00:16
<TabAtkins>
foolip: Added the new object-fit values. Review, please? Also, dbaron said he has some objections to the values, so we'll see what he has to say.
00:43
<AryehGregor>
Hixie, there's a discussion on public-web-security where abarth is arguing CSP is too complicated, but he doesn't seem to have specific examples. I think you said similar stuff before; do you have specific examples?
00:44
<AryehGregor>
It looks way too complicated for authors to me, but I haven't looked closely enough to formulate specific examples.
00:44
<dbaron>
TabAtkins, you forgot to put scale-down in the syntax list at the top of the spec
00:44
<dbaron>
TabAtkins, also, the spec needs a "previous version" link, since it does have one
00:46
<dbaron>
er, top of the section, not top of the spec
00:47
<Hixie>
AryehGregor: i don't recall examples off-hand, but iirc my reaction was the same as yours
00:47
<Hixie>
AryehGregor: the key being that it's a security feature
00:47
<AryehGregor>
Yeah.
00:47
<Hixie>
AryehGregor: so complexity leads to security vulns
00:47
<AryehGregor>
On the other hand, the inflexibility required by simplicity conflicts with the goal of being deployable without large-scale content changes.
00:48
<Hixie>
it is a common dilemma
00:48
<AryehGregor>
But realistically, you're not going to get XSS protection without doing drastic things like banning inline script, so maybe deployability needs to be deemphasized.
00:49
<AryehGregor>
It actually even reminds me a bit of SELinux (although of course orders of magnitude less complicated). Someone was saying something about how authors should use a policy generator. :(
00:49
<AryehGregor>
And of course, with SELinux, the first advice everyone gives when it goes wrong is "Disable it."
00:50
<AryehGregor>
Ideally most of CSP should be reduced to just one on/off switch, like ES5 strict or something.
00:50
<AryehGregor>
I should post these ideas to the list, maybe I will tomorrow.
00:51
<AryehGregor>
Still catching up on stuff, although thankfully I didn't have to ignore *everything* while I was in Israel (or else I'd really have a lot to catch up on).
00:52
<Hixie>
heh, i know that feeling
00:55
<TabAtkins>
dbaron: Ah, right. Sorry about missing that.
00:55
<TabAtkins>
And I'll add a prev version link.
05:28
<yuhong>
Hixie: Have you heard of lcamtuf's mangleme?
05:28
<yuhong>
http://lcamtuf.coredump.cx/mangleme/
08:57
<hsivonen>
zcorpan: webm.html5.org is not yet in bitbucket
09:04
<zcorpan>
ok
09:09
<annevk>
http://dev.w3.org/2006/webapi/progress/Progress.html has a redirect now
09:10
<annevk>
thanks to MikeSmith!
09:20
<Ms2ger>
MikeSmith++
09:32
<erlehmann>
nice http://richwchan.com/2011/01/25/literally-drawing-the-html-5-logo-in-canvas-with-javascript/
09:38
<smaug____>
zcorpan: ping
09:39
<zcorpan>
smaug____: pong
09:39
<smaug____>
zcorpan: so why the need for capturing event listeners if forminput/change would be removed?
09:40
<erlehmann>
sloppiness 2.5 / 7.5 / 3.5 / 3.5 works fine
09:41
<zcorpan>
smaug____: i don't understand the question
09:41
<smaug____>
zcorpan: you're simon, right?
09:41
<zcorpan>
smaug____: yep
09:41
<smaug____>
I'm talking about http://www.w3.org/Bugs/Public/show_bug.cgi?id=11129#c15
09:43
<zcorpan>
oh, do the change and input events bubble?
09:46
<smaug____>
yes
09:46
<zcorpan>
i assumed that they didn't bubble
09:47
<zcorpan>
ok, then a capturing listener would not be needed for the shopping cart
09:48
<zcorpan>
smaug____: for multiple event handlers, i was considering several <output>s that update their own value when something else is changed
09:48
<zcorpan>
e.g. <output name=amountWithoutTax onformchange=...></output> <ouput name=amountWithTax onformchange=...></output>
09:48
<smaug____>
again, there could be one input/change event listeners doing that all
09:49
<zcorpan>
yes
09:50
<smaug____>
I'm having trouble to see the real usefulness of forminput and formchange
09:51
<smaug____>
I wonder when those were added and what discussion lead to it
09:51
<zcorpan>
i think they've been part of WF2 since 2004 or so
09:52
<smaug____>
aha, so there might not have been too much discussion about them ...
09:53
<smaug____>
and no wonder I haven't been able to find related emails
09:53
smaug____
tries again
09:53
<zcorpan>
i think they're mostly there so you can couple your event handler with the <output> tag instead of putting the logic far away from where it's supposed to update the content (and you don't need to name your outputs)
09:54
<annevk>
doesn't WebKit have them as well now?
09:54
<smaug____>
that use case is still rather vague
09:54
<smaug____>
I think webkit just added them, like few days ago
09:55
<smaug____>
I don't want to add useless features to web platform, if just possible
09:55
<smaug____>
I assume that if they were just added to webkit , they could be easily removed
09:55
<zcorpan>
also, with the form="" feature the forminput and formchange events are broadcasted to the controls that are associated with the form even if they're not descendants
09:55
<smaug____>
there is no content which relies on those events
09:55
<zcorpan>
i don't have a use case in mind though for that
09:56
<smaug____>
even with form="", if is trivial to just check event.target.form in the event listener
09:56
<smaug____>
s/if is/it is/
09:57
<zcorpan>
yes, but it's a bit different coding style. not a strong argument to support the events, but still :)
10:16
<zcorpan>
as for valueAsNumber, it might be a bit faster since you don't need to convert to string and back
10:16
<annevk>
but do you need speed?
10:17
<annevk>
does not seem like the thing that needs an optimization path
10:17
<zcorpan>
yeah, dunno
10:20
<smaug____>
ah, yeah, valueAsNumber doesn't look too useful either
10:20
<smaug____>
does Opera implement it?
10:20
<smaug____>
or webkit?
10:22
<zcorpan>
opera does
10:23
<zcorpan>
seems webkit too
10:23
<zcorpan>
but not firefox 4
10:24
<smaug____>
but I guess it is not really use anywhere yet
10:24
<smaug____>
so perhaps there is still chance to get rid of it
10:25
<zcorpan>
file a bug?
10:26
<MikeSmith>
annevk: fwiw, I had problems before with trying to get some htaccess working on dev.w3.org and only managed to figure it what worked through trial and error. I think they have something misconfigured on that server
10:26
<MikeSmith>
but glad it's working now anyway
10:27
<annevk>
did you remove ./ or some such?
10:27
<annevk>
or made it into an absolute path?
10:27
<MikeSmith>
yeah
10:28
<MikeSmith>
and used RedirectMatch instead of Redirect
10:28
<annevk>
oh
10:28
<MikeSmith>
btw, I also fixed the borkedness that was causing http://es5.github.com/ to load so slowly
10:29
<MikeSmith>
if you try it now it should work find
10:29
<MikeSmith>
*fine
10:29
<MikeSmith>
bypass your cached copy of course
10:31
<Rik`>
Am I dreaming or is Facebook really encouraging invalid HTML ? http://developers.facebook.com/docs/opengraph
10:31
<Rik`>
<meta property="og:title" content="The Rock"/>
10:31
<Ms2ger>
Yes
10:32
<Ms2ger>
They're encouraging fake namespaces too
10:32
<Rik`>
wtf ?
