00:03
<jamesr>
Hixie, <foreignObject> is kind of too broken to hand author stuff today
00:03
<jamesr>
tons of crap is broken with it, so not much content exists using it, so UAs aren't motivated to fix i
00:03
<jamesr>
t
00:05
<Hixie>
what's broken?
00:06
<tabatkins_>
I'm going to run with roc's statement that it seems fine to just elide <foreignObject> for HTML elements.
00:08
<roc>
most of the things that are broken in <foreignObject> are just UA bugs
00:08
<Hixie>
well they're certainly not going to go away if we make it more subtle :-)
00:13
<Hixie>
oh jeez, a sudden influx of i18n bugs
01:04
<jamesr>
roc, agree re: they are UA bugs. the rest of my statement still applies in that case
01:54
<roc>
I think Gecko's foreignObject support is pretty good
01:55
<roc>
our main limitation is that we don't have accelerated compositing for the contents of <foreignObject> yet but we're working toward it
03:02
<annevk>
'I am sorry that an overt adherence to "process" blinds some people to positive gains made when people come together and talk to each other, rather than assuming nothing but malevolent actions' coming from John Foliot, that's an instant classic
03:21
<annevk>
yay, seems my email address transition is more or less working
03:35
<annevk>
apart from public-html-testsuite that is
03:44
<annevk>
why is there no in:spam has:label in Gmail?
03:44
<annevk>
or label:*
03:56
<Hixie>
208
03:56
<Hixie>
er
04:01
<annevk>
Already Reported?
04:04
<Hixie>
line 208
04:24
<annevk>
merging SVG with HTML keeps coming up, but each time it's a less just try it and see what breaks rather than thought through
04:24
<annevk>
merging SVG and CSS is more interesting I think; similarly to how SMIL got in CSS
04:30
<Hixie>
agreed
04:34
<Hixie>
jesus the more i have to deal with these frameworks the more i hate them
04:35
<Hixie>
(io.socket.js being the main culprit here today)
04:35
<Hixie>
or socket.io or whatever it's called
04:43
<annevk>
from Asa Dotzler: "not going to block k9o on un-prefixing but if there are full-screen capabilities missing that we need to have a successful app ecosystem, we'll block on those."
04:43
<annevk>
I guess a proprietary app ecosystem is now fine?
04:46
<annevk>
Hixie: btw, do you have a POP3 account for ian⊙hc that Gmail imports?
04:46
<Hixie>
no
04:46
<Hixie>
i just forward all e-mail directly
04:46
<Hixie>
ian⊙hc is just an alias
04:46
<rniwa>
annevk: moving to new email address?
04:46
<annevk>
Hixie: how did you setup your native client to be able to email from ian⊙hc?
04:47
<annevk>
rniwa: yeah, annevk⊙an
04:47
<Hixie>
annevk: why would that be a problem?
04:47
<annevk>
Hixie: e.g. when trying to do that on my iPhone it asked for a password
04:48
<Hixie>
password for what?
04:48
<annevk>
but maybe I'm looking at the wrong thing
04:48
<rniwa>
annevk: SMTP or POP3 i suppose?
04:48
<rniwa>
annevk: which SMTP/POP3/IMAP server are you using to send those emails?
04:48
<rniwa>
annevk: of gmail? or of annevk.nl?
04:48
<annevk>
rniwa: using Gmail at the moment to email from annevk⊙an
04:49
<Hixie>
oh i don't e-mail from gmail
04:50
<annevk>
ah, I guess I need to setup some kind of SMTP account?
04:50
<Hixie>
honestly if you just want to use gmail your best bet might just be to get a hosted gmail account
04:50
<Hixie>
on annevk.nl
04:51
<Hixie>
so you can just point your mx records at gmail and so on
04:51
<annevk>
I'd rather not have another Google account
04:54
<annevk>
Hixie: anyway, if hixie.ch is just an alias, how do you send email from there?
04:55
<Hixie>
pine doesn't care where you say you send mail from, it just happily puts the from address on there
04:55
<Hixie>
and the dreamhost smtp servers are fine with it because they know it's one of my dreamhost addresses, i presume
04:55
<rniwa>
annevk: which mail client do you use?
04:55
<Hixie>
i occasionally e-mail from gmail, and gmail seems fine with it too, didn't have to do anything special there
04:56
<rniwa>
annevk: i think you normally setup a separate account with gmail's credential for username/password
04:56
<annevk>
Hixie: oh, I had to authenticate it there
04:56
<annevk>
Hixie: are you sure you're not importing a DreamHost POP3 account into Gmail?
04:57
<Hixie>
there is no dreamhost pop3 account to forward
04:57
<Hixie>
so yes
04:57
<annevk>
rniwa: I haven't really decided yet which email client to use; I'm using Opera at the moment; not sure how good it is at handling multiple accounts
04:57
<Hixie>
ian⊙hc is just a forwarding alias
04:57
<Hixie>
there's no inbox behind it
04:57
<rniwa>
annevk: http://support.google.com/mail/bin/answer.py?hl=en&answer=22370
04:57
<Hixie>
same as *@spam.hixie.ch
04:58
rniwa
finally found a manual on this :D
04:58
rniwa
hates page redesigns that moves URLs around :(
04:59
<rniwa>
annevk: honestly, the best mail client for gmail is gmail itself. i've tried several mail clients that try to support gmail features
04:59
<rniwa>
annevk: but none of them support labels well.
