:39
<zewt>
zcorpan: seems like it may be important to be able to request permission to show os-level notifications in advance; the "ask on first use" permission model (of eg. geolocation) doesn't really apply, since by the time you need to show a notification, it's too late
12:41
<hsivonen>
does anyone remember what's a good summary of the aria:foo vs. aria-foo episode?
12:41
<zcorpan>
zewt: yeah
12:41
<hsivonen>
use case: explaining why smil:foo is bad for the exact same reasons as aria:foo
12:43
<zewt>
is anything actually using the web permissions spec yet?
12:44
<annevk>
hsivonen, there's probably some emails on www-tag explaining that
12:46
<zcorpan>
http://lists.w3.org/Archives/Public/www-tag/2008Apr/0244.html ?
12:48
<hsivonen>
zcorpan: that one explains that IBM threw its weight behind aria-foo but it doesn't explain the technical problem fully
12:48
<hsivonen>
looking at my own emails to www-tag, I don't see a nice summary among my own emails, either
12:48
<annevk>
http://lists.w3.org/Archives/Public/www-tag/2008Apr/0235.html
12:48
<annevk>
is mine, not very good though
12:48
<annevk>
but highlights some of the technical issues
12:49
<hsivonen>
annevk: thanks. that's the best one so far
12:50
<zewt>
heh the quote in that mail
12:50
<zewt>
doesn't it basically say "if we spend some cost, we can change the cost/benefit ratio"
13:28
<jgraham>
so why doesn't http://software.hixie.ch/utilities/js/live-dom-viewer/saved/858 work in webkit?
13:31
<zewt>
works in chrome
13:34
<zcorpan>
there's something weird with chrome: if i just load a saved livedomviewer, it doesn't run scripts until i edit the script contents in the input
13:35
<zcorpan>
security feature?
13:37
<zewt>
what version? seems normal in 9
13:38
<zewt>
oh, it displays but I don't know if the onload is working
13:38
<zcorpan>
dev 11.0.686.1
13:38
<zewt>
ah yeah contentDocument is undefined
13:40
<zewt>
looks like the data: URL isn't considered same-origin to the page itself
13:41
<zewt>
Unsafe JavaScript attempt to access frame with URL data:text/html,iframe with data: uri from frame with URL http://software.hixie.ch/utilities/js/live-dom-viewer/?... Domains, protocols and ports must match.
13:41
<bga_>
yeah
13:42
<bga_>
in ff all ok
13:42
<bga_>
http: !== data: formally
13:43
<bga_>
zewt document.write is workaround
13:44
<zewt>
oddly doesn't work in FF3 (iframe is undefined); works in 4
13:46
<bga_>
btw this behavior is usefull for sandboxing
13:46
<zewt>
not if it's not consistent, though
13:47
<zewt>
doesn't help sandboxing if it's not sandboxed in FF4
13:47
<bga_>
partial solution for webkit
13:47
<zewt>
i assume chrome is strict about it for some XSS-related reason or something like that
13:49
<bga_>
htmlfile for IE but i havent discovered it deeply
13:50
<jgraham>
Right, I am kind of hoping that that is special magic for data: uris
13:50
<jgraham>
Like Opera and Firefox have
13:50
<jgraham>
Because making data: cross-origin is annoying
13:51
<jgraham>
and document.write doesn't work for my case
14:50
<annevk>
answered 126, left 125 alone (would have had the same arguments as hsivonen)
14:55
<annevk>
Hixie, PDF version link of the spec is broken?
15:10
<webr3>
hsivonen, around?
15:11
<hsivonen>
webr3: yes
15:12
<webr3>
cool, re HTML+RDFa1.1 - would you be happier if there was /no/ mention at all of xmlns in that doc (html host lang for rdfa core)
15:12
<webr3>
or, mentioned only for processors for backwards compat with xhtml+rdfa1.0 with a text/html mime type
15:13
<hsivonen>
webr3: I'd be happier if neither RDFa in HTML nor RDFa in XHTML relied on xmlns:foo
15:14
<hsivonen>
and compat with existing content was achieved by hardwiring property names
15:15
<webr3>
we can't do that though sadly, there are too many to hard wire
15:16
<webr3>
we're stuck int hat we've done everything to get rid of it, but have several million docs out there with it to handle for bc
15:16
<webr3>
ack, moving topic in meeting - will pick up at a later date
15:16
<hsivonen>
webr3: don't those several million all use a handful of vocabs? Facebook, Google, Yahoo!, CC, DC?
