08:07
<othermaciej>
hi folks
08:13
<annevk>
VPN works
08:13
<annevk>
after a while
08:13
<annevk>
not bad
13:27
<foolip>
is anyone in here at TPAC? Hixie?
13:28
Ms2ger
suspects at least kennyluck, othermaciej and homata_
13:30
<othermaciej>
I am
13:31
<Moo^_^>
TPAC?
13:33
<karlcow>
W3C Technical Plenary - Advisory Committee
13:33
<karlcow>
Moo^_^: http://www.w3.org/2010/11/TPAC/
13:33
<karlcow>
also http://search.twitter.com/search?q=%23tpac
13:42
<Moo^_^>
karlcow: thx
13:56
<annevk>
foolip, yeah
13:56
<annevk>
also those Ms2ger mentioned
13:56
<annevk>
Ms2ger, we can put ElementTraversal into Web DOM Core
14:03
hendry
is at TPAC
14:04
<annevk>
oh cool
14:12
<hsivonen>
Ms2ger: while you were away, I took the liberty to land one of your patches
14:13
<hsivonen>
Ms2ger: http://hg.mozilla.org/mozilla-central/rev/6aad432efa02
14:13
<zcorpan>
annevk: why bother touching ElementTraversal?
14:14
<annevk>
it's like HTML5 takes in from HTML4 and DOM 2 HTML and some DOM 0 stuff
14:15
<annevk>
less specs for DOM Core is a good thing imo
14:16
<zcorpan>
html4 and dom2 were underspecified and broken
14:16
<zcorpan>
elementtraversal is fine afaik
15:04
<annevk>
zcorpan, sure
15:05
<othermaciej>
hi all
15:11
<necttocon>
Sup guys
15:11
<necttocon>
I was told to go here for questions etc
15:11
<necttocon>
so um
15:11
<necttocon>
maybe you guys have an idea why my rectangle is rotating around a circle, instead of rotating around the middle of the rectangle?
15:12
<foolip>
Does anyone know who the <video> developer(s) for Chrome are?
15:12
<annevk>
necttocon, URL or it didn't happen?
15:12
<Philip`>
necttocon: When you call rotate(), it rotates around the (0,0) point
15:13
<necttocon>
Philip` oh
15:13
<Philip`>
If you want to rotate around some arbitrary point (x,y), you have to do translate(-x, -y); rotate(a); translate(x, y);
15:13
<necttocon>
Wow
15:13
<necttocon>
Thanks Philip`
15:13
<necttocon>
Will try it out in a sec
15:13
<Philip`>
(unless the translates are the other way round)
15:14
<Philip`>
(I can never quite remember)
15:14
<zcorpan>
i think they're the other way around
15:15
<Philip`>
Trial-and-error is my usual approach to working out what the signs should be
15:15
<Philip`>
which works okay up to about four independent boolean variables
15:15
<necttocon>
Philip`: They should add it to the docs, or did I miss it?
15:15
<Philip`>
necttocon: Which docs?
15:16
<necttocon>
Philip`: specification
15:16
<necttocon>
Sorry, docs != specification
15:17
<Philip`>
I think it's not really the spec's job to teach readers about geometry and transformations and other standard computer-graphics things
15:17
<Philip`>
since that's a pretty large job and isn't specific to canvas
15:18
<Philip`>
so it's probably better to refer to some other document/book to understand the fundamental concepts like this
15:19
<necttocon>
Philip`: I've never worked with rotations/transformations so it's new to me that it always rotates around the (0,0) point
15:22
<Philip`>
http://en.wikipedia.org/wiki/Transformation_matrix#Rotation is what the spec means by "rotation transformation"
15:22
<Philip`>
but the matrix stuff isn't entirely intuitive if you've never seen it before :-)
15:23
<necttocon>
I never read formulas on wikipedia
15:23
<necttocon>
they scare me
15:24
<Philip`>
Any introduction to computer graphics should cover this, though I don't know any specific examples to recommend
15:25
<zcorpan>
"introduction to computer graphics" on google gives plenty of results :)
20:05
<Hixie>
abarth: dude, the http guy just used googlebot against you
20:05
<abarth>
omg
20:05
<Hixie>
oh you already replied to that one
20:05
<Hixie>
i'm surprised you didn't jump on it :-)
20:05
<AryehGregor>
Is there any simple standard JavaScript way to just say "load this script/stylesheet", without having to actually create a script or link element and put it somewhere in the DOM?
20:06
<AryehGregor>
There should be. It's a very common need.
20:06
<abarth>
whenever i interact with that working group, i'm tempted to start my own standards body
20:07
<Philip`>
AryehGregor: XHR + eval?