10:33
<zcorpan>
they should invent a microdata vocab and use that
10:33
<annevk>
ah, Facebook
10:33
<annevk>
that explains the 3.5% RDFa I saw somewhere
10:35
<annevk>
http://tripletalk.wordpress.com/2011/01/25/rdfa-deployment-across-the-web/
10:36
<MikeSmith>
Facebook deserves some love for https://github.com/facebook/jsgamebench though
10:36
<MikeSmith>
hope they do more of that kind of stuff
10:38
<Rik`>
oh so that's RDFa
10:39
<Rik`>
why isn't there any mention of that on the docs ?
10:39
<zcorpan>
<fb:like /> isn't RDFa
10:40
<annevk>
but property is
10:40
<jgraham>
MikeSmith: I heard reports that benchmark isn't very useful, but I haven't confirmed or anything
10:41
<MikeSmith>
jgraham: still happy to see them putting resources into developing such tools at all
10:42
<annevk>
how do you delete apps on android?
10:42
<jgraham>
annevk: Testing document.domain requires different domains (not subdomains), right?
10:43
<annevk>
jgraham, subdomains is fine
10:43
<annevk>
jgraham, with subdomains you can check if they can communicate with each other after setting document.domain, you cannot actually do that with different domains
10:43
<Rik`>
so how do I declare that my HTML page is using RDFa ?
10:43
<annevk>
document.domain is thus a good reason for subdomains :)
10:44
<jgraham>
annevk: Right, but you need different domains to check that they can't, or something
10:44
<jgraham>
I'm pretty sure that we have TCs that require multiple domains
10:44
<jgraham>
(Sigbjorn would know for sure)
10:44
<annevk>
well, you want to test publicsuffix maybe
10:44
<annevk>
but that requires a more complicated setup
10:45
<jgraham>
We need subdomains for sure
10:45
<jgraham>
But I think we also need two domains
10:45
<hasather>
annevk: easiest is to just find it in the app drawer, hold it, move it over the trash bin, hold for two seconds until it tells you to release to uninstall
10:45
<jgraham>
Why is pubsuffix more complex?
10:45
<annevk>
how are two domains different though?
10:45
<annevk>
other than publicsuffix
10:45
<jgraham>
Well that is one reason
10:46
jgraham
asks someone who actually knows :)
10:47
<annevk>
I'm pretty sure you cannot test publicsuffix with two domaisn
10:47
<annevk>
you would want to test all the permutations
10:47
<Rik`>
from a little bit of googling, rdfa in html5 does not require anything special but validator.nu is not marking rdfa as valid yet
10:47
<annevk>
e.g. with some kind of custom DNS setup
10:47
<Rik`>
is it right?
10:47
<annevk>
Rik`, something like that
10:47
<jgraham>
Right, doing a really good test of pubsuffix is hard
10:48
<annevk>
hasather, that does not seem to work :/
10:48
<annevk>
oh well
10:48
<Ms2ger>
Rik`, the cabal doesn't approve of rdfa
10:48
<hasather>
annevk: you can also uninstall from the market, or through Settings > Applications
10:49
<annevk>
ah that works
10:54
<jgraham>
annevk: Sigbjorn's list is "pubsuffix, x-frame-options, crossdomain XSS, cookies, cross-network". Not all apply to W3C ofc but being able to test them is good. Also browsers might have bugs in seperate domains that they don't have in subdomains
10:56
<annevk>
there's no real difference though between domains and subdomains
10:56
<annevk>
it's just a set of dots DNS
10:56
<annevk>
and DNS
10:57
<annevk>
e.g. is example.co.uk a subdomain?
10:58
<Ms2ger>
Pubsuffix could tell you that
10:59
<annevk>
pubsuffix draws arbitrary lines too
10:59
<Ms2ger>
Any spec does ;)
10:59
<annevk>
e.g. dyndns.com iirc is a "TLD"
11:00
<annevk>
Ms2ger, my point is that apart from pubsuffix there's not really any meaningful difference between domains and subdomains
11:00
<Ms2ger>
True
11:27
<MikeSmith>
annevk: you saw http://gmailblog.blogspot.com/2011/01/desktop-notifications-for-emails-and.html ?
11:28
<annevk>
yeah
11:33
<hsivonen>
does anyone happen to have an URL for an HTML page that's served over IPv6 only (no IPv4 A record)?
11:33
<hsivonen>
s/an URL/a URL/
11:34
<annevk>
http://ipv6.google.com/
11:34
<hsivonen>
thanks
11:35
<annevk>
I cannot open that, I wonder when I will drop from the interwebs
11:35
<hsivonen>
Gandi got IPv6 routing
11:35
<hsivonen>
I wonder if html5.validator.nu would break if I added an AAAA record for it
11:36
<hsivonen>
at present, it seems that html5.validator.nu itself can't validate http://ipv6.google.com/
11:40
<annevk>
hmm
11:41
<annevk>
http://googlepublicpolicy.blogspot.com/2010/12/making-copyright-work-better-online.html Google siding with the entertainment industry
11:41
<annevk>
bah
11:41
<jgraham>
annevk: Assuming that browsers work in the same way as specs theorise is not always a great idea
11:43
<annevk>
jgraham, is this still about the domain thing?
11:43
<jgraham>
Yes, I was away
11:43
<annevk>
I'm be interested in seeing an implementation that does it wildly different :)
11:44
<annevk>
In the end though code will be changed and converged... You cannot really anticipate all possible implementation mistakes everywhere
11:45
<jgraham>
Right, but I don't really see the disadvantage of giving ourselves the option for cross domain (not subdomain) tests
11:47
<MikeSmith>
Rik`: validator.nu did have RDFa support previously, but it was removed
11:47
<Rik`>
MikeSmith: since so many pages are using facebook stuff, is there some info we can give?
11:47
<Rik`>
cause right now, it looks like it's not valid at all
11:48
<annevk>
jgraham, maybe because you insist there's a difference :)
11:51
<MikeSmith>
Rik`: for now, the W3C validator has an option to do XHTML+RDFa checking
11:51
<MikeSmith>
those docs are all being served with an XHTML doctype anyway, right?
11:52
<MikeSmith>
and namespace declarations for the namespaces that Facebook uses for their properties
11:54
<Rik`>
MikeSmith: it's HTML
11:54
<MikeSmith>
the doctype? or the mime type?
11:54
<Rik`>
you don't need XHTML to provide a screenshot
11:54
<Rik`>
mime type
11:54
<MikeSmith>
yeah well
12:15
<hsivonen>
MikeSmith: what does XHTML+RDFa actually check?