05:00
<rniwa>
annevk: it's fine as long as you have only one label at a time
05:00
<rniwa>
but as soon as you start adding 2-3 labels, it gets all confused :(
05:00
<annevk>
rniwa: ah k
05:02
<rniwa>
annevk: i highly recommend "preview pane" in labs
05:02
<annevk>
maybe Gmail is efficient enough
05:02
<rniwa>
annevk: yeah, i think the only reason you want to use mail clients to see emails offline
05:03
<rniwa>
i suppose you would do that much more often than i do as a spec. author though...
05:04
<annevk>
rniwa: does preview pane require a minimum width?
05:05
<rniwa>
annevk: you can do vertical split as well.
05:06
<annevk>
Hixie: I guess the problem is that I don't have a DreamHost SMTP setup as far as I know
05:06
<rniwa>
annevk: (it defaults to horizontal split)
05:06
<Hixie>
mail.annevk.nl
05:06
<Hixie>
is your smtp server
05:06
<Hixie>
i believe
05:06
<Hixie>
oh i guess if you don't have an account at dreamhost for your mail you wouldn't have an account on it though
05:07
<annevk>
right :)
05:07
<Hixie>
can you create a bogus account?
05:07
<rniwa>
annevk: setup a local SMTP server :D
05:07
<annevk>
rniwa: I meant that it doesn't split now when enabled
05:07
<rniwa>
oh :)
05:07
<annevk>
Hixie: yeah I could do that
05:07
<rniwa>
annevk: oh that's weird :\
05:08
<annevk>
or find a better email client that allows you to configure the From address
05:08
<rniwa>
annevk: maybe it doesn't work on Opera :\
05:08
<rniwa>
?
05:08
<annevk>
because per the page from rniwa Gmail is fine with it
05:08
<annevk>
rniwa: ah could be
05:09
<Hixie>
on his iphone? :-P
05:09
<Hixie>
as if apple would allow that :-P
05:09
<rniwa>
annevk: fwiw, i used to run smtp server locally to send my emails.
05:10
<rniwa>
not sure if that's practical anymore though given all the sender id nonsense
05:10
<annevk>
rniwa: same problem in Safari
05:12
<rniwa>
annevk: odd
05:12
<rniwa>
annevk: did you select "horizontal split"?
05:12
<rniwa>
annevk: next to < > buttons on the upper right corner
05:12
<rniwa>
there is a button with 3 lines
05:13
<rniwa>
if you click the downward arrow mark there, you can select "no split", "horizontal split", and "vertical split"
05:13
<rniwa>
annevk: maybe it's "no split" by default. although that'll be extremely silly
05:14
<annevk>
oooh
05:14
<annevk>
rniwa: that is the default
05:14
<annevk>
thanks
05:14
<rniwa>
:(
05:15
<annevk>
works fine in Opera btw
05:17
<rniwa>
annevk: great.
05:17
<rniwa>
oh my... my chrome is on rampage :(
05:17
<rniwa>
GPU accleration
05:17
<rniwa>
:\
05:21
<AryehGregor>
Wait a sec. How can a Google+ e-mail in Gmail allow me to add someone to a circle directly in the e-mail? Do e-mails from Google+ get special privileges in Gmail?
05:21
AryehGregor
smells antitrust.
05:24
<annevk>
AryehGregor: you can do the same from Google search results, no?
05:24
<AryehGregor>
I don't know, maybe. But at least that's dedicated UI. This was part of the e-mail.
05:24
<AryehGregor>
I wouldn't have remarked on it if it were an extra button at the side of the e-mail or something.
05:32
<roc>
I think emails from anyone that Google+ knows about get special Gmail UI
06:03
<annevk>
so http://support.google.com/mobile/bin/answer.py?hl=en&answer=138740 suggests I go to m.google.com/sync on iOS to configure the custom From header, but that page says my device is not supported
06:05
<annevk>
oh
06:05
<annevk>
per help forum it's because the UI language is not English, which I never even set to Dutch to begin with, but I guess that's some IP address sniffing
06:05
<annevk>
boohoo
06:10
rniwa
hates lang. setting per IP address sniffing
06:10
<rniwa>
every time I go to a foreign country, i need to figure out a way to get back to english website :(
06:10
<rniwa>
i don't know who decided it's a good idea.
06:12
<annevk>
anyway, I can now email from my iPhone as annevk⊙an and I don't have a custom/DreamHost SMTP setup yet
06:12
<rniwa>
nice.
06:43
<Ms2ger>
jgraham, yt?
