01:11
<Philip`>
http://canvex.lazyilluminati.com/tests/tests/results.html - now hopefully updated to match the spec again (though not checked very carefully so there may be bugs), plus WebKit data
01:12
<Philip`>
(which was quite painful to collect - I need a much better way of gathering results when the browser doesn't provide getImageData to do it automatically...)
01:14
<zcorpan_>
Philip`: wow, nice work
01:15
<Hixie>
Philip`: that's why you want tests that you can determine the pass/fail result for in 200ms or less
01:15
<Hixie>
Philip`: that way collecting results is just a matter of having a page load every test in a row, with you hitting "y" or "n" for each one
01:16
<Philip`>
Hixie: Determining the results is easy - the problem is I've currently just got little checkboxes for each one, and clicking all those is annoying when a computer should be able to do the job much faster :-)
01:17
<Philip`>
although actually the biggest problem was that after the first 180 or so, Safari stopped responding to any interaction with the form, so I had to select the rest of the checkboxes via javascript: in the address bar
01:19
<Hixie>
heh
01:19
Hixie
thinks it takes more than 200ms to determine if those tests have passed or not, though
01:19
<Hixie>
e.g. http://canvex.lazyilluminati.com/tests/tests/initial.reset.security.html
01:20
<Hixie>
it has loads of text
01:20
<Philip`>
Oh, those ones are easy because it does it automatically and colours the background
01:20
<Philip`>
so the text can just be ignored
01:20
<Hixie>
yeah but you have to know that
01:20
<Hixie>
QA people go through tens of thousands of tests, they can't know them all :-)
01:21
<Philip`>
For the ones where the outcome can be determined automatically, the QA people shouldn't even see them since it's not a good use of their time
01:21
<Philip`>
...but that means I need a better way of reporting the results and asking for human confirmation when necessary
01:41
zcorpan_
would love to have elastomania implemented in <canvas>
01:44
<Philip`>
http://canvex.lazyilluminati.com/tests/tests/minimal.initial.reset.security.html - is that version bare enough now? :-)
01:44
<Hixie>
still has three links :-)
01:45
<Hixie>
see http://www.w3.org/Style/CSS/Test/guidelines.html
01:45
<Hixie>
though ignore the filename crap
01:46
<Philip`>
(http://canvex.lazyilluminati.com/tests/tests/minimal.toDataURL.complexcolours.html when it can't determine the result by itself)
01:46
<Hixie>
those are a bit complex
01:46
<Hixie>
the general colour scheme that test cases use is described somewhere, too, hold on
01:47
<Hixie>
oh it's in that same file http://www.w3.org/Style/CSS/Test/guidelines.html#color
01:47
<Hixie>
you never want to use a green colour when it's not a pass
01:47
<Hixie>
never red when it's not a fail
01:47
<Hixie>
(red trumps green if both are there)
01:47
<Hixie>
you want blue when the tester has to look closer
01:47
<Hixie>
navy, specifically
01:48
<Hixie>
also, the colours should geenerally be #008000 #ff0000 #00ff00 and #000080
01:48
<Hixie>
many qa people in browser dev are basically hardwired to act to those colours now
01:48
<Philip`>
It's sometimes hard to be specific about colours when the test is intentionally trying to test multiple colours, though
01:48
<Hixie>
even something slightly outside that range causes them untold amount of time confused :-)
01:48
<Hixie>
yeah
01:48
<Hixie>
there are tests where that doesn't apply
01:49
<Hixie>
but they're rare
01:49
<Hixie>
see e.g. http://www.hixie.ch/tests/adhoc/html/canvas/
01:49
<Hixie>
anyway gotta go
01:49
<Hixie>
ttyl
02:19
<Philip`>
http://canvex.lazyilluminati.com/tests/tests/minimal.toDataURL.complexcolours.html - now bluer than ever before
02:20
<Philip`>
Still not entirely trivial; but when using a sensible browsers there's only 5 out of 201 cases where you should even have to look at it
02:20
<Philip`>
s/browsers/browser/
03:34
<a-ja>
Hixie, et al: question about 3.8.11.3. Distinguishing site-wide headers from page headers....when referencing one/only/single subsection...should that read/mean *child* subsection instead?
03:36
<a-ja>
i.e. one/only/single child subsection rather than just one/only/single subsection
12:58
<annevk>
Does http://waffle.wootest.net/2007/05/16/ruby-the-same-token/ imply we now have a tokenizer in C?
13:04
<Philip`>
That seems to be only talking about Objective-C, which isn't the same
13:07
<annevk>
Oh, ok
13:14
annevk
updates HTML5 presentations
13:16
<krijnh>
annevk: you're going to talk about HTML5 at Info.nl right?
13:16
<annevk>
heh, how do you know? :)
13:17
<krijnh>
annevk: I told them to ;p
13:17
<krijnh>
Well, no, I used your presentation markup to present my CMS
13:18
<krijnh>
And told Tom about your presentation
13:18
<annevk>
ah ok
13:18
<annevk>
:)
13:18
<krijnh>
He was very interested
13:18
<krijnh>
Also, you charge too little ;)
13:19
<annevk>
Oh, I don't care about money
13:19
<krijnh>
Hehe
13:19
<krijnh>
Just kidding
13:19
<annevk>
I figured as much though :)
13:20
<krijnh>
June 1 around 13h right?
13:20
<annevk>
(Although I guess the reason I don't care about it is that I have enough.)