12:15
<MikeSmith>
hsivonen: dunno
12:15
<MikeSmith>
I've not actually tried it myself
12:20
<MikeSmith>
hsivonen: http://dev.w3.org/validator/htdocs/sgml-lib/REC-rdfa-syntax-20081014/ I guess
12:21
<hsivonen>
MikeSmith: so not checking anything but typos in the names of the attributes?
12:21
<MikeSmith>
seems so
12:21
<MikeSmith>
essentially
12:22
<annevk>
omg some people
13:03
<annevk>
what if the moon falls on the earth
13:07
<zcorpan>
i guess that'd be pretty bad
13:08
<jgraham>
Dear lazy-irc: For a SVG testcase I want to align the bottom of some text (specifically, the bottom of the em-square, or whatever the filled area in the X glyph in Ahem corresponds to) with the bottom of a <rect>
13:08
<annevk>
are we assuming it doesn't and build around that?
13:08
<annevk>
o_O
13:08
<gsnedders>
jgraham: You can choose the co-ordinate point that text is drawn from in SVG
13:08
<jgraham>
gsnedders: How?
13:08
<gsnedders>
Including from the baseline
13:09
<gsnedders>
jgraham: I can't remember :P
13:09
<gsnedders>
My normal solution in your position: ask ed.
13:09
<jgraham>
baseline !== em-square aiui
13:09
<gsnedders>
the em-square goes down to the bottom of baseline
13:10
<jgraham>
Really? Which baseline? I am just looking at the diagram in the HTML spec
13:10
<gsnedders>
There is a line called the baseline.
13:11
<gsnedders>
It becomes complicated when you use multiple scripts, but that's a separate issue unrealted to this case
13:12
jgraham
notes it is already complicated
13:15
jgraham
doesn't see anything relevant in http://www.w3.org/TR/SVGTiny12/text.html#CurrentTextPosition
13:16
jgraham
decides to pursue a different approach
13:32
<annevk>
can you change the templates MediaWiki uses from the Wiki instance itself?
13:32
<annevk>
or do you need to change PHP and such somewhere?
13:51
<Ms2ger>
annevk, templates are written in wikitext
13:55
<annevk>
MediaWiki:Sidebar was the one I wanted to change
13:55
<annevk>
navigation was linking to dead pages
14:17
<hsivonen>
hmm. SVG in a big language in terms of vocabulary tokens
14:30
<zcorpan>
hsivonen: looking at sanitation?
14:30
<hsivonen>
zcorpan: yes. I'm assembling lists of interned tokens manually
14:30
<hsivonen>
there's *lots* of tokens
14:31
<hsivonen>
and I'm manually categorizing them as SVG fonts (omitted) and SMIL (behind #ifdef)
14:32
<hsivonen>
one of those tasks that maybe could have used a script, but developing the right script would have taken time considering that it needs to be run once
14:34
<hsivonen>
and the hyphenation vs. camelCase is full of FAIL. Found one bug already.
14:43
<jgraham>
The interesting thing about scripts that need to be run exactly once is that they always end up being run >>once
14:44
<jgraham>
Not saying it is worthwhile in this case though
14:59
<hsivonen>
SVG fonts bloats the language by many, many attributes...
15:09
<annevk>
they gave me a new iPhone...
15:11
<zcorpan>
annevk: SVG fonts gave you an iPhone?
15:11
<annevk>
yeah, didn't work on my Android
15:11
<zcorpan>
aha
15:12
<annevk>
hey, it makes sense to me
15:14
<hsivonen>
it's so sad that they compile the Android WebKit without SVG support
15:14
<hsivonen>
gives opportunity for Firefox and Opera I guess, on the bright side
15:16
<Ms2ger>
But WebKit supports all of HTML5, including SVG!
15:17
<AryehGregor>
Why do they compile it without SVG support? I've noticed that.
15:17
<zcorpan>
HTML5 on Android doesn't include SVG
15:28
<karlcow>
http://www.w3.org/mid/q433k65vs00o7509mrpbq1sll9ukt9uoc3⊙hbhd
15:29
<karlcow>
RELAXNG to JSON
16:07
<annevk>
http://www.infoworld.com/d/developer-world/html-the-standard-failed-585 sigh
16:07
<annevk>
can't even comment
16:08
<TabAtkins>
Man, screw you, Infoworld, and your stupid sign up.
16:09
<TabAtkins>
Also: Argh, that post is a giant pile of misconceptions that we already answered in the FAQ.
16:10
<annevk>
yeah, it is quite weird
16:11
<jgraham>
You know the FAQ is a waste of time, right?
16:11
<TabAtkins>
Clearly.
16:11
<annevk>
I don't think it is
16:11
<jgraham>
The sort of people with a giant pile of misconceptions will never read the FAQ
16:11
<TabAtkins>
I mean, my questions are obviously superior to the plebes.\
16:12
<TabAtkins>
And thus, by definition, not frequently asked.
16:12
<annevk>
jgraham, well yeah, but it's useful for a lot of other people
16:12
<TabAtkins>
Thus I have no need to read the FAQ, QED.
16:12
<jgraham>
I guess it works to tell people RTFM after the fact
16:12
<annevk>
I wonder if Infoworld is read well
16:12
<annevk>
and it is very useful to have all that information gathered somewhere
16:13
<jgraham>
Yeah, having the information is useful
16:13
<jgraham>
I meant "is a waste of time for the purposes of preventing people from publishing their misconceptions all over the intertubes"
16:14
<annevk>
geez
16:14
<annevk>
am I really the person that has to explain everything to Leif?
16:15
<jgraham>
No, you could ignore him like everyone else
16:15
<Lachy>
jgraham, having an FAQ is useful as a resource to point people to when they ask common questions so that people don't have to keep making up new answers all the time.
16:15
<jgraham>
Lachy: Right, see what I said about RTFM above
16:15
<annevk>
that does not seem too nice at this point, but this does not seem to be going anywhere
16:15
<TabAtkins>
annevk: Yeah, what jgraham said. I just stopped talking to him a long time ago.
16:15
<Lachy>
I don't think anyone expects all people to read it first before asking silly questions
16:15
<TabAtkins>
I think I need to filter him like I do Garrett Smith.
16:17
<wilhelm>
He does have a point, though. Giving a certain feature set an unambiguous name does have its advantages.
16:19
<wilhelm>
Opera 10 has certain features, Opera 11 has certain other features. The Web and HTML is much more complex than that, of course, but I can see where he's coming from.
16:19
<jgraham>
wilhelm: The infoworld guy or Leif?
16:19
<wilhelm>
The infoworld person.
16:21
jgraham
now has a theory that even mentioning Leif causes the strange vortex of confusion to occur, whereupon all straightforward conversations become mired in twisty logic and disjoint arguments
16:21
<Lachy>
wilhelm, we should drop versioning in Opera too, and just have the update system work so that it silently updates people to the latest
16:21
<jgraham>
wilhelm: If you follow the Google philosophy then you don't need that for browsers
16:22
<jgraham>
It doesn't seem to be doing them any harm
16:22
<wilhelm>
Lachy: Sure. That doesn't work on all platforms, though.