15:18
<webr3>
we're moving to hard wire them (well for processors) - but doesn't handle all of them, will only handle some
15:18
<webr3>
and we have in charter the bc thing, so can't avoid
15:19
<webr3>
we've made it so authors can forget movign forward, and we don't mention xmlns, /but/ the extra rules needed to handle bc for text/html are goign to make it prominent in that doc
15:19
<hsivonen>
webr3: charters are set by the people in the WG, so it's not really credible to point to charter as something that forces you
15:19
<webr3>
so 50+ mentions in HTML+RDFa1.1 - and 1 in rdfa core :|
15:22
<Philip`>
Instead of mentioning "xmlns" explicitly, you could discourage people from using it by calling it "U+0078 LATIN SMALL LETTER X, U+006D LATIN SMALL LETTER M, U+006C LATIN SMALL LETTER L, U+006E LATIN SMALL LETTER N, U+0073 LATIN SMALL LETTER S"
15:31
<karlcow>
Philip`: not sure, that would encourage perverts
15:34
<annevk>
zewt, hey, you around
15:34
<annevk>
zewt, http://www.w3.org/2006/webapi/track/issues/35
15:34
<annevk>
it seems form submission works by dispatching a synthetic submit event
15:36
<annevk>
it also seems that Ian described "form submission" in the wrong way
15:36
<annevk>
it is triggered by the submit event, but it is not defined that way
15:38
annevk
filed http://www.w3.org/Bugs/Public/show_bug.cgi?id=12230
15:42
<zewt>
looks like we came to the same conclusions he did
15:43
<annevk>
yeah, that's positive
15:43
<annevk>
I'm leaning towards only initEvent() resetting stuff
15:43
<annevk>
and then making the first argument of initEvent optional too
15:44
<zewt>
i think like you suggested, resetting in init*Event makes most sense--prevents the canceled flag from being set and then the cancel flag from being unset (inconsistent state), and avoids both the problems of early-reset and late-reset during dispatch
15:44
<annevk>
yup
15:44
<annevk>
and when we get fancy-objects I guess we should let initEvent take an object
15:44
<annevk>
instead of a series of arguments
15:45
<annevk>
so you can easily set all the properties
15:45
<annevk>
and then "deprecate" init*Event and not allow them on new event objects
15:45
<zewt>
the "set the properties manually" idea would be nice but that seems way too different from the current model where most event properties are readonly
15:45
<karlcow>
http://robert.accettura.com/blog/2011/03/03/wanted-native-js-encryption/
15:46
<annevk>
zewt, oh yeah, we could do that too still...
15:46
<annevk>
zewt, they'd still be readonly while dispatching
15:46
<zewt>
i think it'd need to ... yeah what you said
15:46
<zewt>
so you can't change them for trusted events, in particular
15:46
<annevk>
hmm, gotta run again; should be online at tenish though
15:46
<zewt>
it's tenish now, here :P
15:46
<zewt>
(okay, elevenish)
15:46
<annevk>
timezones
15:48
<zewt>
should also be defined that non-trusted submit events perform their default action, if it isn't (iirc only click is explicitly mentioned)
16:13
<Moo-_-_>
karlcow: wanted native gzip too :)
16:56
<MikeSmith>
if anybody is interested in doing some book writing on HTML5, please let me know. I have an acquisitions editor I can put you in contact with
16:56
<MikeSmith>
ping me here or DM me or e-mail me at mike⊙wo
17:11
<jgraham>
MikeSmith: I totally read DM as "Dungeon Master"
17:11
<MikeSmith>
heh
17:48
<TabAtkins>
Can <object>'s contents be thought of as accessibility fallback? I'd dismissed that offhand earlier, but anne is bringing it up as a solution.
17:49
bfrohs
thought that's what it was already
17:49
<TabAtkins>
I thought it was functionality fallback, but I may be drawing a distinction that doesn't exist.
17:49
<jgraham>
I thought this distinction was important for <video>?
17:49
<jgraham>
Or was that some other distinction?
17:50
jgraham
tired
17:50
<TabAtkins>
I don't... I don't know. I'm confused now.
17:50
<jgraham>
In any case I disagree with anne I think
17:50
<jgraham>
<object> is worse than <canvas>
17:50
<zewt>
using canvas as an IMG seems questionable, since they're very different things (eg. progressive image rendering is very IMG-specific)
17:51
<jgraham>
zewt: It seems natural to load an image with a declarative src and then draw on it
17:51
<TabAtkins>
jgraham: In that case you can just use <img> and then draw it in, though.
17:51
<zewt>
what would it mean? would it be immutable while it's loading (like an img), and only allow drawing ops after the load finishes?
17:52
<TabAtkins>
I'd be inclined to say "don't touch it until you see a load, because it'll explode".
17:52
<zewt>
(eg. act as if you stuck an img on top of the canvas, then blitted the img onto the canvas and deleted the image at onload)
17:52
<TabAtkins>
Alternately, interacting with it between the time you set @src and when 'load' fires will abort the load.
17:53
<TabAtkins>
That counts as "explode"
17:53
<TabAtkins>
jgraham: If you could comment on why <object> is bad news, that would be nice. I can't formulate a forceful enough objection.
17:55
<jgraham>
TabAtkins: For all the normal reasons that <object> is bad
17:55
<TabAtkins>
I'm insufficiently familiar with why <object> is bad, I've just picked up the religion.
17:55
<zewt>
a lot of stuff currently specced for img would seem to need to be defined for canvas (onload, "fully decodable" and so on)--i'd hope there's a simpler way
17:55
<jgraham>
It tries to do many things and none of them well
17:56
<jgraham>
So you don't get media-specific apis and so on
17:56
<TabAtkins>
zewt: All those are already defined, so it would just be a matter of "do [this]".
17:57
<zewt>
they're defined for images, canvas isn't complicated by them
17:57
<TabAtkins>
jgraham: Is there much in the way of media-specific APIs for <img>? I grok why we did <video>/<audio> instead of extending <object>.
