00:01
<jamesr_>
roc, even if we (chrome team) want to do new features this way (channel restrictions but no prefix), that's generally not sufficient for us to actually do it unless apple agrees
00:02
<roc>
I see, they might be worried about landing stuff unprefixed on any branch?
00:03
<jamesr_>
right, roughly. we don't have separate SVN branches for our releases, for instance
00:03
<jamesr_>
we land stuff in trunk and then we cut release branches off of trunk and they advance through the channels as a numbered branch
00:03
<roc>
Can you easily support having both prefixed and unprefixed aliases with independent enabling/disabling? I assume so
00:04
<roc>
anyway, if it hasn't been discussed on webkit-dev yet, I hope it can be soon
00:05
<jamesr_>
we've talked about this at webkit meetups, not sure if it's been discussed on the mailing list
00:05
<jamesr_>
i know simon and tab are flying/will fly soon back from paris
00:05
<jamesr_>
and they'll probably both have input
00:05
<roc>
none of this affects what we have to do for existing stuff of course
00:26
<BenoitRen>
Leaving. Laters!
04:54
<josephg>
hey guys
04:55
<josephg>
does anyone here know anything about mutation observers?
04:55
<josephg>
I'm playing with them in chrome canary, but they're quite buggy… I wonder if I should file the bugs I'm finding.
06:29
<jamesr_>
josephg, you mean the new API
06:29
<jamesr_>
josephg, ?
06:29
<jamesr_>
josephg, i don't know much about them in particular, but you should definitely file bugs!
06:29
<jamesr_>
it's new stuff and some rough edges are probably expected
06:29
<jamesr_>
bugs.webkit.org, minimal testcases preferred but not required
06:30
<jamesr_>
josephg, you can cc me (jamesr⊙co) if bugzilla will let you do that
06:30
<jamesr_>
cheers
07:16
<josephg>
jamesr_: yeah the new api. I filed a bug on chromium - should I file a bug on webkit as well?
07:29
<josephg>
jamesr_: http://code.google.com/p/chromium/issues/detail?id=113584
07:48
<mhausenblas>
hsivonen, congrats for http://validator.w3.org/nu/ - KUTGW!
08:02
<zcorpan>
Hixie: why did https://www.w3.org/Bugs/Public/show_bug.cgi?id=15941 get in the "other Hixie drafts" bucket?
08:16
<zcorpan>
"Instead of coding against a single CSS specification developers will need to code against changing vendor prefixes." http://www.webmonkey.com/2012/02/webkit-isnt-breaking-the-web-you-are/
08:16
<zcorpan>
of course, -webkit-foo is stable and the spec is the one that's changing
08:17
<zcorpan>
i don't understand the argument "it'll break the standards process"
09:12
<jgraham>
zcorpan: People seem to believe as an article of faith that -vendor- is needed to allow innovation in CSS (but not, apparently in other areas like HTML). The desire to not critically examine that particular belief is the number 1 cause of tortured, convoluted, and often nonsensical, reasoning in this whole discussion.
09:15
<zcorpan>
i see
09:17
<annevk5>
so someone wants to advertise a commercial good on blog.whatwg.org?
09:17
<annevk5>
is that okay?
09:17
<annevk5>
in particular: http://alaramills.com/store.html
09:18
<annevk5>
does not really strike me as something we want to do
09:18
<jgraham>
It sounds kind of not-OK to me
09:21
<GlitchMr>
Not interested in this Periodic table of HTML5 elements
09:23
<GlitchMr>
It seems less useful than for example http://www.smashingmagazine.com/2009/07/06/html-5-cheat-sheet-pdf/ (if you want something printable of course :P) and it's paid...
09:24
<GlitchMr>
Artistically, I like the format of it, but that's all.
09:26
<jgraham>
I think you can stop advertising them in the irc logs now ;)
09:26
<annevk5>
thanks for the feedback
09:26
<annevk5>
the samshingmag is free jgraham
09:26
<jgraham>
No, I meant the other thing
09:27
<jgraham>
I imagine that "it seems less useful than" will still make it more likely that people will investigate the link :)
09:29
<zcorpan>
oh, oh, but surely http://simon.html5.org/html-elements is the most useful of all! and FREE!!
09:30
<annevk5>
anyway I declined
09:31
<zcorpan>
I ALSO SELL VIAGRA PILLS
09:31
<zcorpan>
what? i didn't say anything
09:31
<GlitchMr>
Sorry, you were logged in IRC logs...
09:44
<annevk5>
cats
09:44
<Ms2ger>
dogs
09:44
<jgraham>
pigs, but after 9 months you slaughter them, and then bacon.
09:45
<annevk5>
btw
09:45
<annevk5>
tomorrow is my talk on XML5
09:45
<annevk5>
it'll be a bit over 20 minutes so there's not too much time
09:46
<annevk5>
anything I should cover?
09:54
<Ms2ger>
Vendor prefixes
09:57
<annevk5>
dude
09:58
<Ms2ger>
Yes dude?
09:58
<annevk5>
XML already has its own prefix hell
09:58
<annevk5>
no need for making it worse
09:59
<Ms2ger>
We could start using namespaces called http://www.mozilla.org/newlayout/xml/parsererror.xml
09:59
<Ms2ger>
Oh wait, we do
09:59
<jgraham>
Ms2ger: Oh may, you nailed it.
09:59
<jgraham>
*man
09:59
<Ms2ger>
Opera must be really evil to implement that instead of http://www.opera.com/newlayout/xml/parsererror.xml
09:59
<jgraham>
Vendor prefixes have failed because they're not URIs!