07:05
<annevk>
MikeSmith: so public-html-testsuite is still sending emails to my old address and I think that's the last list
07:06
<annevk>
MikeSmith: I had unsubscribed for a few days now though so something seems broken
07:08
<[tm]>
annevk: away from my PC but will check it when i get back
07:09
<[tm]>
i think the reason is that nobody had sent a message to the list, so the subscriber info had not been updated
07:10
<annevk>
I thought so too, but then last night now emails arrived in my annevk⊙oc inbox
07:12
<annevk>
and I have tried unsubscribing about four times it seems per my sent history
07:17
<[tm]>
ok
07:17
<[tm]>
I'll try to get it figured out tonight
07:31
<Ms2ger>
annevk, as usual, please don't interpret "Asa Dotzler" as "Mozilla" ;)
07:54
<zcorpan>
TabAtkins: woah! let's not change the parser around foreign content to be a radically different design...
08:00
<jgraham>
OK, is this a conspiracy? Just when I want to talk to Ms2ger and AryehGregor, they both leave :(
08:05
<zcorpan>
also, let's fix <svg>.innerHTML = '<g/>' *first*
09:37
<annevk>
while trying to find the Hypertext CG (not a Community) mailing list I found http://lists.w3.org/Archives/Public/public-backplane-comments/
09:37
<annevk>
takes you back
09:52
<zcorpan>
[tm]: you know what would be nice? memorable shortlinks to all the specs in http://platform.html5.org/
09:54
<kennyluck>
That sounds cool indeed.
09:56
<zcorpan>
[tm]: btw, i don't see a link to selectors4
09:56
<annevk>
you mean stuff like platform.html5.org/html and such?
09:57
<zcorpan>
annevk: yeah, but ideally shorter :)
09:58
<kennyluck>
we should have reserved tr.org. http://tr.org/html5 and such
09:59
<annevk>
kennyluck: w3.org is pretty short
10:00
<annevk>
kennyluck: you could e.g. have w3.org/r/xxx or r.w3.org/xxx
10:00
<jgraham>
sp.ec/html ?
10:00
<zcorpan>
jgraham: wfm
10:00
<annevk>
.ec requires 3 characters I think
10:01
<jgraham>
Yeah
10:01
<annevk>
it's also expensive
10:03
<jgraham>
Indeed. websp.ec probably isn't worth it then :(
10:05
<zcorpan>
can we use just html5.org/html etc?
10:07
<annevk>
zcorpan: seems like a pain to maintain
10:09
<zcorpan>
annevk: why?
10:10
<annevk>
zcorpan: because it's at the root of the site so I would be the bottleneck
10:12
<zcorpan>
hmm. would it be possible to have a script somewhere that [tm] can fiddle with that creates redirect rules for the root, or some such?
10:13
<zcorpan>
oh selectors is under css
10:14
<annevk>
not without giving mike access to all of html5.org, which is fine, but currently it's connected to most of my other domains
10:26
<annevk>
so I'd prefer a new domain or a subdomain
10:26
<annevk>
but I'm not sure it's really worth it
10:32
<Philip`>
I thought nobody was meant to call it HTML5 nowadays anyway
10:32
<Philip`>
so html5.org is an archaic domain name
10:34
<annevk>
Philip`: nostalgic
10:34
<annevk>
Philip`: also, html.org was way expensive
10:41
<zcorpan>
html60.org?
10:47
<annevk>
ah yeah, html.org owners rejected a USD 12.000 offer and will consider everything above 20.000
10:48
<annevk>
oh hmm
10:48
<annevk>
html.org is now in the hands of w3.org
10:49
<annevk>
so maybe they put some of that Member money into domain buying fun?
10:49
<smaug____>
that would be odd
10:49
<annevk>
also appears to be using GoDaddy which is somewhat disappointing
10:50
<annevk>
smaug____: http://whois.domaintools.com/html.org shows it's owned by the W3C now
10:50
<Ms2ger>
jgraham, so
10:50
<Ms2ger>
There's two things I'd like to do
10:50
<annevk>
and I have an emailed exchange with the previous owner about the price
10:51
<Ms2ger>
One is { timeout: Infinity } (hmm, maybe that even works right now)
10:51
<Ms2ger>
The other is telling a test "I've given up on you, if you get any more results, don't tell me about them, because I don't care"
10:52
<Ms2ger>
Because I'm afraid of test results ending up reported for the next test
10:52
<Ms2ger>
(ISTR that happening with mochitests)
10:53
<zcorpan>
annevk: you should have bought it and sold it to the w3c :-P
10:54
<zcorpan>
maybe w3c want to buy html5.org as well?
10:54
<zcorpan>
for ONE MILLION DOLLARS
11:04
<zcorpan>
"first version of Web Sockets" http://lists.w3.org/Archives/Public/public-webapps/2012AprJun/0557.html
11:07
<annevk>
quick, what's our position?
11:07
annevk
throws up a coin
11:08
<jgraham>
Ms2ger: I think my solution does that
11:08
<jgraham>
Depending on what you mean by "test"
11:09
<jgraham>
So in your testharnessreport.js you would set explicit_timeout to true
11:09
<jgraham>
That would be line timeout:infinity
11:09
<jgraham>
*like
11:09
<jgraham>
Then when you want to give up on a test you would call timeout()
11:10
<jgraham>
That would set the status of any tests without results to "Not Run", and cause the page to try and report its results in the normal way
11:10
<jgraham>
and prevent anything further happening
11:10
<kennyluck>
TabAtkins, yt?
