08:15
<odinho>
Lachy: 21:54 < dfltr> Sort of a random question: http://www.whatwg.org/specs/web-apps/current-work/multipage/selectors.html#selector-enabled states that inputs with type="hidden" should be omitted from :enabled, but currently no browsers support this in their qSA implementations. Is this an in-progress feature or a known issue?
08:17
<Lachy>
odinho, I mailed whatwg about that issue 2 months ago. Hixie hasn't addressed it yet.
08:20
<odinho>
Lachy: Oookay :D
08:44
<jgraham>
zcorpan: Did you say you had a tool for minimising SVG?
08:44
<zcorpan>
http://simon.html5.org/tools/js/svg-optimizer/
08:44
<jgraham>
Thanks
08:48
<zcorpan>
it doesn't perform some obvious optimization like removing useless namespace declarations, metadata, etc, but that's usually simple to do manually
08:49
<jgraham>
Yeah, it would be nice if it removed inkscape-induced cruft
08:53
<heycam>
jgraham, you can try SVG Scour, which I think had removing inkscape cruft in mind
08:53
<heycam>
http://codedread.com/scour/
08:54
<jgraham>
heycam: Thanks
11:02
<jgraham>
I would like to propose the onlick content attribute
11:03
<jgraham>
Because I seem to be unable to type "click"
11:05
<odinho>
jgraham: propose it on the list
11:06
<jgraham>
Alternatvely I want a device with a humidity sensor
11:15
<krijn>
*fingers crossed*
11:17
<krijnh>
\o/
11:22
<zcorpan>
krijnh: hmm?
11:23
<krijn>
Almost up!
11:23
<zcorpan>
oh
11:24
<zcorpan>
nice
11:24
<krijnh>
http://twitter.com/krijnhoetmer/status/236415363259183104
11:26
<zcorpan>
looks like a case from 1995
11:26
<Ms2ger>
Nice wall
11:27
Ms2ger
furiously refreshes
11:27
<Ms2ger>
\o/
11:27
<krijn>
Hey, you're working on some bullshit HTML standard from last century, I'm working on a crappy desktop server!
11:27
<krijn>
Stop ignoring the fact XHTML will take over some day!
11:29
Ms2ger
ignores krijn instead
11:29
<krijnh>
:)
11:29
<krijnh>
Me too!
11:29
<krijn>
Ohnoes!
11:29
<krijnh>
http://krijnhoetmer.nl/irc-logs/ up again?
11:30
<Ms2ger>
It is
11:31
Ms2ger
waves at the peanut gallery
11:31
<krijnh>
Is irc.w3.org still relevant?
11:32
<krijnh>
Is this topic already old?
11:32
<krijnh>
Should I stfu again?
11:32
<odinho>
All that smart stuff I wrote in the down time, noone will ever see it...
11:33
<zcorpan>
krijnh: you mean #html-wg?
11:33
<odinho>
Unless you take donations of logs? :P
11:33
<odinho>
Although I guess i should s/smart/stupid/ and not donate any logs so people see the reality :P
11:33
<krijnh>
zcorpan: yeah
11:34
<krijnh>
odinho: I could fill up the gaps, yes
11:37
<krijn>
Stevef: there, you can power up my PageRank again via Twitter ;)
11:46
<jgraham>
krijn: You realise that you could buy a Raspberry Pi for like 25 USD and it would be better than that thing, right? :p
11:47
<krijn>
No
11:49
<jgraham>
It would be smaller and use less electricity. That's better along some axis. I admit that it wouldn't allow you to use floppy disks.
11:50
<krijn>
Nor my awesome mIRC logger
11:52
<Ms2ger>
Didn't rniwa implement testharness.js support for webkit?
11:53
<jgraham>
I thought he did something at least
12:01
<AryehGregor>
Oh, did he?
12:01
<AryehGregor>
Awesome!