09:59
<Ms2ger>
jgraham++
10:00
<annevk5>
making them URLs might actually solve the problem
10:01
<jgraham>
annevk5: Not if your argument is "they would be hard to type"
10:01
<jgraham>
Because no one actually types CSS anymore
10:01
<jgraham>
Not the cool kids anyway
10:03
<annevk5>
I was going for too confusing
10:03
<annevk5>
but I was also mostly joking
10:03
<jgraham>
:)
10:04
<annevk5>
the main problem with this XML5 talk is that I'm not sure we want the hassle of changing XML anymore
10:04
<Ms2ger>
Heh
10:04
<Ms2ger>
"XML5: I dunno if we want this"
10:05
<jgraham>
Well if we had a non-sucky XML it might solve the problems that the Web Components people are having with templates, or something
10:05
<annevk5>
if XML has to become some kind of authoring format used by web developers I think XML5 is pretty much required
10:05
<annevk5>
but it seems we moved past that and nobody cared
10:06
<Ms2ger>
We have XML5, no?
10:06
<annevk5>
jgraham: well yeah, that's certainly true, and SVG/Math and Components might have provided a compelling story for it
10:06
<Ms2ger>
It's <br /> sent as text/html
10:06
<annevk5>
but now all is text/html anyway
10:06
<annevk5>
Ms2ger: only works in Belgium
10:07
<Ms2ger>
lulz
10:08
<annevk5>
whoa
10:08
<annevk5>
http://arstechnica.com/tech-policy/news/2012/02/jury-rules-that-eolass-interactive-web-patent-is-invalid.ars
10:08
<annevk5>
can we countersue now?
10:08
<annevk5>
we implemented all kinds of silly techniques iirc
10:09
<Ms2ger>
Alright
10:09
<Ms2ger>
If anybody said something useful in an email in my inbox that has "prefix" in the subject, send it again
10:10
<Ms2ger>
And add "responsive" to that
10:14
<jgraham>
YOu didn't like my email "Vendor prefixes for responsive web design nirvana" then?
10:19
<annevk5>
Ms2ger: just fix all the DOM bugs, mkay
10:21
<annevk5>
seems like it is time to fly
10:21
<annevk5>
bah
10:22
<annevk5>
from -6 to -13
10:22
<annevk5>
and there's hail
10:22
annevk5
wants to go back on vacation
10:28
<Ms2ger>
jgraham, I dunno if I liked it, it's moved to /dev/null
10:40
<MikeSmith>
annevk5: right after you got back you pinged me about the thing
10:40
<MikeSmith>
but I forgot what it wa
10:40
<MikeSmith>
was
10:41
<MikeSmith>
we were going to do something but we had to wait for some consensus call from webapps WG
10:41
<MikeSmith>
or www-dom list or something
10:41
<MikeSmith>
and I was all ready to do it
10:41
<MikeSmith>
but I forgot what it was I was all set to do
10:42
<MikeSmith>
haha that posters costs $47.99
10:51
<MikeSmith>
yee hah for CORS for XHR in IE10
11:08
<MikeSmith>
https://twitter.com/#!/ryah/status/167743028637872128
11:10
<zcorpan>
i saw a tv commercial the other day, for a smartphone. it included "internet, email and facebook"
11:11
<MikeSmith>
heh
11:11
<MikeSmith>
zcorpan: they know their market I guess
11:11
<MikeSmith>
annevk5: https://twitter.com/#!/mnot/status/167792607890644992
11:12
<zcorpan>
let's change CORS to suck less! wait, first we need to prefix all implementations...
11:28
<bga>
new regexp engine http://code.google.com/p/brre/
11:29
<MikeSmith>
Object Pascal
11:29
<MikeSmith>
cool
11:59
<Taggnostr>
I'm looking for broken html pages with lot of markup errors to test a parser, does anyone know where can I find it? (does the html validator at w3.org keep an high-score list of the sites with most errors?)
12:00
<Ms2ger>
Maybe try google.com?
12:00
<MikeSmith>
Taggnostr: Alexa keeps a list like that
12:00
<Taggnostr>
I found http://www.trentmueller.com/Top-10-Websites-with-the-Worst-HTML_Article/
12:01
<Taggnostr>
but those sites now have less errors
12:01
<Taggnostr>
MikeSmith, a list of sites with broken html?
12:01
<jgraham>
Taggnostr: http://code.google.com/p/html5lib/source/browse/#hg%2Ftestdata%2Ftree-construction
12:02
<MikeSmith>
http://www.alexa.com/topsites
12:02
<jgraham>
heh
12:04
<zcorpan>
MikeSmith: do the stars indicate the brokenness?
12:04
<MikeSmith>
haha
12:04
<Taggnostr>
I already tested the parser with 1000 popular sites, but I was looking for websites like http://www.dokimos.org/ajff/
12:04
<MikeSmith>
number of errors per page is not a great metric
12:04
<MikeSmith>
it's the quality of the errors that matter
12:05
<MikeSmith>
any jackass can ring up the error meter just by putting a bunch of &s or such
12:05
<MikeSmith>
Taggnostr: oh man
12:05
<MikeSmith>
how'd you find that one?
12:06
<MikeSmith>
this is gold
12:06
<Taggnostr>
googling for "worst websites" or something similar
12:06
<MikeSmith>
the security warning at the top is genius
12:06
<Ms2ger>
Forget all the other browsers and
12:06
<Ms2ger>
down with the Web 2.0 net police.
12:06
<Taggnostr>
but it only has 375 errors on the main page
12:07
<MikeSmith>
Error: meta element between head and body.
