00:30
<Hixie>
hsivonen: you sent four mails right?
00:30
<Hixie>
about the parser?
00:38
<webben>
hsivonen: re http://lists.w3.org/Archives/Public/public-html/2007Jul/0064.html and http://lists.w3.org/Archives/Public/public-html/2007Jul/0067.html, a problem with FIrefox 1.04 (of all things) is not really a problem with AT, is it? I'm also not sure why we're evaluating old methods of embedding Flash rather than the newest techniques: http://alistapart.com/articles/flashembedcagematch/ . (These methods are continuing to evolve.)
00:39
<webben>
(Firefox 1.04 didn't have effective screen reader support.)
00:48
<Philip`>
I was beginning to think that browsers weren't that bad at drawing lines, but then I reached arcTo and my hopes were shattered :-(
07:54
<Hixie>
sweet
07:54
<Hixie>
it works
07:55
<Hixie>
the "table model" algorithms for forming a table totally work and generate actual tables that are what the markup was!
07:55
<Hixie>
i guess i shouldn't be so surprised but that's pretty damn cool.
07:56
<Hixie>
tomorrow i can do the scope="" and headers="" stuff
07:56
<Hixie>
now that i have actual tables
08:08
<annevk>
what language did you wrote your model in?
08:13
<hsivonen>
Hixie: yes, I sent 4 emails.
08:49
<annevk>
"were many things which WHATwg want to delete, are mentioned (ACRONYM, non-visual rendering of TABLES amongst them.)"
08:50
<annevk>
FO 4
08:50
<annevk>
these people are crazy
10:06
<zcorpan_>
what's the reason that i like testing parsing more than testing other things? :)
10:07
<annevk>
parsing is trivial? :p
10:07
<zcorpan_>
perhaps :)
10:07
<zcorpan_>
but not really
10:07
<annevk>
btw, any chance you'll make some script that converts html5lib testdata to browser usable tests?
10:08
<annevk>
because making html5lib tests is trivial
10:08
<zcorpan_>
hmm, i could look into that
10:08
<annevk>
someone from mozilla already did some work
10:08
<zcorpan_>
pointer?
10:08
<annevk>
I'm afraid I don't have a pointer
10:09
<annevk>
the only thing you need to do is get the result DOM
10:09
<annevk>
iterate over it
10:09
<zcorpan_>
load the tests in an iframe
10:09
<annevk>
yeah
10:09
<annevk>
and from the restult DOM you make a string and compare that with the #document data
10:09
<zcorpan_>
ah yep
10:10
<zcorpan_>
seems pretty straightforward
10:11
<zcorpan_>
do the existing tests assume that scripting is enabled? or disabled?
10:11
<annevk>
I think enabled, but scripting support is not there
10:12
<zcorpan_>
might perhaps be good to split up the tests that assume that scripting is enabled or disabled (or neither)
10:13
<annevk>
I suppose we could make extensions
10:13
<annevk>
#data --scripting-disabled
10:13
<zcorpan_>
perhaps
10:15
<annevk>
don't know what's best
10:16
<zcorpan_>
not sure if it's possible to enable/disable scripting on a per test basis
10:17
<zcorpan_>
my port would probably only work with tests that don't assume that scripting is disabled
10:18
<annevk>
sure
10:18
<annevk>
you could have some server-side script that organizes the tests for you or something
10:18
<annevk>
in that case you can simply make a scripting-enabled and scripting-disabled folder
10:18
<annevk>
dunno
10:21
<annevk>
to start it would be nice to have some JS that does DOM -> #document\n...
10:21
<zcorpan_>
yeah
10:21
<annevk>
then we can compare browsers to html5lib more easily
10:22
zcorpan_
looks at the live dom viewer code
10:23
<zcorpan_>
hmm, what if the dom isn't a tree
10:24
<annevk>
Hixie has solved that issue
10:24
<annevk>
I suppose you just keep a pointer around and check if you haven't encountered the node before
10:24
<zcorpan_>
yeah, but the html5lib test output format doesn't support non-trees
10:25
<annevk>
oh
10:25
<annevk>
elementname - graph
10:25
<annevk>
just output something that makes it non-conforming
10:38
<annevk>
(if we have that script it can be integrated with the thing david has)
10:39
<jgraham>
http://lxr.mozilla.org/seamonkey/source/parser/htmlparser/tests/mochitest/
10:39
<annevk>
http://hasather.net/html5/parsetree/
10:39
<annevk>
jgraham, cool
10:48
<hsivonen>
I wonder if sayrer is writing a new HTML parser or changing the old one
10:49
<hsivonen>
(I wouldn't want to change the old one)
10:49
<annevk>
dunno, seems like mrbkap is still fixing the old one
10:49
<annevk>
(from the couple of bug numbers referenced there)
10:50
<annevk>
they also checked out all the old tests
10:55
<annevk>
jgraham, btw, how easy is it to extend all treebuilders to support two more variables for DOCTYPE?