11:12
<zcorpan>
what does <script defer> do if the script is created after window.onload?
11:14
<jgraham>
zcorpan: From memory, that won't ever get run (per spec)
11:14
<jgraham>
But my memory is not that good
11:14
jgraham
adds it to a list of thinsg to test
11:16
<zcorpan>
is that what ie does?
11:18
<jgraham>
Dunno, I would need to change computers to find out
11:20
<zcorpan>
me too
11:20
<zcorpan>
but i'm curious so maybe i'll check :)
11:21
<jgraham>
http://software.hixie.ch/utilities/js/live-dom-viewer/saved/1513 shows the script running in Gecko and WebKit
11:22
<jgraham>
Oh, but that test is wrong
11:22
<jgraham>
Needs to be parser-inserted
11:43
<tabatkins_>
kennyluck: Here.
11:45
<kennyluck>
TabAtkins, I am wondering if you can list http://lists.w3.org/Archives/Public/www-style/2012Apr/0152 in the list of CSS3 V&U issue list. I have a patch for Gecko pending working group resolution on this → https://bugzilla.mozilla.org/show_bug.cgi?id=751805
11:46
<zcorpan>
jgraham: http://software.hixie.ch/utilities/js/live-dom-viewer/?saved=1514 ie9 runs the script
11:46
<kennyluck>
I noticed that in your Syntax draft, url() can't be escaped, but I guess David wants a working group resolution.
11:47
<jgraham>
zcorpan: Yeah, because defer only has an effect on parser-inserted scripts
11:47
<zcorpan>
oh
11:47
<jgraham>
So the question is whether it is possible to create a parser-inserted script after load fires
11:48
<jgraham>
I thought http://software.hixie.ch/utilities/js/live-dom-viewer/saved/1515 might be close, but I haven't checked what document.close actually does
11:56
<Ms2ger>
jgraham, a timeout() function? Does that exist in code or only in your mind? :)
12:04
<odinho>
In HTML5 spec, is several dt's after another supposed to OR-ed or AND-ed?
12:06
<jgraham>
Ms2ger: Only in my mind
12:06
<annevk>
it's a relationship between <dt> and <dd>, so OR
12:06
<annevk>
but maybe we should document that somehow
12:06
<odinho>
annevk: OK, like I thought.
12:06
<jgraham>
Ms2ger: But it would simply be exposing the internal Tests.prototype.timeout
12:07
<annevk>
other glenn was confused by this too
12:07
<odinho>
annevk: Yes, I tried finding it in http://dev.w3.org/html5/spec/introduction.html#structure-of-this-specification
12:08
<annevk>
at some point I'm going to write "Boilerplate"
12:08
<annevk>
with Boilerplate Conformance, Terminology, etc.
12:08
<annevk>
and then other specs just say "See boilerplate"
12:08
<annevk>
plus changes
12:09
<odinho>
annevk: Would be A+. Your html5.org in itself (just linking a few links to very relevant stuff) is really useful .
12:12
<annevk>
it's getting kind of annoying to copy that stuff all over and then we find something worthy of clarifying and boom, 10 specs need to be updated and kept in sync
12:13
<zcorpan>
annevk: or make Anolis insert the boilerplate
12:14
<MikeSmith>
zcorpan: I'm happy to set up shortlinks wherever -- under platform.html5.org if that's where people would like them
12:15
<annevk>
zcorpan: a) that doesn't make it clear it's boilerplate b) that doesn't make additions to the boilerplate clear
12:16
<zcorpan>
annevk: fair enough
12:17
<jgraham>
hsivonen: AFAICT http://software.hixie.ch/utilities/js/live-dom-viewer/saved/1515 is a bug in gecko
12:17
<Niloy>
is there any plans to standardise webkit's css reflection?
12:18
<annevk>
Niloy: best to look through www-style
12:18
<Niloy>
okay thanks
12:19
<zcorpan>
MikeSmith: cool. only problem is finding a good way to make the urls short enough to be worthwhile :)
12:19
<Ms2ger>
Niloy, yes-ish
12:19
<Ms2ger>
(That's all I know about it, though)
12:23
<Ms2ger>
jgraham, I'd much appreciate the code, too ;)
12:23
<MikeSmith>
zcorpan: could do what annevk said earlier, about making them html.org/foo urls instead; but I currently don't have perms to do that -- would require some changes from annevk first I guess
12:24
<Ms2ger>
MikeSmith, you mean html5.org?
12:25
<MikeSmith>
yeah
12:27
<odinho>
If it's apache, just having a .htaccess with the redirects there would be enough I guess. Someone at the F2F used that for a "personal shortening service", don't remember who. :P
12:27
<jgraham>
Oh, W3C owns html.org? Who knew?
12:27
<MikeSmith>
is node. isEqualNode(foo) supported across browsers?
12:28
<annevk>
jgraham: I did since earlier today :p
12:28
<MikeSmith>
jgraham: I didn't know. Are you sure it does?