12:07
<Taggnostr>
(the people who signed the guestbook seem to like it though)
12:07
<MikeSmith>
that's the kind of thoughtful error I'm talking about
12:08
<Taggnostr>
I assume that a website made with word and photoshop and put together with frontpage has enough creative errors
12:10
<jgraham>
Taggnostr: Seriously, have you tried the tests I linked to
12:10
<jgraham>
?
12:10
<jgraham>
If you pass those you shoud do fine on websites
12:11
<Taggnostr>
I'm looking at them, but I was searching for real-world pages (it's also easy for me to wget them and use them in my tests)
12:11
<Taggnostr>
thanks for the link btw!
12:11
<jgraham>
Passing the tests is much more likely to be helpful than finding real world pages
12:11
<jgraham>
They are much easier to debug too :)
12:12
<jgraham>
The only thing that they really lack is crazy-deep nesting and so on
12:13
<zcorpan>
Taggnostr: what do you use the parser for?
12:13
<Taggnostr>
it's the Python built-in parser
12:13
<zcorpan>
ah
12:14
<zcorpan>
well i can only concur with jgraham, you really want to pass the testsuite
12:14
<Taggnostr>
my goal right now is to make it able to parse everything without errors, the next goal is to make it parse everything as correctly as possible (following the html5 standard)
12:14
<jgraham>
I don't understand the difference between those goals
12:15
<zcorpan>
just implement the parsing algorithm in the spec, it covers all cases
12:15
<Taggnostr>
right now when the parser finds invalid markup it just raises an error and give up parsing, and there's nothing you can do
12:15
<jgraham>
Unless you consider "without error" to simply mean "not throwing". In which case lambda html:"" meets the goal
12:16
<Taggnostr>
and I'm changing that so that it tries to figure out what to do with the invalid markup and keep going till the end
12:16
<Ms2ger>
You're modifying an existing parser?
12:16
<Taggnostr>
zcorpan, I can't start from scratch, so I'm adapting what I have to get closer to the specs gradually
12:16
<Taggnostr>
yes
12:16
<Ms2ger>
I... wouldn't recommend that
12:17
<zcorpan>
why can't you start from scratch?
12:17
<Ms2ger>
I don't know of anybody who's done that
12:17
<Taggnostr>
well, so far I managed to have it parse a list of 1000 pages successfully with 0 errors, so it's going quite well
12:18
<Taggnostr>
zcorpan, because I could just use html5lib if I wanted a python parser that follows the specs
12:18
<jgraham>
Pretty sure we had this conversation before
12:18
<Taggnostr>
we did
12:19
<jgraham>
But if I were you I would wrap html5lib in an API that looks more like the stdlib
12:19
<jgraham>
Rather than trying to make the internals of the stdlib more like html5lib
12:19
<zcorpan>
so you want something that is closer to the spec, but still isn't 100% compliant?
12:20
<Taggnostr>
but the goal is providing a better html parser in the stdlib, and improving the existing one is the only viable solution (replacing it with html5lib is not really an option)
12:20
<Taggnostr>
zcorpan, I want to get as close as 100% compliant as possible
12:21
<zcorpan>
why is replacing it with html5lib not an option?
12:21
<Taggnostr>
but if it does the wrong thing with some obscure corner cases it probably doesn't matter too much now
12:23
<Taggnostr>
zcorpan, for several reason, if html5lib is added to the stdlib it will have to follow the python release schedule (so you have to wait many months between releases), it needs someone willing to maintain it, it needs to have a compatible license, it has to be "good enough" to get in the stdlib and replace the existing one and have everyone to switch and so on
12:24
<jgraham>
So it would get a faster release schedule than today?
12:24
jgraham
has neglected html5lib recently
12:25
<jgraham>
The license shouldn't be a problem
12:25
<Taggnostr>
how often do you make html5lib releases?
12:25
<Taggnostr>
for python is more or less 16 months for new features and maybe 4 for bug fixes
12:25
<jgraham>
Well, uh, the last one was a few years ago. It is horribly out of date compared to trunk. This is somewhat embarassing
12:26
<Taggnostr>
are people even using it?
12:26
<jgraham>
It turns out that at any given moment there is always something more fun to do than make a release
12:26
<jgraham>
Yeah, we get a steady flow of bug reports
12:27
<jgraham>
and Mozilla use it in their html sanitizer, for example
12:27
<Ms2ger>
Philip`_, so about your canvas tests...
12:27
<Ms2ger>
jgraham, oh?
12:27
<jgraham>
(most of the bug reports as Invalid, I should say)
12:27
<jgraham>
Ms2ger: The one you use on your websites
12:27
<Taggnostr>
do they use the development version then?
12:27
<jgraham>
Nothing in the browser ofc
12:27
<Ms2ger>
Oh
12:27
<Ms2ger>
Those people use git
12:27
<jgraham>
And for that they are to be praised
12:27
<jgraham>
:p
12:28
<jgraham>
Taggnostr: I am not sure. A release is long, long overdue. And if I could do one right away I think I would now feel guilty enough to do one
12:29
<jgraham>
But we will see how the feeling persists to this evening
12:29
<Ms2ger>
jgraham, today isn't do what the hell you want day?
12:30
<Taggnostr>
these are some numbers about Python html parser btw: http://dpaste.com/699552/
12:30
<jgraham>
Nope :|
12:31
<Taggnostr>
the percentage represent the pages parsed till the end with no errors
12:31
<Taggnostr>
(with a sample of ~1000 pages)
12:34
<Philip`_>
Taggnostr: Have you tried parsing e.g. a PDF file and seeing if that hits any errors?