17:57
<TabAtkins>
zewt: I don't understand what distinction you're trying to draw. All the algorithms are well-defined and can be trivially reused.
17:58
<TabAtkins>
None of them depend on the name of the element.
17:59
<zewt>
(let me reread your mail and think for a bit)
17:59
<zewt>
(not that I'll necessarily have any ideas)
17:59
<zewt>
hmm... seems like the obvious, but impossible, solution would be to put the alternate data in the children of img; impossible since img can't have children, of course
18:00
<TabAtkins>
Yes, indeed. That's why I'm turning to the idea of using the *other* image element.
18:00
<zewt>
my first impression would be to define another tag which is completely synonymous with <img> but has children, eg. <cimg>, sharing 99% of the spec and (preferably) the same object type--existing only to syntactically allow it to have descendents
18:01
<TabAtkins>
Didn't somebody suggest doing that for <image>?
18:01
<zewt>
don't know
18:01
<bfrohs>
TabAtkins: I think so
18:14
<zewt>
another possibility would be something like <img src=foo.jpg id=foo><fallback for=foo>...</fallback>, which degrades a bit more nicely
18:15
<TabAtkins>
That seems weirder.
18:15
<zewt>
(modelling label, of course)
18:19
<TabAtkins>
presumably <fallback> wouldn't render if the element that it was for rendered correctly?
18:19
<zewt>
yeah, something like that--something like "hide this element by default, and show it if the target is in an error state"
18:25
<zewt>
degradation would require users to have a script to hide <fallback> when it's not supported, but image loading would be unchanged
18:45
<AryehGregor>
Hixie, could you get me sample execCommand() usage in the wild?
18:45
<AryehGregor>
Probably I should also research some major WYSIWYG implementations.
19:04
<TabAtkins>
Is there any strong technical objection against the 'beforeprint' event that IE fires?
19:06
<TabAtkins>
Google docs changes the rendering of the document for printing if you hit Ctrl+P or select Print from the in-app menu, but it can't tell if you hit Print from the browser's menu.
19:06
<Ms2ger>
media=print
19:07
<TabAtkins>
That doesn't let you run arbitrary JS.
19:07
<TabAtkins>
Unless you can have matchMedia?
19:07
<TabAtkins>
Hmm...
19:07
<Ms2ger>
I was about to mention that
19:12
<TabAtkins>
I suggested that we implement matchMedia instead.
19:13
<TabAtkins>
Now, back to writing an email summarizing a subtle technical issue with floats.
19:14
Ms2ger
screams at CSS2.1
19:14
<TabAtkins>
I'd just scream at floats. They're defined SO BADLY.
19:14
<TabAtkins>
I don't think anyone recognized at the time how complex they are.
19:15
<Ms2ger>
They're not hard, if you use them as intended
19:15
<TabAtkins>
No, they're always hard. You just don't realize it.
19:15
<Ms2ger>
Just don't start making float-based layouts
19:16
<Ms2ger>
(Hi most of the web)
19:16
<TabAtkins>
Perfectly natural-looking use-cases turn out to be subtly odd
19:16
<Ms2ger>
dropcaps!
19:16
<TabAtkins>
For example, I'm almost certain the spec doesn't cover the case exercised by: http://software.hixie.ch/utilities/js/live-dom-viewer/saved/859
19:17
<TabAtkins>
(Remove the / in front of the float to swap behavior.)
19:17
<TabAtkins>
The float drops to the second line, some of the content following it jumps to the previous line, while some of the content that was previous on the same line as it drops to the line below it.
19:17
<TabAtkins>
s/previous/previously/
19:17
<zewt>
sounds like some kind of dance routine
19:17
<Ms2ger>
I'm pretty certain I can find undefined things in most of CSS2.1
19:18
<Ms2ger>
Even if they're not explicitly called out
19:18
<TabAtkins>
Oh, indeed.
19:18
<Ms2ger>
Hence the scream
20:00
AryehGregor
realizes his definition of node position is totally, totally broken
20:26
<emersonvinicius>
Hello everybody
20:27
<emersonvinicius>
sorry my english, but what the diference of spec whatwg instead w3c?
20:28
<Ms2ger>
http://www.whatwg.org/specs/web-apps/current-work/multipage/introduction.html#is-this-html5? explains it better than I ever could
20:28
<emersonvinicius>
Ms2ger: thx
20:40
<AryehGregor>
Note to all who may be concerned: doing simultaneous alert()s in fifty or so iframes at once makes Firefox unhappy.
20:41
<TabAtkins>
Anyone mind if I rewrite every sentence involving floats to a new, subtly backward-incompatible version that actually makes sense?
20:41
<AryehGregor>
In what spec?
20:41
<Ms2ger>
CSS
20:41
<Ms2ger>
And no
20:41
<zewt>
at least in ff4 i'll get to stop killing the entire browser every time I inadvertently cause an alert loop. heh
20:42
<AryehGregor>
Yeah, but that doesn't help if you create fifty alerts from fifty separate iframes. Apparently the "don't show more" box doesn't work then.
20:43
<TabAtkins>
Sounds like a good rick-rolling situation.
