00:53
Philip`
finds half a dozen XSS vulnerabilities in cam.ac.uk pages
00:54
<Philip`>
(I'm surprised that most (i.e. more than 50%, though not a lot more) actually escaped their output correctly)
00:54
<Philip`>
*most forms
01:06
<dbaron>
So, a canvas text question: http://www.whatwg.org/specs/web-apps/current-work/multipage/the-canvas.html#font has wording for em, ex, larger, smaller, bolder, lighter, etc. that is relative to computed values on the canvas element. If the canvas element isn't currently in the document, what computed values should be used?
01:07
<dbaron>
(We have an intern implementing this who wants to know... :-)
01:08
<Philip`>
dbaron: "If the computed values are undefined for a particular case (e.g. because the canvas element is not in a document), then the relative keywords must be interpreted relative to the normal-weight 10px sans-serif default."
01:08
<dbaron>
oh, right
01:08
<dbaron>
ok
01:08
<Philip`>
( http://html5.org/tools/web-apps-tracker?from=1733&to=1734 )
01:08
<Philip`>
(which was a couple of days ago)
01:52
<Philip`>
Ooh, success - a week after I complained, the IE8 Tech Beta newsletter got republished as PDF instead of OOXML
01:55
<Hixie>
oops
01:55
<Hixie>
my mail client stuck a ***DHSPAM*** marker on the last mail i sent
01:55
<Hixie>
i hope dave doesn't take offence
02:46
<dbaron>
Hmmm. So the changes annevk made in the editor's draft of css3-mediaqueries require failing the Acid3 media queries test.
02:46
<dbaron>
I think I preferred it the old way, too.
02:47
<othermaciej>
what are the changes?
02:47
<dbaron>
it's the question of whether "not all and (bogus)" should match
02:47
<dbaron>
the spec changed to say the query should be ignored
02:47
<othermaciej>
I see
02:47
<othermaciej>
but that breaks the logic of not, kind of
02:47
<dbaron>
whereas before unknown features always failed
02:47
<othermaciej>
the new way you describe seems less future-compatible
02:48
<dbaron>
probably, yeah
02:48
<dbaron>
I just won an argument with anne about a similar case for the media type part.
02:48
<dbaron>
But I hadn't noticed it had also changed for features.
02:48
<dbaron>
Or maybe I noticed but didn't realize it was a change.
02:48
<dbaron>
or something
02:50
<dbaron>
of course, i wrote a bunch of other tests now that depend on anne's way
02:50
<dbaron>
hooray for self-contradictory sets of regression tests
04:30
<Hixie>
so uh
04:31
<Hixie>
anyone remarked yet on the ultimate irony of sunava posting a link to an attachment on his internal corp e-mail as his feedback?
04:31
<roc>
you're tempting me to subscribe to public-html
04:31
<Hixie>
this was on public-webapps
04:31
<Hixie>
the new public-webapi / public-appformats list
04:32
<dbaron>
is that a working link or a requires-auth link?
04:33
<Hixie>
requires auth
04:33
<Hixie>
if one was cynical one could see this as yet another stalling tactic (they promised the feedback last november originally), but i honestly think in this case he just made a mistake
04:33
<dbaron>
I'm cynical.
04:34
<Hixie>
aha, he posted a new link
04:34
<Hixie>
http://code.msdn.microsoft.com/xdsecuritywp
04:34
<Hixie>
requires agreeing to a license
04:34
<dbaron>
read it carefully
04:34
<Hixie>
oooh, the license grants us their patents
04:35
<dbaron>
save a copy, then
04:35
<Hixie>
i ain't agreeing to it
04:35
<Hixie>
i'll ask him to forward the paper without the license
04:35
dbaron
heads home
11:47
<Hixie>
MikeSmith: see the bottom of the message i just sent
11:47
<Hixie>
nn
11:48
<MikeSmith>
Hixie: will do
11:48
<MikeSmith>
'night
13:29
<zcorpan>
hmm, shouldn't <video controls> be considered interactive for the purposes of content models?
13:29
<zcorpan>
should user events on the native controls interface be eaten or not?
13:30
<roc>
they shouldn't be visible to content
13:30
<zcorpan>
makes sense
16:18
<zcorpan>
what should happen if you try to do HTMLMediaElement.empty = 42; or HTMLMediaElement.prototype.EMPTY = 42; ?