12:34
<Taggnostr>
nope
12:35
Philip`_
vaguely remembers a few difficulties with such files when parsing lots of random pages, though maybe that was just because he hadn't implemented a length limit before then
12:38
<Philip`_>
Taggnostr: If you want a larger selection of real-world pages to test, http://www.dotnetdotcom.org/ has something like half a million
12:40
<Taggnostr>
thanks
13:18
<zcorpan>
hmm. should track bugs be in the texttracks CG component?
13:19
<hsivonen>
sigh. http://www.change.org/petitions/microsoft-mozilla-opera-dont-make-webkit-prefixes-a-de-facto-standard
13:21
<zcorpan>
we won't. it already is. we'll make it a de jure standard.
13:21
<zcorpan>
that page doesn't load for me btw
13:22
<hsivonen>
WFM in Opera
13:22
<zcorpan>
ah now it loaded
13:24
<hsivonen>
glad to see PPK calls the BS on the call for action
13:26
<Ms2ger>
annevk5, so, what do you think about the DOMTokenList / space separated token list thread?
13:35
<zcorpan>
i don't follow how PPK recognizes that the problem lies with prefixes, yet proposes other prefixes
13:36
<zcorpan>
having -alpha-foo, -beta-foo, foo and -webkit-foo seems worse than just foo and -webkit-foo
13:49
<jgraham>
zcorpan: I refer you to what I said ~4h 37m ago
13:50
<jgraham>
Which really applies to -anything-, not just -vendor-
13:52
<zcorpan>
yeah
14:19
<MikeSmith>
"specs are not magical"
14:19
<MikeSmith>
sad but true
14:19
<MikeSmith>
I wish they were magical
14:33
<hsivonen>
amen to http://krijnhoetmer.nl/irc-logs/whatwg/20120210#l-693
14:37
<MikeSmith>
"This is not a game where the objective is to win independent of the merits of the arguments. This is all about the merits of the arguments."
14:37
<MikeSmith>
well said
14:37
<MikeSmith>
http://lists.w3.org/Archives/Public/public-html/2012Feb/0110.html
14:38
<MikeSmith>
that should be put word-for-word into the W3C process doc
14:39
<Ms2ger>
Along with "longdesc is broken, don't waste your and my time on it"
14:40
<soapyfish>
Ahoy!
14:43
<karlcow>
MikeSmith: missing mushrooms?
14:43
<jgraham>
I wonder by what factor the number of characters in emails about longdesc attributes outweighs the number of characters of useful text pointed to by longdesc attributes.
14:43
<karlcow>
cf http://krijnhoetmer.nl/irc-logs/whatwg/20120210#l-1029 ;)
14:44
<MikeSmith>
karlcow: yeah, but I make my own substitutes -- windowpane
14:44
<karlcow>
heh
15:10
<annevk>
MikeSmith: it was about sending email for resolved bugs on www-dom
15:11
<annevk>
weather in Prague btw is better than expected, as long as you don't go outside for more than ten minutes
15:13
<MikeSmith>
annevk: oh so I did that already
15:13
<MikeSmith>
I recommend you go for the lulz in your talk
15:13
<jgraham>
annevk: So you mean the weather *indoors* in Prauge is better than expected
15:13
<jgraham>
?
15:14
<Ms2ger>
Will the talk be recorded?
15:16
<annevk>
I think there might be live broadcast even
15:16
<annevk>
not sure though
15:17
<annevk>
MikeSmith: cool that it's already done!
15:19
<MikeSmith>
"Rorschach test for standardistas" is good
15:21
<karlcow>
Roooarrrschach test even
15:23
<Taggnostr>
http://www.w3.org/TR/html5/tokenization.html#end-tag-open-state links to "tag name state", and "tag name state" includes attributes. Does that mean that <foo></foo some="attr"> is not a parsing error?
15:24
<MikeSmith>
Taggnostr: that's a parsing error
15:24
<MikeSmith>
Taggnostr: is the default python parser written in python?
15:24
<Taggnostr>
MikeSmith, yes
15:25
<Taggnostr>
how is that a parsing error?
15:25
<MikeSmith>
because it is?
15:25
<MikeSmith>
I don't have the spec open
15:25
<Taggnostr>
if I follow the page, after </ there's "f", so it's the second entry that redirects to "tag name state"
15:26
<MikeSmith>
but it might help you to read the code of the validator.nu/Mozilla parser
15:26
<MikeSmith>
the source
15:26
<MikeSmith>
lemme get you a URL
15:26
<MikeSmith>
it's an error-reporting parser
15:26
MikeSmith
can't remember if html5lib is error-reporting
15:26
<Philip`_>
Taggnostr: I think the error occurs when the tokeniser emits the end tag token
15:26
<MikeSmith>
Taggnostr: when in doubt I read hsivonen code
15:26
<Philip`_>
since the spec says it's an error to emit end tag tokens that have attributes
15:27
<MikeSmith>
rather than the spec
15:27
<Taggnostr>
Philip`_, is that after the parsing?
15:27
<MikeSmith>
I find hsivonen code much easier to follow and it has comments that reference the spec
15:27
<Taggnostr>
MikeSmith, I think html5lib is error reporting
15:27
<Philip`_>
"When an end tag token is emitted with attributes, that is a parse error." (http://www.whatwg.org/specs/web-apps/current-work/multipage/tokenization.html#tokenization)
15:28
<MikeSmith>
Taggnostr: ok
15:28
<MikeSmith>
Taggnostr: http://hg.mozilla.org/projects/htmlparser/file/default/src/nu/validator/htmlparser/impl is worth perusing
15:28
<Taggnostr>
Philip`_, so I should parse all the attributes in the end tag normally, and then simply ignore them and emit only the tag name?