16:22
<jgraham>
and it is less true for browsers (which are by-definiton single implementation) than for specs
16:22
<TabAtkins>
Lachy: You can keep versioning, if the update would be more silent. ^_^
16:22
<jgraham>
where a single "version" doesn't correspond to anything concrete in the real world
16:23
<Lachy>
TabAtkins, sure, like Chrome's versoining which exists, but which isn't publicised anywhere
16:23
<hsivonen>
annevk: Re: waves of negative blog comments, jd calls in seagull linking
16:23
<TabAtkins>
Yes.
16:23
<hsivonen>
happens to his blog when Gruber links :-)
16:24
<jgraham>
Presumably after a while Chrome will give up on version numbers entirely
16:24
<jgraham>
If browsers manage to make versionless UA strings
16:25
<jgraham>
(well that isn't a precondition exactly, but if the only place you can observe the version number is in the about dialog, it doesn't mean anything anymore)
16:25
<paul_irish>
There are a lot of reasons why versions can't be scrubbed from UA strings, unfortunately
16:25
<annevk>
wilhelm, how was it clear with HTML5 what we implemented? seems exactly the same problem
16:25
<wilhelm>
We have a test suite internally testing form interaction. It deliberately tests only the HTML4 feature set (text, password – not number, date, email), but of course adheres to the HTML[5] spec for the technical details. That is a useful distinction.
16:25
<jgraham>
paul_irish: I thought Mozilla were pushing toward fake versioning
16:25
<jgraham>
Or maybe no versioning
16:26
<annevk>
wilhelm, you think?
16:26
<jgraham>
wilhelm: Only if it corresponds to some implementation detail
16:26
<paul_irish>
jgraham: in the UA string? eesh. i should dig into that. :/
16:26
<annevk>
wilhelm, split feature sets are useful I suppose
16:26
<annevk>
wilhelm, but that does not have to do much with HTML4 vs HTML5
16:26
<jgraham>
paul_irish: What would break if you just froze the UA string at some arbitary point?
16:27
<paul_irish>
nothing would break, but there would be a lot of features/bugs that emerge after that point that would be undetectable.
16:28
<wilhelm>
annevk: Yes. In this particular case, the HTML-1997 feature set was easy to test – the HTML-2009 feature set more difficult. Saying "Let's focus on the HTML4 feature set for now" was a useful shorthand.
16:29
<miketaylr>
you might also be locked in to a buggy codepath, even if the bug was fixed post-freeze. e.g., https://github.com/marijnh/CodeMirror/blob/master/js/editor.js#L7-19
16:30
<jgraham>
My (biased) experience is that UA sniffing *always* ends badly
16:30
miketaylr
agrees
16:31
<jgraham>
So I am inclined to think that the problems caused by making it harder would be smaller than the problems it causes
16:31
<annevk>
wilhelm, fair enough
16:31
<annevk>
wilhelm, I guess in the future that would be old vs new :)
16:32
<annevk>
wilhelm, the danger is of course that people might look at HTML4 and find bugs relative to that
16:33
<paul_irish>
jgraham: this is what i'm getting at. https://github.com/Modernizr/Modernizr/wiki/Undetectables basically unless all these things are detectable without using UA.. then authors will need the UA+version for when feature detection doesnt suffice.
16:34
<Rik`>
paul_irish: or unless UAs implement half baked things
16:34
<miketaylr>
which is the real problem, imo
16:35
<Rik`>
like webforms UI…
16:35
<paul_irish>
indeed. :)
16:36
<jgraham>
paul_irish: The flipside of that is that people make broken assumptions based on UA detection. Like "all mobile browsers are webkit"
16:37
<paul_irish>
yes. in 95% of cases, the UA sniff is unnecessary. i'm just saying there is a legit 5% in there.
16:38
<paul_irish>
but.. authors mostly suck. that's why we "deprecated" $.browser in jQuery. basically saying "stay away! figure it out youself, bro!"
16:38
<Rik`>
some of the undetectables in that page are useless to detect
16:39
<Rik`>
registerProtocolHandler()
16:39
<Rik`>
if you detect that it's not supported, what is your fallback ?
16:39
mhausenblas
waves to paul_irish - any news re http://html5rocks.com CORS enabled? last time I checked it seemed not to be the case - just wondering when to update http://enable-cors.org/#who
16:39
<karlcow>
jgraham: the issue is not really about having versionless UA string, but having *all user agents* doing that at the same time :)
16:40
<karlcow>
indeed there will be massive breakage around
16:40
<jgraham>
karlcow: I'm not proposing versionless
16:40
<karlcow>
what do you propose
16:40
<jgraham>
I'm suggesting you can freeze the version number
16:40
<paul_irish>
mhausenblas: everything at www.html5rocks is Access-Control-Allow-Origin:* now :)
16:41
<mhausenblas>
hu? hmmm then I need to check my checker :)
16:41
<karlcow>
same issue
16:41
<mhausenblas>
ha!
16:41
<mhausenblas>
ok, true paul_irish www.html5rocks.com is but I checked at html5rocks.com
16:41
<karlcow>
jgraham: if browsers implementers don't do it at once. There will be no changes
16:42
<annevk>
Rik`, you don't always need fallback
16:42
<wilhelm>
annevk: Absolutely. Freezing parsing rules, for example, would be insane. But grouping features together and giving them a name is useful.
16:43
karlcow
is going back to user agent sniffing issues ;) for opera
16:49
<kennyluck>
wow, great! paul_irish! (re. everything at www.html5rocks is Access-Control-Allow-Origin:* )
16:49
<paul_irish>
♥ cors
17:00
<thecolor3>
anyone here happen to have any history with the jw player html5 fallback system?
17:07
<annevk>
jgraham, your example seems highly unlikely
17:07
<annevk>
jgraham, probably because you do not author a blog :p
17:08
<annevk>
I have never seen it done that way
17:09
<jgraham>
annevk: You really think that no one will consider <header> to be the header for the entire page including the title of the content?
17:09
<jgraham>
Maybe not on a blog
17:09
<jgraham>
But in general
17:09
<annevk>
well, in that case grouping them is not too problematic
17:09
<annevk>
there's only one subject
17:09
<annevk>
you would care mostly about the other headings anyway
17:10
<jgraham>
You would care mostly about the heading that you would lose
17:10
<jgraham>
i.e. the title of the content
17:10
<TabAtkins>
Agreed with jgraham - I think I might do that <header> pattern sometimes.
17:10
<jgraham>
(rather than the title of the site)
17:10
<annevk>
o_O
17:14
<thecolor3>
Can someone please tell me if they have any history with fall back on html5 specifically with jw player?
17:14
<hober>
what was jgraham's example?
17:16
<Ms2ger>
thecolor3, doesn't look like it
17:16
<thecolor3>
Ms2ger: yea so it seems
17:16
<hober>
oh, I see it now (caught up on email)
17:17
<annevk>
thecolor3, maybe in #html5
17:46
<karlcow>
will there be an API for NFC? http://venturebeat.com/2011/01/25/iphone-5-ipad-2-nfc/
17:48
jgraham
is fooled by the NFC !== Normal Form C thing yet again
17:48
<karlcow>
sorry, Near Field Communications
17:48
jgraham
hopes this technology doesn't catch on
17:48
<jgraham>
Or I will be confused the whole time
17:49
<jgraham>
Moreso than normal I mean
17:49
<karlcow>
jgraham: I think you will be confused the whole time ;)
17:49
<annevk>
I was too
17:49
<AryehGregor>
I was briefly confused before I figured out what was meant.