12:28
<jgraham>
MikeSmith: Well whois says that
12:28
<MikeSmith>
wow
12:28
<annevk>
I asked Ted about it on www-archive
12:28
<annevk>
fwiw
12:29
<annevk>
it's a) registered by GoDaddy which is bad and b) I'm kind of curious if W3C really paid that much
12:29
<MikeSmith>
I doubt W3C would pay a bunch of money for it
12:30
<annevk>
MikeSmith: I think isEqualNode() is supported yes (not sure about all edge cases)
12:30
<MikeSmith>
OK
12:30
MikeSmith
didn't know
12:30
<annevk>
MikeSmith: yeah, so I tried getting this domain January 2011; guy asked USD 20.000
12:30
<MikeSmith>
wow
12:30
<MikeSmith>
that's nuts
12:31
<annevk>
he said he had an offer for 12.000 but wanted more
12:31
<annevk>
with a typo lol "We will consider all serious propositions abone $20,000"
12:31
<MikeSmith>
maybe Ted ended up claiming it through trademark or something
12:31
<MikeSmith>
heh
12:31
<TabAtkins>
kennyluck: Sorry, mirc wasn't binging me for mentions.
12:32
<annevk>
MikeSmith: ah yeah
12:32
<kennyluck>
TabAtkins, nah, no problem.
12:32
<TabAtkins>
I intend to resolve that so that it's completely impossible.
12:33
<TabAtkins>
In other words, so that url() doesn't allow it (because it's all parsed as a URL), and \url() is an unknown function.
12:34
<TabAtkins>
image() will allow attr(), because it takes a <string> type value (not STRING tokens, like url()).
12:34
<kennyluck>
TabAtkins, I am in favor of that for what it's worth. The whole question is whether \url() can be considered a <url>.
12:35
<annevk>
is \u the same as u?
12:35
<annevk>
because then yes
12:35
<kennyluck>
I think WebKit as well as the formal grammar in CSS2.1 disallow this. Gecko allows this, so is a test case in the CSS2.1 test suite. How weird...
12:36
<annevk>
kennyluck: sounds like a bug in the grammar
12:36
<kennyluck>
This has been raised multiple times throughout the history of www-style, but I guess there was never a clear resolution.
12:37
<annevk>
kennyluck: if the grammar is not {u}{r}{l}'(' ... ')' or something close to that
12:38
<annevk>
allowing @document-\url ... but not background:\url() seems really weird
12:38
<annevk>
now removing escapes from what should be literals...
12:38
<annevk>
I would support that, but it seems a little late
12:39
<annevk>
especially considering the current dynamics of the CSS WG
12:39
<kennyluck>
annevk, David is in favor of what you think I guess. I think TabAtkins is against. I am probably against.
12:40
<kennyluck>
In any case, I need to fix the grammar or the test suite.
12:40
<kennyluck>
s/I/We/
12:48
<annevk>
kennyluck: dbaron? pointer?
12:49
<annevk>
kennyluck: and why would anyone be against removing complexity? compat concerns?
12:49
<kennyluck>
annevk, https://bugzilla.mozilla.org/show_bug.cgi?id=751805#c3
12:49
<annevk>
odinho: fwiw, the question in that Origin header bug is about the Origin header not being required by HTML
12:50
<annevk>
odinho: because HTTP is not required
12:50
<annevk>
odinho: so you could opt to not include the Origin header and it's not entirely clear to the submitter of the bug what the requirements with respect to CORS are then
12:50
<TabAtkins>
kennyluck: No, \url() shouldn't be a <url>. It's an unknown function. You only get a <url> from URL tokens, or from the image() function.
12:51
<annevk>
TabAtkins: why url() not work like any other syntax construct in CSS?
12:51
<annevk>
would /\
12:51
<annevk>
either you handle escapes everywhere or you don't
12:52
<TabAtkins>
Because how does "\url(foo bar ! /* baz */)" work? It definitely *doesn't* act like "url(foo bar ! /* baz */)".
12:52
<zcorpan>
TabAtkins: http://software.hixie.ch/utilities/js/live-dom-viewer/saved/1516 opera and gecko apply it
12:52
<odinho>
annevk: If it's never sending CORS requests, then they will always fail I guess. So in No CORS-mode you'll always get taint, in the CORS modes you'll always fail. -- I find the question a bit strange.
12:52
<kennyluck>
so does IE, but IE is in general just broken concerning CSS escape sequences...
12:53
<odinho>
s/CORS requests/an Origin header/
12:53
<annevk>
odinho: HTML CORS requests are not dependent on supporting CORS or something
12:53
<annevk>
odinho: the dependency chain is somewhat weird
12:53
<annevk>
it'll be fixed once we have fetch v2
12:53
<annevk>
but we don't
12:55
<zcorpan>
TabAtkins: i don't follow what the problem is with "\url(foo bar ! /* baz */)"
12:56
<TabAtkins>
zcorpan: If you're not invoking the special URL token parsing, it looks like FUNCTION IDENT IDENT DELIM COMMENT CLOSE-PAREN.
12:56
<TabAtkins>
And you can lose data there unless you're preserving comments and exact whitespace.
12:57
<TabAtkins>
Even worse is what happens if you don't close the paren.
12:57
<TabAtkins>
close the comment, i mean.
12:57
<TabAtkins>
Then you get different results no matter what, because the URL token doesn't recognize a comment inside of itself.