15:29
<Philip`_>
The tree construction algorithm never looks at the attributes of an end tag token, so you don't need to actually bother storing them when tokenising
15:29
<Taggnostr>
but I have to go through them in order to figure out where the tag ends
15:30
<Philip`_>
Yeah, I think you need to at least partially tokenise them, so you can handle </foo bar=">" baz> correctly
15:31
<Taggnostr>
yep, so in that case I'll just emit a foo, and ignore the rest
15:31
<Philip`_>
Storing the attributes then ignoring them later is presumably not going to hurt efficiency in practice (compared to having the tokeniser skip over them without storing anything), because almost nobody puts attributes in end tags in practice
15:32
<Philip`_>
but the spec doesn't care how you implement it as long as the visible end result (i.e. the DOM tree) is the same as what the spec says it should be
15:32
<Taggnostr>
there's also the case of e.g. </li<ul>, I think this should emit 'li<ul' as token name
15:33
<Ms2ger>
Indeed
15:33
<Philip`_>
By "should", do you mean you think the spec says that, or you think the spec ought to say that?
15:33
<Taggnostr>
the python "parser" just does tokenization, it doesn't build the tree
15:33
<Ms2ger>
The spec does say that, IIRC
15:33
<Taggnostr>
I mean that if I read them correctly that's what I should emit
15:34
<Philip`_>
That does seem to be the case
15:35
<Taggnostr>
MikeSmith, the code you linked doesn't seem too clear to me
15:35
<MikeSmith>
Taggnostr: so what does python use for tree building by default?
15:35
<MikeSmith>
Taggnostr: blame hsivonen
15:35
<MikeSmith>
it conforms to the spec anyway
15:35
<MikeSmith>
most of the time the spec is very clear
15:35
<MikeSmith>
as long as you know where to look
15:35
<MikeSmith>
problem is, some of the stuff is kind of spread around in the spec
15:36
<MikeSmith>
and once you know where it is it's clear
15:36
<Taggnostr>
MikeSmith, there's nothing in the stdlib for that, one option is BeautifulSoup, otherwise there are other alternatives like lxml and html5lib
15:36
<MikeSmith>
ah
15:36
<MikeSmith>
yeah
15:36
<Taggnostr>
MikeSmith, apparently all that I need is in http://www.w3.org/TR/html5/tokenization.html , but with all those states it might get a bit confusing to follow
15:36
<Philip`_>
One problem with only doing tokenisation and not building the tree is that it's hard to parse stuff like <script><div></script> properly, since it's not obvious when you should switch to the spec's 'script data state' or whichever state it is
15:37
<MikeSmith>
Taggnostr: be glad I guess that you don't have to implement tree-building
15:37
Philip`_
doesn't know whether "hard" means "impossible" or merely "not directly specified"
15:37
<Taggnostr>
luckily that's not my problem, for that I would emit a start 'script', a start 'div', and an end 'script'
15:38
<jgraham>
Taggnostr: Wow, really?
15:38
<Taggnostr>
jgraham, yes
15:38
<MikeSmith>
Taggnostr: fwiw seems like you are doing the wise thing by asking here instead of just trying to read teh spec in isolatin
15:38
<Philip`_>
What about <circle/><svg><circle/></svg> ?
15:39
<Taggnostr>
iirc <circle/> is seen as start+end, and the default implementation emits a start and an end
15:39
<Taggnostr>
so start circle, end circle, start svg, start circle, end circle, end svg
15:40
<Taggnostr>
jgraham, why are you surprised?
15:40
<Philip`_>
It should be more like start-circle, start-svg, start-circle, end-circle, end-svg, to match what a full HTML5 parser would do
15:41
<Philip`_>
What about <script><!--</script> ? (Should be start-script, text-"<!--", end-script; otherwise some pages will get all of their content slurped into the comment, which is probably undesirable)
15:41
<Taggnostr>
I think even <br/> emits start-br, end-br (unless you override the start/end method
15:42
<Taggnostr>
that should do the right thing, i.e. read everything between <script> and </script> and emit it as data
15:43
<Philip`_>
That makes sense for <br> and <br/>, since they're both equivalent to XML's "<br/>", but <circle/> is equivalent to XML "<circle>" if outside of <svg> etc or equivalent to XML "<circle/>" if inside of <svg> etc
15:43
<Philip`_>
so whether it should be treated as self-closing is context dependent
15:44
<Taggnostr>
can you elaborate on this a bit?
15:46
<Philip`_>
See e.g. http://livedom.validator.nu/?%3C!DOCTYPE%20html%3E%0D%0A%3Ccircle%2F%3Ex%3Csvg%3Ey%3Ccircle%2F%3Ez%3C%2Fsvg%3E
15:47
<Philip`_>
The "x" is a child of the first circle element, but the "z" is a sibling of the second circle element
15:49
<Philip`_>
(i.e. "<circle/>x<svg>y<circle/>z</svg>" is equivalent to "<circle>x<svg>y<circle></circle>z</svg>")
15:49
<Taggnostr>
I was looking for a dom viewer, this will be useful, thanks!
15:51
<Taggnostr>
so the first circle is parsed following the html rules so the / is discarded, whereas inside <svg> an xml parser is used and the / implies a self-closing tag?