20:43
<zewt>
i havn't squinted hard enough at the "pause" part of the spec, but I'd expect an alert (of any modal type) to pause scripts on the whole window, so you can only have one alert at a time even across iframes
20:43
<zewt>
(i know that's the case with a single document; just not sure off-hand how that translates to iframes)
20:45
<zewt>
oh, since it pauses the event loop, and iframes are in the same event loop, it should pause all iframes in the window together
20:51
<zewt>
same bug as async scripts running during prompts, i guess
21:03
<Hixie>
AryehGregor: i dunno how much use it'll be but http://aupeople.com.au/1/aupeople2/read.php?tid=60099 is a random sample of some pages with "execCommand(" and "<script"
21:03
<Hixie>
AryehGregor: i don't have an easy way to get pages that have JS files that use execCommand(, which unfortunately is going to be the majority of your interesting cases
21:03
<Hixie>
AryehGregor: but e.g. i'm sure google docs uses it
21:03
<AryehGregor>
Ah, okay.
21:03
<AryehGregor>
Yeah, it's probably best to look at specific high-profile examples.
21:04
<Hixie>
yeah
21:04
<bfrohs>
Google Docs, AFAIK, does not use execCommand anymore. They ran into a ton of inconsistencies with browsers and I believe they ended up creating their own API for it in javascript (create characters off-screen, take measurements, and move them into place).
21:05
<Hixie>
hm, possible
21:05
<TabAtkins>
Docs does EVERYTHING manually.
21:05
<Hixie>
it'll still be interesting though as it'll be able to show the use cases that are important
21:06
<bfrohs>
Yeah, that's very true.
21:06
<bfrohs>
Gmail/hotmail/etc WYSIWYG editors as well.
21:21
<annevk>
I really like my idea of removing all initXXXEvent in favor of initEvent(object)
21:21
<annevk>
well, removing all future initXXXEvent
21:26
<jamesr__>
i never understood why i have to create a mouse event and then init it before it's really a mouse event
21:26
<zewt>
it just sets the properties, since most of them are read-only
21:27
<annevk>
so yeah, the alternative would be removing all future initXXXEvent in favor of read-write attributes before and after dispatch
21:28
<annevk>
but the initEvent() design is probably slightly better, as that also cleanly resets the "trusted flag" and maybe also the "propagation flag" and such if we decided to do that
21:28
<zewt>
could deal with the getters problem by just defining it as operating on a copy, eg. var copy = {}; for(key in obj) { copy[key] = obj[key]; } but sorting the keys, which would be a simple way of defining how getters are called
21:28
<zewt>
(doesn't work for structured clone, but works here since it's not recursive)
21:28
<annevk>
heycam can sort that out :)
21:29
<annevk>
though it should allow arbitrary nested objects, e.g. for CustomEvent
21:29
<zewt>
having properties be writable when not in dispatch would be nicer, but sort of unusual behavior
21:31
<zewt>
wonder if there's any precedent for that--properties which are only conditionally read-only
21:31
<heycam>
i like the idea of just writable properties on Events
21:31
<heycam>
instead of 50-argument init methods
21:31
<zewt>
either idea gets rid of the horrible init methods
21:31
<othermaciej>
giant init methods suck
21:32
<othermaciej>
otoh it is weird if you can alter the property of an event being dispatched by the UA
21:32
<othermaciej>
I guess altering any properties would have to clear the "trusted" flag
21:32
<zewt>
othermaciej: needs to be read-only when the dispatch flag is set
21:33
<zewt>
that'd be weird--setting a property would effectively imply preventDefault for most eventts, which would be confusing
21:34
<annevk>
heycam, the idea was to have initEvent(object) instead
21:34
<heycam>
annevk, ah. that works too.
21:34
<annevk>
writable attributes does not solve the problem of resetting some flags not exposed by attributes
21:34
<heycam>
right, ok
21:34
<heycam>
i wasn't really considering them
21:35
<annevk>
e.g. the moment you modify you want the "trusted flag" to be false
21:35
<heycam>
aha
21:35
<heycam>
well that you could do
21:35
<annevk>
defining that for each attribute would be kind of weird
21:35
<heycam>
but you couldn't then reset it to true
21:35
<heycam>
actually you never want that do you
21:35
<zewt>
i think the current readonly flags in events should stay readonly (at least during dispatch), not silently clear trusted
21:35
<annevk>
heycam, yeah, it would never be set back to true
21:36
<heycam>
you could just make the writable attributes also set the internal trusted flags, it's easy enough
21:36
heycam
doesn't mind either way
21:36
<annevk>
yeah, it could be done, it just seems less intuitive
21:37
<annevk>
zewt, this is for when a trusted event is dispatched and then modified after dispatch
21:37
<zewt>
annevk: right, I'm saying you shouldn't be able to modify it during dispatch at all
21:38
<annevk>
I don't think I suggested otherwise
21:38
<zewt>
<annevk> e.g. the moment you modify you want the "trusted flag" to be false <- modify when? if you modify outside of dispatch it doesn't matter (trusted will be reset on dispatch anyway)
21:40
<zewt>
i guess i'm assuming the trusted flag is only important during dispatch
21:42
<annevk>
i want isTrusted to always be correct
21:42
<annevk>
anyway, if not for the "trusted flag" it would also go for the "canceled flag"
21:42
<zewt>
then initEvent(object) seems simpler--one entry point to clear the flag
21:42
<annevk>
and for that using initEvent() is also better, as otherwise you'd have to set an attribute to reset it...