16:20
<zcorpan>
http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C!DOCTYPE%20html%3E%0A%3Cscript%3E%20Node.ELEMENT_NODE%20%3D%2042%3B%20w(Node.ELEMENT_NODE)%20%3C%2Fscript%3E
16:20
<zcorpan>
opera and firefox say 42, safari says 1
16:24
<annevk>
42 is obviously the correct answer
16:24
<annevk>
oh wait, that's what you set it to yourself
16:24
<annevk>
:)
16:25
<zcorpan>
i'd expected an exception to be thrown
16:25
<zcorpan>
same as readonly attributes
16:25
<zcorpan>
does webidl say?
16:27
<annevk>
i guess it should
16:27
<annevk>
setting .empty diretly shouldn't work though, I think
16:28
<zcorpan>
um, i meant .EMPTY
16:29
<zcorpan>
but why shouldn't it work?
16:29
<annevk>
.empty sounded readonly
16:29
<annevk>
.EMPTY is different :)
16:30
<annevk>
I guess WebIDL should say what happens when setting constants
16:30
<zcorpan>
there's no .empty on HTMLMediaElement
16:30
<annevk>
meh
16:31
<zcorpan>
web idl is public-webapps right
16:32
<annevk>
yeah
16:53
<zcorpan>
does ie support some constant in the dom?
16:56
<annevk>
I don't think so
16:58
<Philip`>
http://www.opera.com/docs/changelogs/linux/950/ - "Canvas.getImageData and Canvas.putImageData methods, including support for creating an ImageData object using the ImageData interface" - too bad it's a totally non-standard way of creating ImageData objects, and the standard way is not supported
17:04
<Philip`>
"xml tags within HTML documents are now given a documentElement attribute." - I don't see that...
17:13
<annevk>
Philip`, yeah, we should probably mention that
17:15
<Philip`>
(Also the path transformation thing is not fixed)
17:19
<Philip`>
(Also, there's a security bug)
17:20
<annevk>
bugs.opera.com/wizard
17:22
<annevk>
and :/
17:49
<Philip`>
(annevk: https://bugs.opera.com/show_bug.cgi?id=337562)
18:19
<Philip`>
(I suppose it's not a particularly severe problem, since it reveals at most four bytes of memory, but it's the principle that matters)
20:15
hsivonen
is now curious about garbage collecting the XHR document
21:03
gsnedders
blames Hixie for making the spec so unclear
21:03
<Hixie>
well that's not one i can easily duck the blame on
21:04
<Hixie>
you know, all these people complaining about how xhr depends on html5 bring up a very good point
21:04
<gsnedders>
That HTML 5 should suck less?
21:04
<gsnedders>
:P
21:04
<gsnedders>
"Let candidate section be the section that contains the previous candidate section in the outline of current outlinee." — if it's the first time we're running that, _what_ previous candidate section?
21:04
<Hixie>
well no, it's just that if the people who want html5 split out for what they want, we'd just have a massive tangle of interdependent specs
21:05
<Hixie>
gsnedders: um
21:05
<gsnedders>
Hixie: Hey, you wrote that! :P
21:05
<Hixie>
i know, i know
21:05
<Hixie>
and this is the simplified version :-/
21:06
gsnedders
wonders whether to be an asshole and just implement something HTML 4 like for TOC construction :P
21:09
<Hixie>
so would <body><h1> be a case where you hit that?
21:09
<gsnedders>
I haven't thought about how you'd hit it yet :P
21:11
<gsnedders>
Hixie: No, you wouldn't hit that
21:11
<Hixie>
<body><h1><h2>?
21:11
<gsnedders>
Hixie: You'd hit "If the current section has no heading, let the element being entered be the heading for the current section."
21:11
<gsnedders>
Hixie: That has an implied </h1>, right?
21:11
<Hixie>
yeah
21:11
<Hixie>
i don't see how you can hit this clause without a previous section
21:12
<gsnedders>
what's the previous candidate section though?
21:13
<Hixie>
the value of "candidate section" when you entered that step
21:13
<gsnedders>
Hixie: Huh?
21:14
<gsnedders>
Hixie: We don't change candidate section until then, so, what? The section before candidate section in the current outline?
21:14
<Hixie>
it's doing candidate section = outline of current outlinee . getSectonThatContains(candidate section)
21:14
<gsnedders>
OK.