17:49
<annevk>
until I saw vendturebeat.com and Apple product names
17:50
<karlcow>
5x more NFC than NFC ;)
17:50
<jgraham>
Hopefully someone figures out that Near Field Communications is not the sort of name you can sell to the mass market
17:50
<jgraham>
and calls it something else
17:50
<karlcow>
jgraham: you mean like GPS :p
17:50
<jgraham>
karlcow: dammit
17:50
<karlcow>
hehe
17:51
<TabAtkins>
I was confused for an entire lunchtime conversation over that.
17:51
<jgraham>
It doesn't have to be true that you can't sell it, you just have to have them believe it
17:51
<karlcow>
in fact GPS as the same issue than HTML5, GPS is used for geoip, celltower triangulation, etc. for everything-location-through-waves
17:52
<AryehGregor>
To most people, a "GPS" is the thing in your car that tells you where to drive.
17:52
<AryehGregor>
:)
17:53
jgraham
votes we rename it Hvísla
18:08
<annevk>
since when did I become a member of the press?
18:08
<annevk>
yesterday some marketing agency asks my address details for "free" gifts
18:08
<Ms2ger>
Since you got a blog and a twitter account
18:08
<TabAtkins>
Interesting.
18:08
<annevk>
today I get some kind of press release about a code competition and invitation to talk with the CEO if I want more details...
18:08
<annevk>
is there a list somewhere?
18:08
<annevk>
I want off
18:09
<Ms2ger>
You want X-Do-Not-Act-Like-I-Am-A-Journalist
18:10
<annevk>
:)
18:10
<annevk>
that just leads to the inevitable X-Do-Act-Like-I-Am-A-Journalist-I-Really-Mean-It
18:41
<TabAtkins>
"Web developers have tried to compensate for this problem by creating IPv6 -- a system that recognizes six-digit IP addresses rather than four-digit ones." ~http://www.foxnews.com/scitech/2011/01/26/internet-run-ip-addresses-happens-anyones-guess/
18:42
<TabAtkins>
On that note, though, does anyone know why ipv6 is named ipv6? It's got 16 bytes, so it's not following the same naming scheme as ipv4, unless they're just taking it mod 10.
18:43
<TabAtkins>
...Oh. It's just version 6, and ipv4 is version 4. No relation to the byte size at all.
18:44
<AryehGregor>
Yes.
18:45
<AryehGregor>
I know I've seen some blog software where the italics button produces <em>. Does anyone know what I'm talking about?
18:46
<Dashiva>
hiptsercat/I still use/ipv5/
18:46
<TabAtkins>
I think wordpress does/did that.
18:46
<AryehGregor>
Yeah, but the Wordpress blogs I'm looking at don't seem to have markup buttons.
18:46
<AryehGregor>
At least in WebKit.
18:46
<AryehGregor>
Maybe they do in other browsers?
18:46
<TabAtkins>
They do in the article posting interface.
18:46
<TabAtkins>
Just not in comment posting.
18:47
<AryehGregor>
Ah.
18:50
<AryehGregor>
Grr.
18:50
<AryehGregor>
I know I've asked this more than once before, but where are some stats on how often various elements are used, more detailed than Google's 2005 study?
18:51
<Philip`>
http://philip.html5.org/data/tag-count-pages.txt ?
18:51
<AryehGregor>
There we go, thanks.
19:10
<foolip_>
so, uh, anybody have a great idea about how to proceed on http://lists.w3.org/Archives/Public/public-html/2011Jan/0415.html ?
19:12
<TabAtkins>
I really don't know. It doesn't make a ton of sense to continue on with the issue if you're trying to raise a larger bug.
19:13
<annevk>
foolip_, it seems like you do not need to withdraw it
19:14
<annevk>
foolip_, it seems in scope as well... it's just that your solution is slightly more complex
19:14
<foolip_>
the change I want would conflict with the change Julian wants, so the question is if I can just push for what I want and let the conflicts happen as they may
19:14
<annevk>
(though better, imo)
19:15
<foolip_>
it would seem a bit sneaky if, for example, I just propose this change, Hixie does it, and when Julian's issue has been resolved there's no longer any application/octet-stream to change the wording on
19:15
<annevk>
if you are still unsure I would reply and ask that, but I do not think you have to do anything at this point
19:16
<foolip_>
also, just looking at the size of the changes, my proposal is pretty much guaranteed to draw more objections
19:16
<foolip_>
if that should happen, how should one interpret that decision when we want to kill Content-Type?
19:17
<karlcow>
Philip`: you should put a date at the top of this txt file.
19:17
<annevk>
foolip_, it can be re-raised
19:17
<annevk>
foolip_, if there's sufficient support it will happen that way
19:18
<MikeSmith>
nice, selecing Settings from a Gmail desktop notification crashed my chrome
19:18
MikeSmith
looks around for atwilson
19:19
<karlcow>
with a baseball bat?
19:27
<foolip_>
annevk, ok, do nothing it is
19:27
<foolip_>
any Mozillans in here?
19:27
<foolip_>
roc, maybe?
19:28
<foolip_>
it'd be nice to hear if their position on this has changed at all
19:28
<roc>
position on what?
19:29
<roc>
oh that
19:29
<Ms2ger>
Racing to the bottom of the poisoned well
19:29
<roc>
that's my line!
19:30
<foolip_>
Ms2ger, right, dropping right to the bottom instead how slowly and painfully being dragged down, creating a mess along the way :)
19:30
<Ms2ger>
I guess we've lost already :)
19:30
<foolip_>
s/how/of/
19:31
<foolip_>
right, I think it's easier to just give up and do what IE, Chrome and Safari are already doing
19:32
<roc>
I agree
19:32
<roc>
I don't like it, but I agree
19:32
<Ms2ger>
How about we all just implement it, then?
19:32
<roc>
I should mention it to our video guys again to make sure they're not going to rebel though
19:33
<foolip_>
roc, please do :)
19:33
<foolip_>
I assume you're already "sniffing" for the difference between audio/ogg and video/ogg anyway :)
19:33
<foolip_>
that is, just decoding it and seeing what happens
19:33
<roc>
yes of course
19:34
<roc>
although we will have to add new sniffing code since currently we instantiate a decoder based on MIME type
19:34
<roc>
back in a bit
19:35
<MikeSmith>
http://twitter.com/#!/pbakaus/status/30344292123807745 regarding what's lacking in canvas to make it useful for game programming… "needs more features like click detection and the ability to have sprites rather than just set and wipe pixels"
19:38
<foolip_>
roc, OK, good to know that you're also prepared to give up :)
19:38
<foolip_>
or should it be :/
19:38
<foolip_>
MikeSmith, drawImage ?