13:20
<annevk>
Yeah
13:20
<krijnh>
I'll try to be there then as well
13:20
<annevk>
Oh, you work for them now?
13:20
<krijnh>
Freelance, yes
13:21
<krijnh>
3 days a week or something
13:21
<annevk>
So we get to meet after all :p
13:21
<krijnh>
Yay!
13:21
<krijnh>
;]
13:21
<krijnh>
That was my only reason for bringing it up, hehe
13:22
<krijnh>
Nah
13:37
<annevk>
Oh I see, it's a superset of C
13:38
<annevk>
it looks quite different though...
13:45
jdandrea
observes he can't generate conforming HTML 5 using XSLT 1.0. The meta content-type is automagically generated when the head element is encountered while using the html output method!
13:45
<jdandrea>
Workaround: Use XSLT 2.0 (not yet a recommendation) and set include-content-type to no, or hardcode the head element/attributes (also bleah). Ahh well ...
13:45
<jdandrea>
Other ideas?
13:46
<krijnh>
And the meta content-type isn't allowed?
13:46
<Philip`>
XSLT5?
13:46
<jdandrea>
XSLT5! Now there's an idea.
13:46
<jdandrea>
Should use something akin to <meta charset="UTF-8">
13:46
<jdandrea>
But XSLT 1.0 will output <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
13:47
<jdandrea>
XSLT 2.0 lets you suppress that.
13:47
<krijnh>
That's non conforming?
13:47
<annevk>
There has to be an option to suppress that
13:47
<jdandrea>
I believe so. (I could be wrong?)
13:47
<annevk>
Yeah, it's non-conforming
13:47
<jdandrea>
annevk: There is, in XSLT 2.0.
13:47
<annevk>
Hah, XSLT 2...
13:47
<jdandrea>
hehe
13:48
<jdandrea>
Noting it here: http://code.google.com/p/gsa-xhtml-stylesheet/issues/detail?id=30&can=2&q=#c3
13:48
annevk
hopes browsers won't have to implement that
13:48
<jdandrea>
I always thought that was messy (in XSLT 1.0). Now I know why!
13:48
<annevk>
Maybe you should output "XML compatible" HTML5
13:49
annevk
wonders how <!DOCTYPE html> is inserted
13:49
<jdandrea>
Meaning output-method set to xml? Can do.
13:49
<jdandrea>
In XSLT 2.0, yes, good question. In 1.0, you can't have multiple outputs, so I have a template that handles it (hardcoded, again - alas).
13:49
<jdandrea>
Ref: http://code.google.com/p/gsa-xhtml-stylesheet/
13:50
<jdandrea>
The new version is atoning somewhat (HTML 4.01 Strict, (X)HTML 5, etc.) ;)
13:50
<annevk>
Maybe Python is more suitable than all this XML fancyness :)
13:50
<jdandrea>
Maybe.
13:51
<jdandrea>
This project is intended as a drop-in replacement for the default XSLT on the Google Search Appliance.
13:51
<jdandrea>
Otherwise, yes, python has some good possibilities here.
13:52
<jdandrea>
Two more errors to squash and I've got (X)HTML 5 conformance. Yay.
13:52
jdandrea
notes that's before considering new HTML 5 elements.
13:53
annevk
goes t fetch some food and have lunch
13:53
jdandrea
gets ready to head out for the day, waves
13:55
<jdandrea>
annevk: Ahh, XML compatible HTML5 - of course! I just groked that. That would probably work nicely.
14:25
<Philip`>
(http://canvex.lazyilluminati.com/tests/tests/results.html - added data for old Firefoxes - it's slowly getting better)
15:29
zcorpan_
wonders why old firefoxes would be interesting
15:34
<Philip`>
It's useful if you're trying to write content that's compatible with the old browsers people still use, and want to know which features and bugs they have
15:36
<zcorpan_>
ah
15:36
<zcorpan_>
ok
15:36
<zcorpan_>
i thought it was more aimed at implementors
15:37
<Philip`>
It probably is, but if it's easy to get the results then one might as well aim partly at other groups too :-)
16:53
<annevk>
http://me.mywebsight.ws/2007/05/15/xhtml-2-and-html-5-who-will-win/ is funny
16:56
<Dashiva>
"If (X)HTML 5 is about not being too hard for authors who are too lazy to use XHTML 2, then the English language is about not being too hard for authors who are too lazy to use Esperanto."
18:56
<annevk>
Is the Ruby html5lib simply converted Python code?
18:56
<annevk>
It very much looks that way...
18:59
<Philip`>
Yes, hence problems like http://code.google.com/p/html5lib/issues/detail?id=39 where the straightforward Python->Ruby translation doesn't quite work
19:02
<annevk>
Who's twoggle?
19:04
<Philip`>
Tim Fletcher, it seems
19:06
<Philip`>
(as in http://text.rubyforge.org/files/lib/text/figlet_rb.html and suchlike)
19:09
<annevk>
Ah, http://intertwingly.net/blog/2007/05/19/Ruby-HTML5-Parser-Tests-Pass says that too
22:13
<Hixie>
http://forum.mootools.net/viewtopic.php?pid=15695#15695
22:15
<jruderman>
"are we really in 2nd Browser-War? is it really between XHTML 2.0 and HTML 5?"
22:17
<Hixie>
it gets better
22:17
<Hixie>
he talks about reading the minutes... of a 2004 meeting