15:53
<jgraham>
Yes
15:53
<jgraham>
Sort of
15:53
<Philip`_>
It's not a real XML parser - it's just a different mode in the tree construction algorithm which handles the token's self-closing flag differently
15:53
<Philip`_>
(in particularly, by treating the token as self-closing, rather than by ignoring the flag)
15:54
<Philip`_>
s/ly//
15:54
<Taggnostr>
ok
15:54
<Taggnostr>
I wonder why </br> is seen as a <br> whereas <hr> is ignored
15:55
<Taggnostr>
is hr gone from html5?
15:55
<Taggnostr>
s/<hr>/</hr>/
15:56
<Philip`_>
Maybe I'm confusing things since I don't know whether you're aiming to output a properly-nested stream of start/end tokens that correspond to an XML serialisation of the DOM produced by the HTML5 parser (which requires unlimited buffering to do it properly), or a usually-incorrectly-nested stream of tokens that match the HTML5 tokeniser's output (which requires an explicit self-closing flag so consumers can decide how to handle the token in a way tha
15:56
<Philip`_>
...that matches HTML5), or some mixture
15:56
<jgraham>
Don't you use irssi, Philip`_ ?
15:56
<Philip`_>
</br> is a special case due to the insanity of HTML
15:56
<Philip`_>
jgraham: Yes
15:57
<Taggnostr>
I just tokenize and emit tokens as soon as I parse them, I don't build trees and/or keep states
15:57
<jgraham>
/load splitlong.pl ?
15:59
<Philip`_>
http://www.whatwg.org/specs/web-apps/current-work/multipage/tree-construction.html#tree-construction says "↪An end tag whose tag name is "br" Parse error. Act as if a start tag token with the tag name "br" had been seen. Ignore the end tag token."
16:00
<Philip`_>
jgraham: That might be convenient
16:05
Philip`_
can't think of enough to say to test whether it actually works, but assumes it will since it didn't give any error messages or anything, so he conditionally thanks jgraham for the suggestion
16:06
Ms2ger
conditionally thanks Philip`_ for updating his canvas tests
16:07
<MikeSmith>
btw, I think Hixie likes his "no-quirks mode" term
16:07
<MikeSmith>
but who knows
16:07
<Philip`_>
Ms2ger: warning: conditional expression is constant
16:07
<MikeSmith>
sometimes you can't predict when Hixie will change his mind
16:08
<MikeSmith>
I like "no-quirks mode" myself
16:08
<Ms2ger>
MikeSmith, it's a lie, though
16:08
<Ms2ger>
"no-quirks mode" has plenty of quirks
16:08
<MikeSmith>
many things are a lie
16:08
<Hixie>
the problem is "standards mode" is a sillier name :-)
16:08
<Hixie>
and "almost-standards" doubly so :-)
16:09
<MikeSmith>
heh
16:09
<Ms2ger>
Because we have no standards?
16:09
<Hixie>
if we spec it all, it's all standards mode
16:09
<MikeSmith>
yeah, please "almost standards"? c'mon
16:10
<Wilto>
Something something dating joke.
16:10
<MikeSmith>
heh
16:10
<MikeSmith>
Wilto: please hang out here more often
16:10
AryehGregor
votes for "quirks mode", "marginally more quirks mode", and "appreciably more quirks mode"
16:10
<Wilto>
I’ll be like the Jar Jar Binks of #whatwg.
16:11
MikeSmith
seconds AryehGregor proposal
16:11
<Wilto>
Quirkiness should be measured on a scale of “zero” to “Zooey Deschanel.”
16:12
<Ms2ger>
AryehGregor, I still hope that "appreciably more quirks mode" will become appreciably less quirky
16:12
<AryehGregor>
Me too.
16:12
Philip`_
was going to suggest "quirks mode", "quirkier mode", "quirkiest mode", but realises that won't be extensible in the future when we discover an even weirder mode that browsers have to be compatible with
16:12
<Ms2ger>
Also, removing document.all entirely
16:12
AryehGregor
was also going to, but realized that the first mode would have to be called "quirky mode" instead of "quirks mode" for consistency
16:13
<Philip`_>
AryehGregor: The naming inconsistency is just another quirk
16:13
<karlcow>
we should pick up the name of modes according to a random selection of insect names
16:21
<MikeSmith>
so I started http://platform.html5.org/history/ recently
16:21
<MikeSmith>
for anybody who cares to help with the forensics
16:22
<MikeSmith>
https://github.com/sideshowbarker/platform.html5.org/tree/master/history
16:23
<MikeSmith>
ara
16:23
<MikeSmith>
Do Not Track support in Opera
16:24
<MikeSmith>
http://my.opera.com/desktopteam/blog/2012/02/10/core-dnt-mail-themes
16:25
<karlcow>
MikeSmith: well I have to write something about this.
16:25
<MikeSmith>
karlcow: merveilleux
16:25
<karlcow>
Because the DNT header on the client side is… peanuts. It's not really the part which matters. And the BIG HUGE battle will be on the server side
16:26
<MikeSmith>
yur caps hurts my eyes
16:26
<karlcow>
I hope I will take the time this afternoon on ODIN blog http://my.opera.com/ODIN/blog/
16:26
<MikeSmith>
"BIG HUGE" is good those in most context
16:27
<karlcow>
MikeSmith: I like to hurt you :p
16:29
<MikeSmith>
on an unrelated note, how do I say "peanut gallery" in your France language?
16:31
<PandaZ>
jquery
16:31
<PandaZ>
oops
16:31
<Ms2ger>
Gallerie d'arachide?