19:39
<foolip_>
and <canvas onclick>
19:58
<twisted`>
question... content: url(image.png); seems to replace the image in WebKit but not Gecko
19:58
<twisted`>
is this a bug or just not implemented?
19:59
<twisted`>
(fx 3.6 btw :))
20:02
<twisted`>
same in the firefox 4 beta 10
20:03
<smaug____>
twisted`: replace which image?
20:03
<twisted`>
smaug____: it's a span with text that gets replaced
20:04
<Philip`>
karlcow: Even better than that, the date is before the top of the text file, in the HTTP header :-p
20:04
<smaug____>
twisted`: hmm, dbaron should know...
20:05
<karlcow>
Philip`: :p
20:05
<Philip`>
(None of my data stuff is really meant to make sense out of its original context)
20:05
<karlcow>
yep
20:05
<Philip`>
(which is usually some IRC conversation or email etc)
20:06
<twisted`>
smaug____: not a big issue I can also do a hide of the text and then do a background
20:06
<twisted`>
but this looks nice a nicer solution
20:06
<Philip`>
(and there's too much required context and too many caveats to make the data presentation self-contained)
20:06
<smaug____>
twisted`: it is possible that content: url(image) isn't implemented
20:06
<karlcow>
Philip`: you should add
20:06
<karlcow>
X-Context: "None of this data stuff is really meant to make sense out of its original context"
20:07
<karlcow>
in the HTTP headers
20:07
<karlcow>
:p
20:08
<smaug____>
twisted`: apparently CSS2.1 isn't too strict what should happen with content: url(some_resource)
20:08
<Ms2ger>
content isn't implemented on elements, only for pseudo-elements in Gecko
20:08
<smaug____>
I doubt anyone supports content: url(somepage.html)
20:09
<smaug____>
ah, right, Ms2ger is most probably right
20:09
<smaug____>
:)
20:09
<TabAtkins>
Presumably it *should* work like an iframe. But yeah, very underspecified.
20:10
<Ms2ger>
Does that work in WebKit already? I thought it was just Opera...
20:10
<zcorpan>
seems to work in webkit
20:11
<AryehGregor>
That's vaguely horrifying.
20:11
<AryehGregor>
Although I can't say exactly why.
20:11
<zcorpan>
i mean element { content:url(image.png) } works in webkit
20:11
<AryehGregor>
Oh, that's not so horrible.
20:11
<AryehGregor>
I thought we were talking about HTML.
20:11
<twisted`>
hmm
20:11
<twisted`>
well I'll just stick the image directly in it
20:11
<twisted`>
:)
20:12
<TabAtkins>
zcorpan: Though, it doesn't work the way I think it *should*; it doesn't act like a replaced element, but rather like an element with an anonymous <img> child.
20:12
<twisted`>
just need to figure out if the image of a single-quote I put in the content can be moved around
20:12
<TabAtkins>
When I pick up Generated Content I'll figure out whether we can make that automatic or have a switch or what.
20:12
<zcorpan>
TabAtkins: i think it does in opera too
20:12
<karlcow>
what's happening with svg instead of png
20:12
<karlcow>
and with html called from svg ?
20:13
<smaug____>
based on CSS 2.1, content applies only to :before and :after
20:13
<zcorpan>
TabAtkins: iirc there's a good reason for it but i'm not sure what it is
20:13
<Ms2ger>
With a difference between |url(image)| and |url(image) ""|?
20:13
<TabAtkins>
Ms2ger: That's what I'm thinking, yeah.
20:13
<Ms2ger>
I guess that would probably work
20:13
karlcow
has to finish this email, but it seems like a fun thing to test
20:13
<annevk>
please don't make content:url() work for anything but images
20:14
<zcorpan>
TabAtkins: maybe it's because it's supposed to be consistent with how content: "foo" url(image) "bar"; would be handled
20:14
<annevk>
also, when directly applied 'content' does make some kind of replaced element, no?
20:14
<annevk>
maybe I misremembered
20:14
<TabAtkins>
zcorpan: Yeah, but it makes it impossible to do actual image replacement, which is annoying.
20:15
<annevk>
we should just special case content with a single url()
20:15
<annevk>
imo
20:15
<karlcow>
well annevk that would reverse the snafu
20:15
<TabAtkins>
annevk: Yeah, that's what I'd like to do. I'll explore it later when I pick up the draft, though if anyone wants to forge ahead with an implementation, feel free.
20:15
<karlcow>
HTML5 would be then included in CSS3
20:15
<Ms2ger>
Tab, how many specs are you editing now, and how many are you planning to take up in the near future? :)
20:15
<zcorpan>
TabAtkins: what if you want image replacement and some text after like the header on anne's blog?
20:15
<zcorpan>
TabAtkins: and want to resize the image?
20:16
<annevk>
note that my header changed
20:16
<TabAtkins>
zcorpan: use content on the elemetn and on ::before?
20:16
<TabAtkins>
zcorpan: Alternately, don't do that.
20:16
<zcorpan>
TabAtkins: maybe we could have some sort of pseudo-element to reach each part of the generated content?
20:16
<annevk>
zcorpan, XBL
20:17
<zcorpan>
or yeah, we could try not to overengineer it and keep it simple and pragmatic :)
20:17
<annevk>
lets not introduce more special pseudo-elements for scenarios that can be solved with XBL
20:20
<TabAtkins>
annevk: FYI, you show up in autocomplete after typing "anne va", and are the first result when I add the "n".
20:20
<TabAtkins>
annevk: The results for "anne v" are much better, though.
20:21
<annevk>
:)
20:22
<karlcow>
some sites are using https://developer.mozilla.org/en/JavaScript/Reference/Global_Objects/Object/defineSetter because of the Pigment JS library
20:24
<karlcow>
s/pigment/figment/ pfff dyselxya
20:26
<TabAtkins>
link to the engine? search results are surprisingly unhelpful here.
20:27
<zcorpan>
TabAtkins: btw, using ::before wouldn't help with that case since it ends up being pretty much the same
20:28
<TabAtkins>
Okay. I dunno quite what the example is anyway now, since Anne's header changed.
20:37
<zcorpan>
it used to be something like h1 { content:url(daddy) " weblog" }
20:37
<zcorpan>
s/daddy/logo/
20:37
<TabAtkins>
...
20:37
<TabAtkins>
That's an odd substitution.
20:38
<Dashiva>
The keys are like right next to each other
20:38
<TabAtkins>
Only on the DADDYLOGO keyboard layout.
20:38
<zcorpan>
yah, DADDYLOGO is what i use
20:40
<TabAtkins>
Never understood the claimed performance benefits of DADDYLOGO. I'd have to decide which of three keys to press every time I want a D!
20:40
<TabAtkins>
Seems like it would slow you down.
20:41
<AryehGregor>
No, because you could pick the closest one.
20:41
<AryehGregor>
Clearly it would be much faster to type, e.g., "Daddylogo" in this layout than in QWERTY.
20:42
<TabAtkins>
Point.
20:43
dglazkov
can hear TabAtkins giggling in his cube. Clearly I missed a great DADDYLOGO keyboard discussion.