21:14
<Hixie>
i'll rephrae it
21:14
<gsnedders>
please do
21:15
gsnedders
has no getSectonThatContains, hmm
21:15
<Hixie>
the only way you can hit this i think is <body><h1/><h2/><h2/>
21:16
<gsnedders>
Which _is_ kinda common :)
21:16
<Hixie>
while processing the second <h2>; and it'll get hit when candidate section is the section for the first <h2/>
21:16
<Hixie>
in which case you can always go up to the <h1>'s section
21:16
<gsnedders>
Yeah, right.
21:16
gsnedders
stops being confused
21:17
<Hixie>
it'll always work because the one case where it would fail is handled as the first "Otherwise" for heading content (before this sublist)
21:17
<gsnedders>
Yeah.
21:17
<Hixie>
clarification deployed
21:17
gsnedders
reloads
21:18
<gsnedders>
Yeah, that's clearer
21:44
<Hixie>
annevk: one way to sidestep a number of the problems w.r.t. the document becoming null (the issue relating to garbage collection for xhr) would simply be to say that the object in question can't be garbage collected while there's a reference to it from xhr.
21:47
<Hixie>
anyone need a dreamhost invitation code btw?
21:48
<krijn>
Not me
21:48
<krijn>
Add it to the issue tracker :)
21:48
<Hixie>
heh
21:48
<krijn>
If you know which, of course
21:49
<Hixie>
gsnedders: i added a bunch of xrefs in the outline section
21:49
<Hixie>
and an example
21:49
<gsnedders>
Hixie: thx
21:51
<gsnedders>
Hixie: How about something crazy like <body><section><h1>foo</h1></section><h1>Place me as a heading of the body, plz?
21:51
<Hixie>
what about it?
21:51
<gsnedders>
Hixie: A crazier example, that proves you can't just build a TOC sequentially (:()
21:53
<Hixie>
you can still do it in one pass
21:53
<Hixie>
you just have to buffer the outline of a section until you get its header
21:54
<Hixie>
it's far better style generally to create a data structure and then serialise it
21:54
<Hixie>
than it is to try and do it all at once
21:54
<Hixie>
especially in these days of ample ram
21:56
<gsnedders>
Hixie: Yeah, that's what I'm doing
22:59
Philip`
thanks the TV Licensing people, since he can't remember his house number or postcode but fortunately it's written on the front of a letter they sent some time last year (presumably demanding that it is highly illegal to not have a TV license (and clearly unthinkable to not have a TV) though I can't be sure since I haven't bothered opening the letter yet)
23:23
<Hixie>
dbaron: i can make it static
23:23
<dbaron>
Hixie, was it either before?
23:23
<Hixie>
dbaron: do you have the same request regarding 'currentColor' in the various places where that can be set?
23:23
<Hixie>
it was underdefined before
23:23
<dbaron>
for canvas colors?
23:23
<Hixie>
i hadn't really thought about it :-)
23:23
<Hixie>
yeah
23:23
<dbaron>
yeah, the same concerns would apply
23:23
<Hixie>
right-o
23:24
<Hixie>
will fix shortly, you can assume both will be static unless someone disagrees with your e-mail and gives a good reason why it should be dynamic
23:25
<dbaron>
FWIW, it's actually one of our interns (Eric Butler) working on this (this being fixing canvas text bugs).
23:26
<Philip`>
Seems a bit weird that ctx.font=ctx.font wouldn't be idempotent
23:26
<Philip`>
Uh
23:26
<Hixie>
dbaron: cool
23:26
<Philip`>
I don't mean idempotent
23:26
<Philip`>
I just mean it'd be a bit weird that it could have an effect
23:26
<dbaron>
we could make it live
23:27
<dbaron>
Though in some cases it could be a good bit of work for each text-drawing operation
23:27
<Hixie>
live seems complicated and bug prone to me, but i agree that setting something to itself having an effect is something i've tried to avoid before
23:27
<dbaron>
particurlarly if the canvas is deeply nested inside something that's display:none
23:27
<Philip`>
(but canvas.width=canvas.width already has a significant effect (clearing the canvas), so it wouldn't be any worse than that)
23:27
<Hixie>
yeah
23:28
<dbaron>
who else implements this part of the canvas text stuff?
23:28
<Philip`>
dbaron: Nobody
23:29
<Hixie>
it's very new
23:29
<Hixie>
i'm not aware of any implementations
23:38
<Hixie>
Lachy: you probably want to change the abstract for your draft
23:39
<Hixie>
er, the intro, i mean
23:39
<Hixie>
actually nm.
23:39
<Hixie>
i'm on crack,.
23:40
<Hixie>
re your namespaces section, the xml: prefix example is misleading since you never have to declare that prefix anyway