16:41
<MikeSmith>
gallery of spiders?
16:41
<Ms2ger>
No 'n'
16:42
<MikeSmith>
dammit why can't France people just use ENGLISH like the rest of the world
16:42
<MikeSmith>
I mean it's quaint and amusing
16:42
<MikeSmith>
but please
16:42
<MikeSmith>
enough is enough
16:42
<karlcow>
peanut gallery? which meaning for peanut
16:43
<karlcow>
sex or fruits
16:43
<Wilto>
Woah, I tuned back in at a weird time.
16:43
<Ms2ger>
There are no others
16:43
<karlcow>
Wilto: with me, it's always NSFW for americans crowd
16:45
<MikeSmith>
karlcow is a sexiste
16:45
<MikeSmith>
with the e on the end
16:45
<karlcow>
:)
16:46
<MikeSmith>
cf. alcholiste
16:48
<AryehGregor>
ARGH.
16:49
<AryehGregor>
https://www.w3.org/Bugs/Public/show_bug.cgi?id=15508
16:49
<MikeSmith>
AryehGregor: hah
16:50
<MikeSmith>
more of the same on the way
16:50
<Ms2ger>
Ah, CSS
16:50
<AryehGregor>
What do you mean?
16:51
<MikeSmith>
I mean that lots of stuff getting implemented hard and fast these days without much thought about how it should work with existing features
16:51
<MikeSmith>
or other proposed features
16:51
<AryehGregor>
It's not an implementer here, it's someone from Adobe.
16:52
<MikeSmith>
Adobe is an implementor
16:52
<AryehGregor>
I mean, maybe Adobe pays some WebKit people, I don't know.
16:52
<MikeSmith>
um
16:52
<AryehGregor>
For all I know he could work on transforms for WebKit.
16:52
<AryehGregor>
I was assuming not, but maybe that was unjustified.
16:52
<MikeSmith>
you know who Dirk is, right?
16:52
<AryehGregor>
. . . no. :(
16:52
<MikeSmith>
ah
16:52
<MikeSmith>
well
16:52
AryehGregor
shouldn't make assumptions
16:52
AryehGregor
definitely shouldn't make assumptions in a publicly-logged IRC chat
16:52
<MikeSmith>
Dirk's been working on Webkit since before it was Webkit
16:52
<karlcow>
"absolutely" and others… are words that I would absolutely bannish.
16:53
<AryehGregor>
Heh, okay.
16:53
<AryehGregor>
My bad.
16:53
<MikeSmith>
no
16:53
<MikeSmith>
no way to know
16:53
<AryehGregor>
Well, I could have looked before making assumptions . . .
16:53
<MikeSmith>
since we don't record this stuff anywhere
16:53
<MikeSmith>
anyway, he knows a thing or two about a thing or two
16:55
<gsnedders>
AryehGregor: Adobe have quite a few people working on WebKit, on stuff like regions.
16:55
<AryehGregor>
Yeah, I somewhat knew that.
16:55
<AryehGregor>
Should have thought before I spoke.
16:55
<AryehGregor>
I think this would be a bad change, but it's not the end of the world.
16:55
AryehGregor
doesn't think CSS transforms should be brought in line with SVG transforms at all
16:58
<MikeSmith>
maybe should give up on SVG transforms at all
16:58
<MikeSmith>
or whatever they're called
16:59
<AryehGregor>
They should be deprecated presentational attributes.
16:59
<AryehGregor>
Or not deprecated.
16:59
<AryehGregor>
Since SVG is meant for display.
16:59
<AryehGregor>
Or whatever.
16:59
<AryehGregor>
I don't car.e
16:59
<AryehGregor>
care.
16:59
<AryehGregor>
But I don't want us complicating CSS to be more compatible with them.
16:59
<MikeSmith>
Doug and heycam|away and others might not be so keen on that
16:59
<AryehGregor>
Which part?
17:00
<MikeSmith>
dropping SVG transforms
17:00
<AryehGregor>
I'm being hasty and rash here. I'm tired, and I shouldn't be so vocal just because I'm in a group of like-minded people. I still think the change is a bad one, though.
17:00
<MikeSmith>
I don' think SVG actually calls them transforms
17:00
<AryehGregor>
Well, obviously keep them for compat.
17:00
<AryehGregor>
I think it does, no?
17:00
<MikeSmith>
I dunno
17:01
<MikeSmith>
don't know it well enugh
17:01
<MikeSmith>
but anyway
17:02
<MikeSmith>
I think for web author-developers, if they can do something usng CSS, that's what they're going to use
17:02
<MikeSmith>
if they have a CSS way to do it, they don't care if SVG can do it because after all for them them wtf is SVG anyway
17:03
<Wilto>
For what it’s worth, yeah, I’ll attest to that.
17:03
<Wilto>
Not that there’s any excuse for not keeping up with the New Hotnesses™, but… yeah.
17:03
<AryehGregor>
Right, exactly.
17:03
<AryehGregor>
People know about CSS but not SVG.
17:04
<AryehGregor>
I don't think it's worthwhile to make CSS more similar to SVG, but rather the reverse.
17:04
<Wilto>
Right. Not saying I wouldn’t be interested, but the lower the barrier to adoption the better.
17:04
<AryehGregor>
I think my disagreement with Dirk is mostly because he has a strong SVG background, and I'm just about at the point where I can make a red circle without having to look stuff up. :)
17:04
<AryehGregor>
(except the namespace)
17:04
<Wilto>
I put borders on things for a living; I like things to be simple.