21:44
<Hixie>
yeah this CDATA paragraph is ridiculous
21:44
<Hixie>
what was i thinking
21:45
<zewt>
i was assuming there'd still be initEvent() for the base event flags, and only the additional properties would be writable
21:46
<zewt>
eg. e.initEvent("click", true, true); e.offsetX = 10; ...
21:47
<Hixie>
(please don't let events be modified during dispatch)
21:47
Ms2ger
tries to get Hixie back in the business of writing evil test cases
21:54
<annevk>
Hixie, some browsers allow it, but yeah, that's the plan
21:54
<annevk>
zewt, eww
21:57
<zewt>
might be sort of handy that you can initEvent(object) with the same object multiple times (though of course you could implement that manually, copying properties from a saved object)
22:05
<Hixie>
nessy: so, it's likely going to be time to actually do this multitrack thing soon. it seems there's been lots of discussion about it in public-html; is there a summary anywhere of the current thinking?
22:05
<nessy>
Hi Hixie
22:05
<Hixie>
in particular, list of use cases, implementation constraints, anything like that?
22:06
<nessy>
try this: http://www.w3.org/WAI/PF/HTML/wiki/Media_Multitrack_Media_API
22:06
<nessy>
I'll be working with Eric Carlson on a change proposal
22:07
<nessy>
the proposal that Microsoft did is too limited and only restricted to audio tracks
22:08
<Hixie>
that page is great
22:09
<Hixie>
i considered reopening http://www.w3.org/Bugs/Public/show_bug.cgi?id=9452 but i wasn't sure whether that was appropriate or not
22:09
<Hixie>
since we had agreed to assign it to the tf
22:09
<Hixie>
and then it got closed and escalated while assigned to the tf
22:10
<AryehGregor>
Okay, deleteContents() is probably finished. Hurrah.
22:12
<annevk>
zewt, I don't see why we'd disallow that
22:13
<zewt>
just mean, one (minor) handy side-effect of the init(object) approach
22:14
<annevk>
ah k
22:16
<annevk>
once we have that initEvent(object) stuff we could also make that automatic
22:16
<annevk>
document.createEvent(interface, initObject)
22:16
<nessy>
yeah, I think they worked through all the remaining open bugs and just escalated them to issues - though some slipped the cracks - I don't claim to understand
22:16
<annevk>
or maybe new MouseEvent(initObject) at some point... dunno though
22:17
<annevk>
would be a lot of additional things to test for somewhat minor benefit
22:17
<annevk>
pretty easy to make these things way more complicated than they need to be because the problems are so simple
22:17
<nessy>
Hixie: you can either try and work through it now and understand all the options etc, or you can wait until the call for change proposals closes and you have some more thought-through proposals to work through
22:17
<zewt>
doesn't seem very important to have a lot of redundant shortcuts for event init
22:18
<nessy>
the main problem we still haven't looked at is how to actually render all this stuff - and how to address it through CSS
22:18
<nessy>
that's not in that wiki page
22:19
<Hixie>
nessy: well i don't want to step on any toes here. If the bug was filed today, I would probably just address it, but then there wouldn't be pending CPs and whatnot.
22:19
<zewt>
personally I'm not crazy about the separate fireEvent() suggestion for the same reason--doesn't seem worth the redundancy
22:20
<nessy>
Hixie: no toes to step on - people are looking for a solution and are prepared to put in effort to help get there - but we all want the best possible outcome
22:20
<Hixie>
nessy: I don't want to have people feel like they have worked on something and then have me just come along and change it, but i also don't want y'all to work on something if you'd rather i just went ahead and did my thing like i usually do :-)
22:20
<nessy>
it's not a simple task though, so if you have a good solution, it would be awesome to push it!
22:20
<annevk>
zewt, right, I think initEvent(object) is probably the simplest approach to something sane here
22:20
<Ms2ger>
AryehGregor, how about you write tests for intersectsNode, then I can get it into Gecko ;)
22:21
<zewt>
yeah
22:21
<nessy>
how about you do your thing like you usually do, but put it forward as a change proposal rather than pushing it into the spec?
22:21
<Hixie>
nessy: well i don't have anything right now, but i'm sure i would come up with something (maybe good, maybe not) if i were to go through that wiki page
22:21
<annevk>
zewt, the advantages are no crazy initXXXEvent methods, it makes event interfaces extensible, and does not change much about the current model
22:21
<AryehGregor>
Ms2ger, maybe I'll take a look tomorrow.
22:21
<Hixie>
nessy: i hate writing CPs
22:21
<Hixie>
nessy: they take 10 times longer than writing spec text
22:21
<nessy>
just add another option to the wiki page then?
22:21
<nessy>
or engage in the discussion to state which option(s) you prefer?
22:22
<nessy>
it's good to come up to speed so we can shoot down the bad change proposals/options early IMHO
22:22
Ms2ger
will be out for the weekend
22:23
<Hixie>
nessy: what is the direction you and eric are likely to go in for your CP?