12:57
<TabAtkins>
But a FUNCTION token would.
12:58
<zcorpan>
but url(foo bar ! /* baz */) isn't valid anyway, is it?
12:59
<TabAtkins>
No, I believe that's valid.
12:59
<TabAtkins>
and corresponds to the url "foo bar ! /* baz */".
13:00
<TabAtkins>
Which isn't a valid url, obviously, but still.
13:00
<kennyluck>
Well, space isn't valid in unquoted url() for what it's worth.
13:00
<TabAtkins>
url(foo bar ! /* baz) is invalid, but it's well-defined and guaranteed to end at the ) character.
13:00
<zcorpan>
browsers don't seem to think it's valid
13:00
<TabAtkins>
Oh, right, sorry.
13:00
<TabAtkins>
That's true.
13:00
<kennyluck>
Seriously, I don't buy any argument here. My patch which changes Gecko to match WebKit has only 10 lines. We just need to agree on something.
13:00
<TabAtkins>
So url(foo/*baz*/).
13:01
<TabAtkins>
Yeah, it's simple in any direction.
13:02
<TabAtkins>
I mean, I suppose I could make the tokenizer look for escaped u, r, and l to invoke the url token parsing. It just doesn't seem necessary.
13:02
<annevk>
but that's what happens everywhere else
13:02
<TabAtkins>
Oh, and here's a more realistic example: url(http://example.com/*)
13:02
<annevk>
it would be highly inconsistent not to look for escapes
13:03
<zcorpan>
TabAtkins: in opera/gecko, url(image?/**/) and \url(image?/**/) both load the image, while url(image/**/) and \url(image/**/) both don't
13:03
<TabAtkins>
This is an invalid url token that ends at the ). It's an unclosed comment if it's a function.
13:03
<zcorpan>
(in live dom viewer)
13:03
<annevk>
what happens for url(x()x)
13:04
<annevk>
or url([)])
13:04
<annevk>
(whether or not it follows the generic error handling)
13:04
<TabAtkins>
Let's see what url([)]) does...
13:05
<zcorpan>
TabAtkins: http://software.hixie.ch/utilities/js/live-dom-viewer/saved/1518 gets applied in opera/gecko, so they don't tokenize a comment there
13:09
<zcorpan>
TabAtkins: i think i'd go with majority and make the tokenizer look for escaped url :-)
13:09
<TabAtkins>
Putting together a decently comprehensive suite real quick.
13:09
<kennyluck>
Anyway, we had http://lists.w3.org/Archives/Public/www-style/2007Dec/0215 from annevk, http://lists.w3.org/Archives/Public/www-style/2010Jul/0499 from Bjoern Hoehrmann, http://lists.w3.org/Archives/Public/www-style/2011Apr/0680 from fantasai all about this for whoever is interested...
13:10
<annevk>
haven't changed my mind on this in 4+ years it seems
13:11
<TabAtkins>
Yeah, I'll do whatever's reasonable and common. Just trying to figure out what it is, comprehensively.
13:11
<TabAtkins>
I did my initial spec from testing in webkit only. ^_^
13:12
<kennyluck>
TabAtkins, well, what you do is matching CSS 2.1 no there is no blame for you.
13:12
<kennyluck>
s/no/so/
13:14
<TabAtkins>
That's true too. Doesn't excuse sticking with 2.1 if it's clear that reality diverges, though.
13:15
<annevk>
I don't think I ever got a reply to that email btw
13:15
<TabAtkins>
The correct answer is "nobody should ever touch the parser itself ever again".
13:15
<annevk>
CSS WG violated W3C Process, news at eleven
13:15
<annevk>
oh, film at eleven, doh
13:17
<Ms2ger>
TabAtkins, hmm, once you have a decently comprehensive suite, you can unprefix!
13:19
<TabAtkins>
Ms2ger: FINALLY
13:20
<annevk>
"PREFIX ALL THE FEATURES"
13:28
<TabAtkins>
All right, everyone non-webkit agrees.
13:29
<TabAtkins>
Escaping the u,r, and l characters are kosher, and it still invokes the URL token parsing.
13:29
<TabAtkins>
So I'll fix that in the spec. Kenny, feel free to patch accordingly.
13:30
<TabAtkins>
That still means that \url(attr(foo)) won't do what that thread wanted, because you get the URL token parsing.
13:30
<TabAtkins>
But like I said, image() handles that fine, because it parses as a FUNCTION instead.
13:31
<kennyluck>
TabAtkins, which means I need to patch CSS2. (my patch changes Gecko to match WebKit)
13:31
<TabAtkins>
Well, I'll patch CSS2.1.
13:31
<TabAtkins>
Don't match Webkit.
13:32
<zcorpan>
patch webkit! :-P
13:32
<TabAtkins>
Yes!
13:32
<kennyluck>
TabAtkins, please do. I already add my mail to the long list of mails about this I mentioned above.
13:32
<Ms2ger>
TabAtkins, only if you have proposed text and a test case ;)
13:34
<kennyluck>
Does anyone have an opinion on whether \u+1234 should be accepted as UNICODE-RANGE or not?
13:34
<TabAtkins>
Ms2ger: I'm writing up an email with proposed text now.