11:36
<zcorpan_>
http://simon.html5.org/temp/dom-tostring.html
11:44
<zcorpan_>
please tell me if i can improve the function, i think it is doing some unnecessary things (looping too many times or something)
11:44
<zcorpan_>
or if i got the format wrong
11:50
<hsivonen>
zcorpan_: well, if you wanted to optimize performance (which isn't really necessary), you could traverse the DOM using an iterative algorithm instead of a recursive one
11:56
<zcorpan_>
hsivonen: hmm, ok. i think i'll leave it as is for now
12:58
<zcorpan_>
the selectors api naming issue will never end
13:02
<annevk>
I wonder when HTML 5 will get issues like that
13:02
<annevk>
hopefully when everything is already shipped in implementations
13:08
<zcorpan_>
selectorQuery/selectorQueryAll seem to be good names
13:10
<annevk>
wfm
13:25
<Lachy>
yeah, I like those names too. It looks like a reasonable compromise between selectElement and cssQuery, but I have a feeling there's going to be someone that doesn't like it
13:27
<zcorpan_>
of course someone won't like it
15:12
<zcorpan>
so why doesn't this work? http://simon.html5.org/temp/html5lib-tests/wrapper.html
15:14
<annevk>
you're not invoking send()?
15:14
<annevk>
with my XHR2 that would not be necessary but it seems that it will not going to fly :)
15:17
<Lachy>
how is XHR2 going to know when to implicitly invoke send()?
15:17
<annevk>
much like Audio(src)
15:17
<annevk>
anyway, not going to happen
15:18
<annevk>
new XMLHttpRequest(method, src) is likely the only shortcut that makes some sense
15:18
<Philip`>
Sounds like about the same as setTimeout(fn, 0) too (i.e. as soon as possible after the current script has finished)
15:18
<annevk>
yeah
15:23
<zcorpan__>
hmm, is tests1.txt parsed as xml or something?
15:25
<annevk>
shouldn't matter for responseText
15:26
<zcorpan__>
Error: syntax error
15:26
<zcorpan__>
Source File: file://loca.....l5lib-tests/tests1.txt
15:26
<zcorpan__>
Line: 1, Column: 1
15:26
<zcorpan__>
Source Code:
15:26
<zcorpan__>
#data
15:27
<annevk>
oh, in Firefox?
15:27
<zcorpan__>
yeah
15:28
<annevk>
they have some issues there I believe
17:17
<virtuelv>
hm, wonder which version of Safari the iPhone is based on
17:18
<virtuelv>
Row 1 and 4+5 of Acid2 fails
17:19
<Philip`>
virtuelv: It calls itself "Mozilla/5.0 (iPhone; U; CPU like Mac OS X; en) AppleWebKit/420+ (KHTML, like Gecko) Version/3.0 Mobile/1A543a Safari/419.3" which sounds like a possibly old WebKit
17:21
<virtuelv>
then this too: http://www.johnmurch.com/2007/07/01/iphone-javascript-and-spec-benchmark/ makes sense
17:22
<Philip`>
Looks like someone tried running Canvex on an iPhone - I wonder what kind of performance they get...
19:35
<jgraham>
annevk: The only issue I can think of with adding more doctype stuff to the treebuilders is several tree types don't really support the concept of doctypes
19:36
<jgraham>
There have been hacked in various ways to support the concept e.g. by having elements with tagnames like "<!doctype>"
19:37
<jgraham>
But where we hacked we can hack a little more
19:37
<jgraham>
:)
19:45
<hsivonen>
jgraham: my plan is not to expose the doctype by default through APIs designed for XML
19:46
<jgraham>
hsivonen: We need it for running tests. By default it is stripped when a special flag is not set
19:46
<hsivonen>
annevk: the doctype node stuff is for browsers. it is rather pointless for parsing libraries meant for non-browser use cases
19:47
<hsivonen>
jgraham: OK. I guess I have to support doctypes in SAX (buffered and streaming)
19:47
<hsivonen>
but supporting them in the DOM just sucks without a backdoor API
19:48
<jgraham>
(that is, it is stripped where it not supported by the underlying tree)
19:49
<hsivonen>
if the use case is making a drop-in library for apps that now use an XML parser, exposing the doctype might be more harm than good
19:50
<hsivonen>
but yeah, supporting doctype exposure in SAX and XOM is reasonable
23:53
Hixie
narowly misses flinging his ipod acress his cube
23:53
<Hixie>
oops
23:57
<Hixie>
right, done scope=""
23:57
<Hixie>
now headers=""