20:50
AryehGregor
relentlessly pursues his campaign to get minor Linux annoyances fixed: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=611299
20:50
<AryehGregor>
I don't know why I'm investing all this effort into that bug.
20:50
<AryehGregor>
It's not like it matters much. I guess that once I took the time to write a patch, I may as well see it through.
20:54
<annevk>
TabAtkins, daddy is one of the other images :)
20:54
<annevk>
TabAtkins, the one that says "SVG Sucks"
20:54
<TabAtkins>
annevk: Ah, that makes sense.
20:55
karlcow
has an evil laugh
20:55
<karlcow>
http://www.la-grange.net/2011/01/27/test/content-css
20:56
<karlcow>
we can insert SVG with Firefox and Opera and have a working foreignObject inside containing HTML
20:56
<annevk>
karlcow, content:url(svg) is also buggy in Opera
20:56
<karlcow>
at least it displays
20:57
<annevk>
only after reload
20:57
miketaylr
has to reload too
20:57
<annevk>
known bug btw, affects my blog too
20:58
<karlcow>
:) updated.
20:58
<AryehGregor>
"Most specifically, ISPs can force a downgrade of https to http, but Sullivan said that Facebook had not seen that happen." http://www.theatlantic.com/technology/archive/2011/01/the-inside-story-of-how-facebook-responded-to-tunisian-hacks/70044/ Is that actually true?
20:59
<AryehGregor>
Maybe they mean if the user doesn't connect with HTTPS to start with, and connects via HTTP only to be redirected to HTTPS.
20:59
<AryehGregor>
(yay STS)
20:59
karlcow
tries to imagine people will use SVG for inserting html in background images
21:01
<karlcow>
I wonder if video are working too
21:01
<karlcow>
videos
21:01
<annevk>
hmm yeah, HTML as images too?
21:02
<annevk>
that would require an update to the image loading spec
21:02
<AryehGregor>
SVG in things like background images or <img> already suppresses script and interactivity and whatever, why not also embedded HTML?
21:02
<AryehGregor>
Or does it just not really matter?
21:04
<annevk>
why would it suppress it?
21:04
<AryehGregor>
I dunno.
21:04
<AryehGregor>
It seems pointless to include HTML in SVGs if they're non-interactive, though. Hmm, I guess you might want to do <video> or something . . .
21:08
<karlcow>
I put links. And they are not working at least. They show up but are not activated http://www.la-grange.net/2011/01/27/test/content-css
21:10
<TabAtkins>
Hm, given that <url> is an <image> type, I guess it makes sense to just accept anything that you could validly point an <img src> at.
21:11
<TabAtkins>
And treat it the same way.
21:13
<Hixie>
foolip_: despite what sam says, in my experience there is plenty of precedent for issues' CPs going way beyond the scope of the original bug's request
21:25
<foolip_>
I'll just go on vacation and see if anything interesting has happened to the ISSUE when I'm back
21:27
<AryehGregor>
lol. http://blogs.technet.com/b/rhalbheer/archive/2011/01/14/real-physical-security.aspx
21:29
<TabAtkins>
Heh, I love nested data urls.
21:32
<twisted`>
what's the most adviceable route for implementing SVG? Do a check if the browser supports it and then replace it using javascript or css?
21:32
<AryehGregor>
twisted`, inline SVG? SVG in <img>, <object>? Something else?
21:33
<twisted`>
AryehGregor: img if supported
21:34
<twisted`>
never used it on sites but considering the support in most major browsers really kicks ass I see no reason not to do it ;)
21:34
<AryehGregor>
Probably the best way is just to blacklist based on UA string. I don't think there's any way to feature-test there, except possibly by trying to load the image and checking for error events, which is slow.
21:34
<AryehGregor>
(note: do not whitelist based on UA string, that's evil)
21:34
<twisted`>
AryehGregor: I was planning more a whitelist :p browsers I know that support it get the SVG and else... just the PNG
21:35
<twisted`>
oh, ok, why not?
21:35
<AryehGregor>
Whitelisting means your page won't work well with any new browsers that show up.
21:35
<AryehGregor>
It will bitrot.
21:35
<AryehGregor>
Blacklisting only punishes new browsers if they do the wrong thing and don't support SVG.
21:35
<AryehGregor>
Which is unlikely for mainstream browsers at this point.
21:35
<AryehGregor>
Whereas the existence of browsers you haven't heard of is basically a certainty.
21:35
<twisted`>
hmm not possible to say IE =< 9 allowed but the rest not? dunno
21:36
<TabAtkins>
Set some value in a conditional comment for IE <= 9, then check if it's set. If not, add the SVG and hope for the best.
21:38
<twisted`>
hmm ok never mind turns out the logo was designed by a n00b and did weird effects
21:38
<twisted`>
and the logo goes all haywire on me as a svg
22:06
<roc>
<video> in an SVG image might currently work in Firefox, but it's about to stop working
22:06
<roc>
we think everyone will need to disable loading of external resources from an SVG image
22:07
<TabAtkins>
<video src=data:video/ogg,...> should still work, then? ^^;
22:07
<roc>
yeah, we'll still support data:
22:07
<roc>
good luck with that
22:07
<TabAtkins>
Oh, wow, I was being facetious. Interesting.
22:09
<roc>
the problem is that if a site allows SVG image uploads, and has an open redirector, the images can "phone home" in unexpected ways
22:09
<TabAtkins>
Yeah, I understand the issue.
22:09
<TabAtkins>
You want to have a defined and limited contract for what "images" can do.
22:09
<roc>
we have an implicit one now
22:14
<roc>
hmm, maybe we can fix this another way
22:22
<abarth>
can't we just remove SVG images? they're so wrong
22:22
<TabAtkins>
What? No! SVG images are great.
22:22
<TabAtkins>
I use 'em, and love 'em.
22:22
<abarth>
for what? introducing security vulnerabilities?
22:22
<abarth>
:)
22:22
<TabAtkins>
For scalable vector graphics!
22:22
<TabAtkins>
I let you fix the security vulnerabilities.
22:22
<abarth>
why not just use SVG in HTML?
22:23
<abarth>
i've used svg images before too
22:23
<abarth>
bu
22:23
<abarth>
t
22:23
<abarth>
they having all these strange problems
22:23
<TabAtkins>
Can I easily reuse a single image in multiple places on the page with SVG-in-HTML?
22:23
TabAtkins
doesn't know enough.
22:23
<abarth>
TabAtkins: with XBL!
22:23
abarth
hides
22:23
<TabAtkins>
>_<
22:24
<zcorpan>
<use>?
22:24
<TabAtkins>
Anyway, if you can do that, then SVG-in-HTML is fine. If not, then you still want SVG-in-<img>, or a similar solution.
22:24
<TabAtkins>
zcorpan: Does that work?
22:24
<abarth>
i'm not sure how use interacts with being in HTML
22:24
<zcorpan>
TabAtkins: i haven't tried to touch it with a stick, so dunno
22:25
zcorpan
has just heard other people curse over <use>
22:26
<TabAtkins>
All these attributes on <use> confuse and enrage me.
22:26
<TabAtkins>
Oh, wait, I was looking at a decade-old WD.