20:07
<AryehGregor>
:/
20:07
<AryehGregor>
I was thinking something more like document.loadScript("http://...");
20:08
<Hixie>
abarth: well, i replied to that e-mail too
20:14
<Ms2ger>
AryehGregor, write a spec and get it implemented? :)
20:14
<AryehGregor>
Ms2ger, maybe I could find an existing spec and try to get someone else to do it for me. :)
20:21
<AryehGregor>
I guess you'd want that to somehow translate into adding something to the DOM in a predictable place, since otherwise you couldn't remove it.
20:22
<Philip`>
What does it mean to remove a script?
20:22
<AryehGregor>
Well, say for styles.
20:22
<Philip`>
Ah, that seems like an independent thing
20:22
<AryehGregor>
Although, what happens if you do remove a <script>?
20:22
<AryehGregor>
Yes, in practice they're going to be quite different.
20:22
<Philip`>
Nothing, I think
20:22
<Hixie>
nothing interesting
20:23
<Hixie>
<script> is only interesting when you insert it into a doc
20:23
<Hixie>
after that it becomes boring
20:23
<abarth>
Hixie: using a dummy document for fragment parsing is causing performance problems. WebKit innerHTML is not 10x slower for trivial values. :(
20:24
<Hixie>
yeah i never expected anyone to actually implement it that way
20:25
<abarth>
i like the model. we'll figure out how to make it fast
20:25
<Hixie>
so long as you don't run scripts and fire no mutation events except the last one(s) at the end, as if it'd all been inserted at once, i think the effect is the same
20:25
<Hixie>
there might be some other subtlties
20:25
<smaug____>
abarth: did you accidentally fix webkit to not have the "optimization" for simple innerHTML values?
20:25
<smaug____>
(the optimization which is against the spec)
20:26
<abarth>
smaug____: possibly. there aren't any optimizations for innerHTML
20:26
<Hixie>
webkit used to do something dodgy with reusing a text node or something
20:26
<smaug____>
yeah, it at least used to reuse the text node in some cases
20:26
<abarth>
boo
20:26
<abarth>
that doesn't seem anywhere close to correct
20:26
<Hixie>
indeed
20:26
<Hixie>
fast though!
20:27
<smaug____>
fast, but no one else does that, IIRC
20:27
<smaug____>
since it doesn't make sense
20:27
<Hixie>
yeah
20:27
<smaug____>
IMO
20:27
<Hixie>
agreed
20:27
<Hixie>
that's why the spec doesn't allow it :-)
20:27
<abarth>
that case is tested in peacemaker
20:28
<abarth>
which is about the lamest benchmark suite i've ever seen
20:28
<abarth>
https://bugs.webkit.org/show_bug.cgi?id=48719
20:29
<jamesr_>
peacekeeper is the worst browser benchmark out there
20:29
<Hixie>
you can still mostly do the optimisation if you just make sure that you don't do it if there are any references from JS to the text node
20:29
<jamesr_>
which is quite a feat
20:29
<Hixie>
(which there will almost never be)
20:30
<abarth>
that's true
20:30
<abarth>
a first step is to recognize the case and use a new text node
20:30
<abarth>
rather than constructing a whole new document
20:34
<AryehGregor>
So the standards-compliant way to add a stylesheet would be something like: var link = document.createElement("link"); link.rel = "stylesheet"; link.src = foo; document.appendChild(link);
20:36
<Ms2ger>
s/document.appendChild/document.head.appendChild/
20:36
<AryehGregor>
What if document.head is null?
20:36
<Hixie>
then your document is not conforming html
20:37
<AryehGregor>
Yeah, I guess it would be pretty hard for that to happen. How could it be null by the time you get to executing script, in text/html?
20:37
<AryehGregor>
Isn't something going to auto-create the <head> at some point?
20:38
<Hixie>
<html onclick="something..." style="...."><!-- network pauses here, user clicks on document -->
20:38
<AryehGregor>
Aha, so it could fail in some cases where your document is conforming.
20:38
<Hixie>
true
20:39
<AryehGregor>
So using document is safer. Although technically HTML5 claims that it's non-conforming if you put the link element outside the head/body, the validator won't notice so no one will care.
20:39
<AryehGregor>
(besides, that seems like a bogus requirement anyway when it comes to script-inserted stuff)
20:39
<Hixie>
you won't be able to insert it outside the root element
20:39
<Hixie>
you'll get a HIERARCHY_ERR
20:39
<Hixie>
if you're going to all this effort, just create a <head> element :-)
20:40
<Hixie>
or do it in such a way that you can't run script before the <head> is seen
20:40
<Hixie>
i mean, a script could always just remove the <head> manually, to not have a <head>
20:40
<Hixie>
it's not clear to me what your use case is though
20:42
<romeo_>
I am looking at tokenization states. Are the cases supposed to be ordered in some way?