13:35
<TabAtkins>
kennyluck: What's the support?
13:35
<TabAtkins>
That's all I really care about. ^_^
13:36
<kennyluck>
TabAtkins, does any browser support 'unicode-range' already? Otherwise it can't be tested...
13:37
<TabAtkins>
Yeah, I think that's supported pretty widely in @font-face.
13:37
<TabAtkins>
I know that at least *some* browsers support it.
13:37
<zcorpan>
does anyone *use* unicode-range?
13:37
<TabAtkins>
Can't answer that. ^_^
13:38
<zcorpan>
it seemed like a misfeature last time i looked at the spec
13:39
<TabAtkins>
Not my place to say, within the context of css3-syntax.
13:41
<kennyluck>
"The CSS code above worked for me in Chrome and Safari on Mac OS X 10.6. but didn’t work in Firefox or Opera. Nor did it work in IE9 on Windows7"
13:42
<kennyluck>
If that's true, I am pretty sure WebKit follows the CSS 2.1 grammar so there's no support.
13:42
<kennyluck>
But well, I don't know if people will think that's inconsistent.
13:47
<TabAtkins>
kennyluck: You can just use the CSSOM to query the property on the font-face rule, and see if it's been dropped or not.
13:48
<TabAtkins>
It would be pretty dumb if every single browser was opposite (in different ways) in whether they support escapes in URL and UNICODE-RANGE.
13:50
<Ms2ger>
It would be appropriate for the web
14:13
<TabAtkins>
Oh, heh, we took the escaped-url as editorial, and then just forgot to make the edit.
14:14
<TabAtkins>
(It's not editorial, of course. No idea why they thought that.)
14:17
<TabAtkins>
Ah, yup, and the change was covered by uri-015; it's just that the spec doesn't match the test.
14:24
<TabAtkins>
So, annevk, no W3C Process was violated. Just forgetfulness. ^_^
14:26
<zcorpan>
ah, so that's how not to violate the process!
14:28
<TabAtkins>
1. Make promises to edit something.
14:28
<TabAtkins>
2.
14:28
<TabAtkins>
3. Profit!
14:29
<zcorpan>
yeah, we'll put in longdesc. promise.
15:20
<jgraham>
This is the second time today that Ms2ger has quit just as I want to talk to him :(
15:20
<jgraham>
AryehGregor: https://gist.github.com/2636212
15:50
<annevk>
TabAtkins_: mwaha
17:22
<Hixie>
TabAtkins_: do you still think we should add .naturalOrientation to <img> to expose a particular facet of the EXIF data?
17:41
<annevk>
Hixie: trying to get emails back to 1000?
18:01
<Hixie>
annevk: got to hit 0,0,0 sometime this year!
18:02
<annevk>
you still want to proof your abandoned timeline?
18:03
<annevk>
because I think it might have included some assumptions about feature freeze too :)
18:05
<annevk>
http://w3cmemes.tumblr.com/post/22659085815 ref?
18:09
<annevk>
MikeSmith: is sysreq⊙wo the place to bug with TLS issues?
18:09
<MikeSmith>
annevk: yup
18:10
<annevk>
ooh
18:10
<annevk>
site-comments is public
18:10
<annevk>
I'll use that
18:10
<MikeSmith>
yeah
18:10
<MikeSmith>
that's better
18:27
<MikeSmith>
annevk: your old address is removed from the testsuite list now
18:27
<MikeSmith>
I had you in the list under "Non auto removable"
18:28
<jgraham>
annevk: You want an example of an example where where no-one remembers whether the example is an example of an in joke or not?
18:29
<annevk>
MikeSmith: ah thanks
18:29
<annevk>
jgraham: too much xzibit?
18:30
<annevk>
fast ij is fast http://lists.w3.org/Archives/Public/site-comments/2012May/0004.html
19:05
<Hixie>
annevk: not especially, i just want to actually get the feedback to zero because that's _always_ the goal. :-)
19:06
<WeirdAl>
yeah, those annoying commenters ;)
20:18
<Hixie>
abarth: what's the status of http://wiki.whatwg.org/wiki/Meta_referrer ?
20:21
<othermaciej>
Hixie: it's in WebKit (though only shipping in Chrome and we might turn it off for Safari) and Mozilla at least at one point allegedly said they'd implement
20:21
<Hixie>
k, guess i'll wait a bit longer then
20:21
<othermaciej>
Hixie: I'm not keen on the existence of the "always" value
20:22
<othermaciej>
Hixie: do you have an opinion?
20:22
<Hixie>
what's the use case for 'always;?