22:23
<nessy>
that's the hard question
22:23
<nessy>
Eric submitted option 8 and I think that's the main one he is after
22:24
<nessy>
but we have already found that option 1 will be necessary in addition so we can make adaptive HTTP streaming work
22:24
<Hixie>
looking at this i think a variation on 6 is probably what i'd mainly propose
22:24
<nessy>
and we have found cases where a total replacement of the resource is necessary, which could be done with option 6 or 7
22:24
<Hixie>
anything that messes with <track> is flat-out wrong since video/audio tracks have nothing to do with text tracks, and as something near the top says, anything that messes with <source> is a nightmare in waiting
22:25
<Hixie>
however, 6 wouldn't really help with multiple tracks in one file
22:25
<nessy>
Philip is mostly for 6 or 7
22:26
<Hixie>
i imagine if i were to try to spec this it'd be a hybrid of all of the above plus some more out-of-the-box thinking
22:26
<Hixie>
e.g. with a controller object as the master and all the <Video>s as slaves
22:26
<nessy>
well, the discussion about 8 is on the list - you should engage with Eric/David about that, because they are very keen on 8
22:26
<nessy>
haha, that's option 4 - and very smil-like
22:27
<Hixie>
it's similar to 4
22:27
<Hixie>
and similar to 6
22:27
<Hixie>
probably wouldn't be an element though
22:28
<nessy>
interesting!
22:28
<nessy>
throw it in the wiki and add it as a proposal to the email thread?
22:28
<Hixie>
i'd have to actually spend the time to consider this, i'm just making stuff up here
22:29
<nessy>
that's how we all started...
22:30
<Hixie>
whenever i make proposals without thinking them through, they end up implemented and with legacy content before i have a chance to correct them
22:30
<Hixie>
so...
22:30
<nessy>
I actually quite like option 8 right now - it leaves the main a/v in place as the controlling entity and the others are clearly dependent
22:30
<nessy>
but I'm, keen to see yours!
22:31
<Hixie>
option 8 is a non-starter imho, it messes with <track> in a nonsensical way
22:31
<Hixie>
it's unclear to me whether external sign-language tracks need to be supported without JS, and what the use case is for inline embedded videos being controlled from JS.
22:31
<Hixie>
are there same pages anywhere?
22:31
<Hixie>
sample
22:31
<Hixie>
not same
22:32
<nessy>
let me see - I stared a collection some time ago...
22:33
<nessy>
how's this http://www.w3.org/WAI/PF/HTML/wiki/Media_Multitrack_Media_Rendering
22:33
<nessy>
incomplete, but has some ideas..
22:35
<Hixie>
we really should look at how people do this today in more detail before we write proposals
22:36
<nessy>
mostly sign language is picture-in-picture or two videos next to each other
22:37
<nessy>
there's not that much more options anyway
22:37
<nessy>
I'm thinking about rendering that we could render all the active video tracks in a video into the single viewport and have two rendering modes: a tiling approach and a picture-in-picture approach
22:38
<Hixie>
yeah but do people use two videos? one video with two tracks?
22:38
<Hixie>
do they let people control both videos separately?
22:38
<nessy>
on modern TV all the time
22:38
<Hixie>
etc
22:38
<nessy>
actually, there's a third display mode, usually used on tV
22:39
<nessy>
the main video is large and a couple of other views are tiled down the side
22:39
<nessy>
when you select one of the smaller tiles, it becomes the big one and swaps position with the big one
22:40
<Hixie>
TVs aren't really good as a precedent here since they have very different interaction modalities and constraints
22:40
<TabAtkins>
I've never seen that used on tv.
22:40
<nessy>
my uncle had a TV like that in Germany 5 years ago - I was quite impressed - probably had a set-top box
22:41
<nessy>
typically used for sports games to allow people to have different perspectives
22:41
<zewt>
annevk: re "Conflicts" - it sort of makes sense to require initEvent to be called, since in practice you probably don't want to take the default and dispatch an event with e.type == ""
22:41
<nessy>
there is only a single timeline though and only one control
22:42
<zewt>
(which I assume is why DOM Events disallows e.type == ""; that seems to be what's checked when requiring initEvent to be called, so it doesn't need a separate "has been initialized" flag)
22:43
<annevk>
zewt, what does DOM3Events say should happen when it is not called? what does it say should happen .type returns? or .eventPhase returns? lots of things are undefined
22:43
<annevk>
zewt, I'm not sure putting arbitrary restrictions in place really makes sense
22:43
<zewt>
not sure; I'm looking more at browser behavior than the spec
22:44
<annevk>
I'm also not sure why he thinks event.type can be null in D3E
22:44
<zewt>
it can't be, in either browser I tried
22:44
<annevk>
e.g. D3E says it has to be at least one character and has multiple places say it cannot be an "invalid string"
22:44
<zewt>
(by spec I don't know)
22:45
<zewt>
(predictably, I havn't read the entire DOM Events spec :)
22:45
<annevk>
would be cool to know what Jacob thought about how the spec was written
22:46
<annevk>
but I should sleep some and then read everything again and then reply
22:48
<TabAtkins>
annevk: Re using <object> for inline fallback on images, <object> is horrible for video and audio, but I'm not sure if there are any APIs we have to worry about for static images that <object> won't expose.