22:27
<TabAtkins>
Dammit.
22:28
<TabAtkins>
It looks like you could maybe use <svg><symbol/></svg> in the head, and then <svg><use/></svg> in the document? Theoretically, at least.
22:30
<zcorpan>
you can't put <svg> in head
22:30
<AryehGregor>
Why not?
22:30
<AryehGregor>
I guess you want it to break out implicitly.
22:30
<zcorpan>
it implies <body>
22:30
<AryehGregor>
In text/html, anyway.
22:31
<TabAtkins>
Okay, well, <symbol> isn't supposed to ever be rendered, so it should be fine in <body> too.
22:34
<TabAtkins>
Hmm. Do document referenes in inline <svg> work across <svg> elements?
22:46
<zcorpan>
mpilgrim: btw, opera 11 has nicer error messages for form validation, in case you have nothing better to do and feel like updating images in http://diveintohtml5.org/forms.html#validation :)
23:02
<roc>
SVG images are nice but some implementations are rather buggy .... cough cough
23:09
<zcorpan>
TabAtkins: the feed would be more useful if the entries contained the diffs
23:10
zcorpan
subscribes to the commits mailing list instead
23:12
<TabAtkins>
zcorpan: They don't? That's a pretty useless feed, then. I'll complain.
23:14
<zcorpan>
<description>Message posted by Tab
23:14
<zcorpan>
Atkins Jr.. via cvs-syncmail</description>
23:14
<zcorpan>
is all
23:14
<TabAtkins>
Yeah, useless.
23:14
<roc>
I see that abarth is discovering that it's hard to make arguments about complexity in the face of use-cases and an implementation
23:15
<abarth>
w.r.t. SVG images or CSP?
23:15
<roc>
well, both
23:15
<zcorpan>
TabAtkins: ideal would be log message as title and full diff as description
23:15
<abarth>
i know we can't get rid of SVG images :)
23:15
<roc>
I learned that lesson in other places too
23:15
<TabAtkins>
zcorpan: Agreed.
23:15
<roc>
Audio WG for example
23:16
<abarth>
its just a funny artifact of how svg came to the world
23:16
<AryehGregor>
roc, a feature can be too complex for *authors* even if you have use-cases and an implementation.
23:16
<AryehGregor>
Particularly security-related features.
23:16
<roc>
I totally agree
23:16
<roc>
it's just a hard argument to make when authors are saying "no, I really want this"
23:17
<zcorpan>
TabAtkins: well, and which spec being updated in title, too
23:17
<roc>
not all authors, just some set of really smart authors --- the sort who tend to get involved in spec discussions, or work for Google
23:19
<roc>
and who think that being able to implement a professional audio mixing desk on a mobile device as a Web app is a completely reasonable requirement for the first version of a spec, cough cough
23:41
<foolip_>
roc, is there any point in joining that Audio WG?
23:41
<foolip_>
I imagine I'd mostly be saying "no" :)
23:42
<roc>
I dunno
23:42
<roc>
your feedback would actually be useful
23:42
<foolip_>
are you trying to reconcile your super-simple API with Google's filter graph?
23:42
<roc>
no, I don't think anyone's trying to do that
23:42
<foolip_>
phew
23:43
<foolip_>
so inventing a 3rd API then I guess?
23:43
<roc>
I don't know what to do
23:44
<foolip_>
I think I'll join after my vacation and see if there's anything fun happening
23:44
<roc>
we could add support for using Workers to generate audio. that would help with latency issues
23:45
<roc>
but fundamentally, if you have a hard requirement for minimal latency plus maximum possible throughput using the GPU to process samples etc, then sure OK you need to bake every effect under the sun into the browser
23:45
<foolip_>
is there any kind of idea about the model yet?
23:45
<foolip_>
push from script using something like writeSamples, or pull from script via a callback?
23:46
<doublec>
Corban Brook was working on a JavaScript bridge implementing the google api on top of the mozilla api:https://github.com/corbanbrook/audionode.js
23:46
<roc>
I don't think the sample graph guys really care about JS that much
23:46
<roc>
in their world, if you have to process samples with JS you've already lost
23:47
<foolip_>
how low latency is low enough for them?
23:47
<foolip_>
what's the use case?
23:47
<foolip_>
I'm thinking the number 1 use case, all categories, is a NES emulator
23:47
<roc>
audio mixing desk
23:47
<foolip_>
if you can do that, nothing else matters :)
23:47
<roc>
instrument synthesis
23:48
<roc>
real-time spatial effects
23:48
<roc>
for games etc
23:48
<roc>
we can do the NES emulator, that's a non-issue :-)
23:49
<foolip_>
is there no minimum latency caused by the number of samples you have to write at a time or some such?
23:49
<roc>
I don't know
23:49
<roc>
you kinda have to be a super audio guru to participate in that WG
23:50
<roc>
and I don't have time to become one
23:50
<foolip_>
of course, if you have a small enough buffer then having to wait for the JS engine is eventually going to fall apart
23:50
<foolip_>
it's just a question of how small is small enough...
23:51
<roc>
so will C
23:51
<roc>
of course, JS will never be quite as efficient as C
23:51
<foolip_>
don't you have audio cards that you can upload sounds to and program the filters and whatnot in advance?
23:51
<roc>
which is why all features everywhere must be baked into the browser
23:52
<roc>
that's another issue yeah
23:52
<foolip_>
I really don't know, to me an audio card is just a D/A-converter
23:52
<roc>
plus the use of the GPU or another specialized processor
23:52
<foolip_>
processing audio isn't that expensive...
23:53
<roc>
it is if you want to mix 300 channels with spatial effects and more
23:53
<roc>
sure, not all Web developers want to do that
23:53
<roc>
but the Audio WG people do :-)
23:53
<foolip_>
what kind of system is that possible in?
23:53
<doublec>
and they want to do it on browsers embedded on mobile devices
23:53
<foolip_>
don't you need really expensive hardware?
23:54
<foolip_>
or can their filter graph thingy work on anything?
23:54
<roc>
I don't know
23:55
<foolip_>
ok, so I'll join the WG and ramble a bit in 2 weeks
23:55
<foolip_>
please stop them from doing crazy stuff in the meantime :)
23:55
<roc>
I'm not optimistic
23:56
<roc>
the point I was making w.r.t. Adam was that if someone wants a feature, and someone else is providing it, it's really hard to get in the middle and tell them they shouldn't do that
23:56
<roc>
on the grounds of complexity, at least
23:56
<foolip_>
uh, so who's providing?
23:56
<foolip_>
Chrome?
23:56
<roc>
Chrome and Safari
23:56
<roc>
the code is in Webkit
23:57
<foolip_>
"The ones that win are the ones that ship."
23:58
<foolip_>
sucky but true
23:58
<doublec>
mozilla guys are providing too though - in that they've landed and documented an api
23:58
<foolip_>
doublec, they? is it not you?
23:58
<doublec>
foolip_: no, it wasn't the auckland office that worked on it
23:59
<foolip_>
oh, ok
23:59
<doublec>
if it was I'd be pushing hard for it in the working group. As it is I'm not sure what the best approach is.