20:44
<romeo_>
E.g. the tag open state is not ordered numerically.
20:44
<romeo_>
It would be faster if small ascii letters was moved up before capital asciis.
20:45
<romeo_>
s/was/were/
20:45
<Hixie>
it's theoretically ordered by unicode code point except that things that cause parse errors come last
20:46
<Hixie>
however i've not always been good about keeping that order consistent
20:46
<AryehGregor>
Hixie, it's very common to want to load stylesheets from script. Currently it's pretty awkward to do.
20:46
<Hixie>
AryehGregor: agreed
20:48
<AryehGregor>
So do you have any ideas for an API to make it nicer?
20:48
<Hixie>
document.addStyleSheet(url, [title], [alternate-p]); ?
20:48
<Hixie>
or document.addStyleSheet(url, [media], [title], [alternate-p]); ?
20:49
<Hixie>
anne might have something cooking along these lines as part of the CSSOM AltSS stuff
20:49
<romeo_>
Okay, thanks.
20:53
<abarth>
AryehGregor: document.head can also be null if you remove the head from the DOM
21:26
<Hixie>
how do i get a url from a File object these days?
21:39
<Hixie>
ok, dropzone is done
21:43
<Hixie>
abarth: his response makes no sense... "ok, so class A and class B of UAs might need to interoperate... but there are other classes (which i am not going to name) that don't need to interoperate"
21:43
<Hixie>
say what?
21:43
<Hixie>
you gotta wonder why these people are writing specs, some times
21:44
<abarth>
they seem to have some aesthetic thats important to them
21:44
<abarth>
which i haven't quite figured out yet
21:44
<hober>
what list is this on? httpbis?
21:44
<Hixie>
yeah
21:45
gsnedders
stopped reading that a while ago
21:45
<abarth>
his mail really makes no sense
21:46
<Hixie>
gsnedders: me too, but every now and then someone says something that hits one of my alarms, and i get to see it again :-)
21:46
<abarth>
he's arguing that the requested file name isn't part of the semantics of Content-Disposition: attachment; filename=foo
21:47
<Hixie>
i added HTTP to the list of specs we need to write, linking to his e-mail to show that the wg asked us to write it and we're not just doing it because of NIH syndrome
22:29
<JonathanNeal>
What do you guys think of http://forums.whatwg.org/viewtopic.php?p=5912&sid=995810a1b65454d366aee56fdc8f504a ?
22:30
<Hixie>
interesting idea
22:33
<JonathanNeal>
More modern websites would validate. I mean, sticking <script>s after <body> seems pretty normal practice to me.
22:36
<Hixie>
sticking <script> before or after </body> today has no effect (the browser moves them all before the </body> anyway), so there's no reason to do that today
22:37
<JonathanNeal>
I'm pretty sure it has an effect. Let me see...
22:37
<Hixie>
the parser essentially ignores </body>
22:37
<Hixie>
if it has an effect, that is significant news
22:39
<JonathanNeal>
I always thought the document was not accessible until it had </body>'d.
22:39
<Hixie>
good lord no
22:39
<Hixie>
the document can be accessible as soon as the browser wants it to be
22:43
<hober>
IE9 will support application/xhtml+xml
22:43
<JonathanNeal>
Thanks Hixie.
22:44
<AryehGregor>
JonathanNeal, that's disproven by just loading any particularly long page. You'll see that the top loads before the full page content has loaded.
22:44
<JonathanNeal>
Maybe it's just in older IEs
22:44
<JonathanNeal>
And I think too legacy or something stupid.
22:44
<Hixie>
IE does have weird behaviour in this space
22:44
<Hixie>
but even in IE you can disprove it if you run a script with an alert() half way through the page
22:44
<Hixie>
amongst other means
22:46
<JonathanNeal>
Sure, I guess I just couldn't access the document.body until it was done or something.
22:48
<AryehGregor>
You can access that as long as the opening tag has been parsed, it just won't contain its full contents.
22:49
<AryehGregor>
If you only want to run after the full document has been parsed, use onreadystatechange or whatever that new feature is called.
23:01
<Rik`>
DOMContentLoaded ?
23:02
<AryehGregor>
Sounds right.
23:34
<cardona507>
anyone know what % of html5 tests were submitted by microsoft?
23:36
<AryehGregor>
You can look yourself to get an idea: http://dvcs.w3.org/hg/html/file/662a167f802d/tests
23:36
<AryehGregor>
approved/ and submission/ are the relevant ones.
23:36
<AryehGregor>
Oh, approved/ doesn't say who submitted them.
23:37
<AryehGregor>
Well, submission/ does.
23:37
<cardona507>
thnx AryehGregor