22:50
<zewt>
seems generally unpleasant to have people start putting images in object tags...
22:50
<annevk>
using <object> if you want rich fallback is the argument we have been using for years
22:50
<annevk>
if it is invalid I would appreciate to learn why, as it works fine for me
22:51
<nessy>
TabAtkins: I've found something similar: http://gigaom.com/video/att-lets-cubs-fans-choose-their-own-camera-angle/ - though this brings in different channels rather than tracks of the same video
22:52
<annevk>
whatwg.org uses <object> for an image by the way
22:54
<TabAtkins>
nessy: Ah, that's a cool use-case.
22:54
<Hixie>
<object> works pretty well for images
22:56
<jamesr__>
i still think the 'fallback-ish' behavior for <canvas> is bizarre
22:58
<zewt>
FWIW, object doesn't have naturalHeight/naturalWidth, or complete (doesn't define the "available" state like <img> does) ... not necessarily critical for the cases where you want rich fallback (eg. you're probably not blitting it into a canvas)
22:59
<Hixie>
jamesr__: how would you do it instead?
23:00
<zewt>
to sum up my earlier two suggestions, 1: have another tag, eg. <cimg>, which is identical to <img> but allows descendants, and put fallback data there; 2: have a <fallback for=id> tag (like <label>) which is only shown if the given id is not
23:00
<fantasai>
hsivonen: just wanted to note that we just ran into a HTML parsing issue with </! starting a tag sequence
23:00
<fantasai>
hsivonen: in case that's an interesting test case for you
23:01
<annevk>
zewt, neither is backwards compatible
23:01
<zewt>
the <canvas> suggestion isn't, either
23:02
<annevk>
well I'm not really considering that one either
23:02
<zewt>
<fallback> can be made to degrade sensibly with a little scripting help, at least
23:02
<annevk>
with scripting everything can be made to work...
23:02
<zewt>
(eg. check if HTMLFallbackElement exists; if not, add fallback { display: none; })
23:02
<nessy>
TabAtkins: here's one with multiple angles on a single game http://www.sportsgeekery.com/2010/10/28/watch-tnts-nba-games-online-with-overtime-extra/
23:03
<annevk>
lets all look at what we can safely remove instead :)
23:18
<nessy>
Hixie: I added a few more examples to http://www.w3.org/WAI/PF/HTML/wiki/Media_Multitrack_Media_Rendering
23:20
<Hixie>
i don't understand the third one down
23:20
<Hixie>
there only seems to be one video on the page mentioned by the blog post cited on that page, the sign-language one
23:22
<Hixie>
the firstone seems to just be single standable videos with one track
23:23
<Hixie>
the second one seems to just be two unrelated videos rendered on top of each other
23:24
<Hixie>
for the fourth one, the olympics thing, it's unclear to me what the requirements are. It looks like just four unsynchronised live streams, which we can do today. If they're to be synchronised, I guess that's more like proposals 4/6 -- separate <video>s that are slaved to each other.
23:25
<Hixie>
unclear to me what the 5th is showing.
23:25
<Hixie>
or the 6th
23:26
<Hixie>
the 7th is burnt-in.
23:26
<Hixie>
presumably pretty much whatever we do will support the equivalent rendering with two tracks, so long as you can sync it somehow.
23:28
<Hixie>
8th is the same as the 2nd, it seems
23:29
<Hixie>
dunno what the 9th is.
23:29
<Hixie>
10th is like the 7th.
23:29
<Hixie>
11th too
23:31
<Hixie>
12th is like the 4th, assuming they're synchronised. four streams slaved to each other with one set of controls.
23:34
<Hixie>
13th is weird, looks like multiple unsynchronised video streams in some incarnations, and more like the slaved case in others.
23:35
<Hixie>
the last one is more like the olymics thing again
23:35
<Hixie>
so in conclusion it looks like we have the following kinds of use cases here:
23:35
<Hixie>
1. unsynchronised video streams next to each other (possible today with <video>)
23:36
<Hixie>
2. an additional track, either in a separate file or in the same file, providing a sign-language overlay, which must be synchronised to the primary track, and presumably which can be turned off and on.
23:36
<Hixie>
3. multiple independent tracks that need to be played along the same timeline
23:37
<zewt>
(do deaf people seriously use sign-language overlays? heh)
23:38
<Hixie>
zewt: some do. There's a lot of variety in the deaf community from what I understand, but certainly there are those that use sign-language overlays.
23:39
<zewt>
that's ... odd, heh
23:39
<Hixie>
why?
23:39
<zewt>
hard to think of why you'd want sign language instead of text, since I don't imagine you can watch someone signing while actually watching whatever video contains whatever's being signed
23:39
<AryehGregor>
I'd imagine you could watch sign-language as well as printed text.
23:40
<zewt>
text you can read in half a second and go back to the video
23:40
<Hixie>
sign-language is much denser than text, and you end up seeing it out of the corner of your eye
23:40
<AryehGregor>
Is the sign language we're talking about here animated, or some type of sequence of symbols?
23:40
<Hixie>
neither, it's live action of someone signing
23:40
<AryehGregor>
Interesting.
23:40
<AryehGregor>
Sign language is developed instinctively, like speech and unlike writing.