17:32
<AryehGregor>
. . . I admit that now I'm getting annoyed at git. Is there really no way to set the default number of context lines for patches?
17:36
<AryehGregor>
Also, git doesn't figure out copies.
17:36
<AryehGregor>
That's one good thing about hg.
17:36
<AryehGregor>
Sigh . . .
17:37
AryehGregor
just hates everything now indiscriminately.
17:38
dglazkov
offers AryehGregor a hug
18:08
<AryehGregor>
I also forgot how it takes at least five minutes to find anything in a git man page, because every command has 200 options, half of them documented by reference to other man pages. Most of which seem gratuitous, insofar as I've never missed them in hg.
18:08
<AryehGregor>
The not-tracking-renames thing is very bad when submitting patches to a non-git project. I think for Mozilla, I'll stick to hg.
18:08
<AryehGregor>
When in Rome . . .
18:09
<AryehGregor>
It's not really nice to submit patches that won't work correctly in hg just because I generated them using git because I don't like hg.
18:54
<jamesr_>
how does 'null' convert to an optional double parameter?
18:55
<jamesr_>
someone's passing 'null' in for the maxWidth parameter of canvas2d's fillText(). trying to figure out if IDL says that it's ignored, or if it converts to 0
18:55
<Ms2ger>
18:56
<Ms2ger>
Er
18:56
<Ms2ger>
Or NaN
18:56
Ms2ger
checks
18:56
<Ms2ger>
jamesr_, +0
18:56
<Ms2ger>
(undefined would be NaN)
18:57
<Ms2ger>
And http://es5.github.com/#x9.3 is the relevant spec
18:58
<jamesr_>
well
18:58
<jamesr_>
the HTML spec says "If maxWidth is present ..."
18:58
<Ms2ger>
It is
18:58
<jamesr_>
what if you say fillText(..., undefined) ?
18:59
<jamesr_>
is it still present, even though the IDL is declared optional?
18:59
<Ms2ger>
It is present
18:59
<jamesr_>
so wait
18:59
<Ms2ger>
Unless
18:59
<jamesr_>
if i had a function with parameters (optional double a, optional double b), how the flip do i call it without 'a' being present?
18:59
<Ms2ger>
[TreatUndefinedAs=Missing] is on the argument
18:59
<Ms2ger>
You don't
18:59
<jamesr_>
void fillText(DOMString text, double x, double y, optional double maxWidth);
19:00
<jamesr_>
ok so i'd have to declare it void foo(optional [TreatUndefinedAs=Missing] double a, optional double b) then do foo(undefined, 1.0) ?
19:00
<jamesr_>
hmmmmm
19:00
<Ms2ger>
You can't omit a if you're passing b
19:00
<Ms2ger>
Impossible
19:01
<jamesr_>
so now i'm not sure what fillText(..., undefined) should do. the first line of the the fillText algorithm reads "If maxWidth is present but less than or equal to zero, return without doing anything; abort these steps."
19:01
<jamesr_>
so if someone passes undefined, that turns into NaN. NaN <= 0 is false, but so is NaN > 0
19:02
<Ms2ger>
Ah
19:02
<Ms2ger>
But there's a catch-all somewhere
19:02
<Ms2ger>
Except where otherwise specified, for the 2D context interface, any method call with a numeric argument whose value is infinite or a NaN value must be ignored.
19:07
<jamesr_>
so what's that mean for null?
19:08
<Ms2ger>
null turns into +0
19:13
<Hixie>
jamesr_: (optional double a, optional double b) is just shorthand for declaring three overloaded operations with (), (double a), and (double a, double b) respectively
19:18
<Ms2ger>
readonly attribute (DOMString or ArrayBuffer)? result
19:18
<Ms2ger>
That works, right?
19:23
<jamesr_>
hm ok
19:48
jwalden
sees days-ago scrollback talking about Mozilla (?) spewing CSS warnings for vendor-prefixed stuff and is quite sure we don't do that (except if it's a -moz- property we don't recognize), seeing as he implemented it
19:50
<jwalden>
well, more day-ago
19:59
<matjas>
hsivonen: can you explain why in old Firefox (3.6) the iframe gets rendered? <style>*{font-family:'</style <iframe onload=alert(1)//';</style>
20:00
<matjas>
i understand why `</style` followed by space acts as end tag for the <style> element
20:02
<Ms2ger>
matjas, treating the < as part of the tag name is an IEism, Gecko would imply a > when seeing a <
20:06
<Wilto>
So this is where you get your dark and arcane secrets, matjas.
20:06
<matjas>
HTML5 parsers seem to discard the iframe in <style>*{font-family:'</style <iframe onload=alert(1)>';</style>
20:07
<matjas>
not sure I understand why though
20:07
<Ms2ger>
Because it's an attribute in an end tag
20:07
<matjas>
i thought </style + space = ETAGO, which would close the element
20:08
<matjas>
but it only does so after the >, i see
20:08
<matjas>
that makes sense
20:08
<matjas>
thanks Ms2ger :)
20:08
<Ms2ger>
ETAGO?
20:16
<matjas>
well, </ = etago
20:16
<matjas>
and </style and </script followed by a space character, >, or / will close their respective opening tag
20:16
<matjas>
i knew that much
22:38
<zewt>
This webpage has disabled automatic filling for this form. <- yeah, uh, webpages shouldn't be able to do that. heh
22:39
<zewt>
This webpage has deliberately inconvenienced you.
23:36
<Hixie>
zewt: aw man i hate that
23:36
<Hixie>
zewt: i made sure to spec autocomplete="off" as optional, but no browser lets me override it as far as i can tell :-(