20:22
<othermaciej>
I do not know
20:22
<Hixie>
trackback, according to the page
20:22
<Hixie>
"A blog hosted over HTTPS might wish to link to a blog hosted over HTTP and receive trackback links"
20:23
<Hixie>
i assume it means "send trackback links"
20:23
<Hixie>
it is kinda sad that moving to https kills referer
20:24
<othermaciej>
the reason for it doing so is historical and weird, but it's created an expectation of referer privacy for at least some users
20:24
<othermaciej>
so for example some people promote using google https search to get better privacy on the search queries used to reach a site
20:24
<Hixie>
yeah
20:24
<othermaciej>
use of the "origin" value does not seem to break that assumption, but "always" would
20:25
<Hixie>
(not sure that actually works btw)
20:25
<othermaciej>
there may well be back channels for all I know
20:26
<annevk>
Google Analytics and Google Webmaster tools provide some info in how users get on your site, including search queries
20:26
<annevk>
and links on search redirects are often redirects
20:26
<annevk>
search results :)
20:28
<othermaciej>
I've heard rumors there is interest in using the "origin" meta referrer value for SSL search but I haven't heard of anyone wanting to use "always"
20:28
<jgraham>
I think people at Opera might have had concerns about "always" too. Although it was very informal discussion, and "people at Opera" might have just been "me" or something
20:28
<othermaciej>
so I would guess search providers care about whatever perceived privacy is gained by not sending the referrer
20:29
<annevk>
but they care about origin so they can promote their existence?
20:29
<othermaciej>
apparently, yes
20:30
<othermaciej>
or at least, they'd like content providers to know what traffic is being sent by their search engine, even if they don't know the exact query in any given case
20:30
<annevk>
because other than that (or CORS usage which does not apply here) it seems like a waste of bandwidth
20:31
<othermaciej>
the desire to send origin-as-referrer seems reasonable to me, though I'm in no position to explain anyone's business case for it
20:31
<annevk>
I guess the query is less and less exposed anyway with the search as you type stuff
20:32
<Hixie>
personally i'd be fine with not ever sending referer, if we send origin
20:32
<Hixie>
origin is useful for security
20:32
<Hixie>
my understanding is that "SEO" people want to know what keywords are used to get to their site
20:32
<Hixie>
(origin can also be a security/privacy problem itself, of course)
20:32
<Hixie>
(e.g. leaking intranet host names)
20:33
<annevk>
not sending Referer will likely break a number of sites unfortunately
20:33
<Hixie>
jesus, the number of ways that the <object> element's algorithm can be invoked has become ludicrously long
20:33
<annevk>
that use it to determine whether or not to display an image and such
20:35
<Hixie>
anyone know if <embed> elements that are display:none also get disabled, like <object> elemnets?
20:35
<Hixie>
oh look at that
20:35
<Hixie>
the test that roc was using to test <object> actually tests <embed>
20:35
<Hixie>
new question...
20:36
<Hixie>
anyone know if <object> elements that are display:none also get disabled, like <embed> elements? :-)
20:42
<jgraham>
Hixie: I seem to recall it matter for compat.
20:42
<jgraham>
That doesn't help much though :)
21:15
<jamesr>
Hixie, in WebKit the behaviors are the same
21:15
<Hixie>
that's what i went with too
21:15
<jamesr>
Hixie, for <object> vs <embed> re display:none
21:15
<Hixie>
more or less
21:15
<jamesr>
Hixie, for <object> vs <embed> re display:none
21:15
<Hixie>
more or less
21:15
<Hixie>
yeah
21:15
<jamesr>
queue a task, eh?
21:16
<jamesr>
which task source? some new one?
21:16
<Hixie>
same one as for other things that trigger the plugin on/off code
21:16
<jamesr>
(wondering if ordering matters vs other tasks)
21:16
<Hixie>
the DOM manipulation task source.
21:16
<Hixie>
i guess that's bad
21:16
<Hixie>
hm
21:17
<jamesr>
i'm not sure we are super careful about that (making sure stuff is in the right task source)
21:17
<jamesr>
not sure anyone is, to be honest. how tested is it?
21:17
<Hixie>
not
21:17
<Hixie>
:-)
21:18
<jamesr>
normally you just toss everything into one queue and it all just works
21:18
<jamesr>
where did you stick this new text?
21:19
<Hixie>
for <embed> it's in the "potentially active" definition and there's an open bug on making sure it is triggered on a task and not sync
21:19
<jamesr>
i may want to jack some of it for a css animations/transitions proposal
21:19
<jamesr>
what section #?
21:19
<Hixie>
and for <object> it's before the huge algorithm
21:19
<Hixie>
uh
21:20
<Hixie>
4.8.3 and 4.8.4 apparently
21:20
<jamesr>
found it (ctrl-f for the win)
21:20
<jgraham>
No, ctrl-f for find is a horrible key combination
21:21
<jgraham>
The Opera/Firefox setup is much nicer
21:21
<jamesr>
actually i don't see this. are your changes on the whatwg.org version yet?
21:21
<Hixie>
should be
21:21
<Hixie>
that's the first version that gets regenned
21:21
<Hixie>
yup
21:22
<Hixie>
http://html5.org/tools/web-apps-tracker?from=7099&to=7100 is the diff
21:22
<jamesr>
aha
21:23
<Hixie>
actually i guess using the dom task source is ok, since we're "guaranteed" that the layout will be redone when the event loop next spins [or i guess when the next time you call an attribute that has to do it synchronously], so it's a well-defined time
21:23
<jamesr>
Hixie, "chnges" in the diff
21:23
<Hixie>
thanks will fix
21:24
<Hixie>
oh oops, i missed <embed> in the last checkin
21:25
<jamesr>
explains that