23:41
<AryehGregor>
Sign languages reliably arise spontaneously whenever you have a critical mass of deaf children and no local sign language for them to learn.
23:41
<zewt>
possibly a similar use case would be animated karaoke tracks of the sort fansubs like to use, which are way outside of webvtt's scope
23:41
<AryehGregor>
Whereas writing was an invention that was invented independently maybe like three times.
23:41
<AryehGregor>
And alphabetic writing was invented once.
23:41
<AryehGregor>
In all of human history.
23:41
<zewt>
having a separate baked video track overlaid on top of the video would be a lot nicer (in theory) then baking them right into the video
23:42
<AryehGregor>
So maybe sign language is much easier to follow than writing, if you grew up speaking it.
23:42
<zewt>
well, signing is more like speech than writing, and speech arises on its own too
23:42
<aho>
i imagine it's more like having someone talking to you
23:42
<aho>
instead of reading something
23:43
<zewt>
(one practical problem with karaoke overlays is video codecs generally don't support alpha, or if they do underneith it's never implemented)
23:44
<aho>
what's the new name of websrt again? :>
23:44
<zewt>
webvtt
23:44
<zewt>
(don't know why it was renamed, much harder to remember, heh)
23:44
<nessy>
there's also signwriting, btw - I think in the long run that has the most potential to become a machine-readable sign language
23:45
<nessy>
Hixie: I will make some notes on the examples
23:45
<aho>
i really wonder why they didn't go with something more straightforward... like... y'know... websubtitles :>
23:46
<nessy>
in general, your conclusion seems sound though
23:46
<aho>
or websubs lol
23:46
<nessy>
I took those examples also so we can see how multiple tracks can be displayed
23:47
<Hixie>
aho: webvtt has the advantage of an obvious filename extension (.vtt) that isn't used by anyone already (unlike .sub)
23:47
<nessy>
vtt = video text track - not that bad IMHO
23:47
<Hixie>
also "sub"-anything is someone misleading since we do more than subtitles
23:47
<aho>
still kinda hard to remember
23:48
<zewt>
we'll get used to it :)
23:48
<zewt>
better than .ASS
23:48
<aho>
oh yea, i remember that discussions from many moons ago... going with some "ASS" based name would have been fun :>
23:48
<nessy>
lol
23:49
<zewt>
don't know any particular discussion, just another common sub format :P
23:50
<aho>
basically, someone made a subset of aas and the idea was to call it small ass or something like that :>
23:50
<aho>
WebAss would have been fine, too
23:51
<aho>
haha
23:51
aho
is easily amused
23:53
<nessy>
(Hixie: re third one down - I was simply looking for an example where the sign language video would sit in same size as main video next to it, but I couldn't find an image better than this - maybe I am mistaken to have thought I'd seen that before)
23:54
<zewt>
sign language track - as a separate element or as an overlay?
23:55
<Hixie>
nessy: that's why it's important to have actual examples. :-)
23:55
<zewt>
(of course you can do both by overlaying two <video> elements, but that's sort of the wrong layer in general I'd think)
23:55
<Hixie>
zewt: that's what we're discussing
23:55
<aho>
zewt, would allow resizing and moving it around though
23:55
<nessy>
Hixie: you're right - I might remove that one and add your comments on all, so we keep track (not just in the bug tracker ;-)
23:55
<Hixie>
nessy: one use case that hasn't been talked about here that i'm interested in is playing two videos back synchronised but at different offsets
23:55
<aho>
full screen would be f-ed though :l
23:56
<Hixie>
nessy: something like the "double youtube" thing but where you can pick the synchronisation point
23:56
<zewt>
yeah, if you're on a TV you probably want something like PIP for that, not side-by-side
23:56
<nessy>
aho: when we attach full screen to page elements, it's possible still
23:56
<aho>
ah yea... there was that thing
23:57
<nessy>
Hixie: is that use case not just a toy? I'm worried we end up with SMIL ...
23:57
<aho>
does any browser support it today?
23:57
<zewt>
programmatic fullscreen doesn't exist yet at all afaik
23:57
<aho>
so far i've only seen very slow video fsm
23:57
<aho>
(full screen mode, not flying spaghetti monster)
23:57
<nessy>
aho, not yet - but it's being discussed
23:58
<zewt>
well yeah all (afaik) video implementations right now are really bad at fullscreening, presumably that'll get fixed
23:58
<Hixie>
nessy: toys are good too :-)
23:58
<Hixie>
nessy: it's the kind of thing that would be good to support if it can be done in a way that doesn't add complexity
23:58
<zewt>
Hixie: being able to adjust the offset is also useful to, for example, quickly adjust off-sync audio or captions
23:59
<Hixie>
yeah, that too
23:59
<nessy>
yeah, but then we get those that say that we need to be able to arbitrarily align any video track and audio tracks and have choices between subparts of them etc
23:59
<nessy>
that really is SMIL and I would prefer to keep such stuff to JavaScript
23:59
<Hixie>
nessy: e.g. if we have a controller + slaves model, you can imagine a model where the controller doesn't have a timeline, it just controls the velocity (and pauses everything whenever any of them stall), which would let you do all of the above with very minimal additional complexity