12:03
<krijn>
lol Stevef
12:03
<Stevef>
krijn: i like to be obliging
12:08
<krijn>
Some background, for those caring: we had a little second hand clothes shop, an office and a house (where my mother, brother and sister lived), but they had to throw away all of that (high server costs! ;) and start over again somewhere else. Server now is in a new (much smaller) clothes shop, but office and family is gone there. Result: a much slower internet connection in the new place (less usa
12:08
<krijn>
ge). If anyone thinks it's too slow, I'm open for donations :]
12:20
Ms2ger
likes http://testsuite.org/
12:50
<AryehGregor>
krijn, what are the requirements for a host?
13:05
<AryehGregor>
Did you say this thing actually runs in mIRC?
13:06
<Ms2ger>
He appears to have said that
13:06
<AryehGregor>
:/
13:06
<AryehGregor>
No wonder.
13:08
<svl>
krijn is just very, very good at paving cowpaths
13:08
<jgraham>
Actually logging IRC doesn't seem to be that hard. krijn's main value proposition is the little yellow markers people can use to stoke up flamewars
13:08
<svl>
or at least at placing the occasional cobblestone on them
13:13
<Stevef>
jgraham: http://krijnhoetmer.nl/irc-logs/whatwg/20120817#l-81 thats BS!
13:49
<AryehGregor>
jgraham, we currently detect unreliable tests by hand, file bugs on them by hand, add comments to the bugs by hand every time a known-flaky test fails using a special web interface, and then if a test starts failing far too much we disable it indefinitely by editing our test manifests by hand. And when too many tests fail on a platform, we just completely silence reporting of tests for that platform until someone goes through all the failure
13:49
<AryehGregor>
s and alters all the test manifests to skip the platform. By hand.
13:49
<AryehGregor>
Oh, and our Android test devices like to randomly die during tests, so that on any given test run a few Android runs will usually fail with no explanation.
13:50
<jgraham>
I see
13:51
<jgraham>
When we add a testit is first run 200 times to check that it produces consistent results. If it doesn't it is marked as unstable and the results are ignored for the purposes of regression detection.
13:53
<jgraham>
Then we can also mark failures in stable tests as non reproducable and query to see which tests are producing too many non reproducable fails
15:20
<dglazkov>
good morning, Whatwg!
15:39
<jarek>
Hi
15:40
<jarek>
what's the official stance on presentational attributes in SVG? Are they considered to be good or bad practice now?
15:41
<jarek>
presentational attributes in HTML were deprecated long time ago, will this also happen with SVG for the sake of consistency?
15:41
<zcorpan>
SVG is an image format, it's inherently presentational
15:42
<jarek>
zcorpan: but it shares a lot of CSS rules with HTML
15:42
<zcorpan>
that doesn't make it any less an image format
15:44
<zcorpan>
(gotta go)
15:44
<shepazu>
jarek: they are still supported, and will be for the foreseeable future. There's nothing wrong with presentational attributes, but if you don't like them, don't use them
15:46
<jarek>
I'm not really convinced that having two ways for doing the same thing is good for the standard
15:46
<jarek>
presentational attributes made sense back when XSLT and SMIL were alive
15:47
<shepazu>
jarek: you think that deprecating all content using presetanational attributes is better for the standard?
15:47
<jarek>
shepazu: in the longer run, yes
15:48
<jarek>
shepazu: there are many SVG documents that are mixing both types (e.g. files coded by hand and then tweaked in Inkscape)
15:49
<shepazu>
jarek: this is a pretty philosophical point, and I'm not convinced that it's a problem in practice, nor that the simplistic "style vs. content" argument is relevant to SVG
15:56
<jarek>
is this theoretically possible that HTML5+CSS4will eventually support all the features of SVG?
15:57
<jarek>
gradients, transforms and filters are already there
15:58
<jarek>
the biggest missing feature is <path> element
15:59
<shepazu>
jarek: it's not theoretical… HTML5 already supports all the features of SVG… it's called SVG
16:00
<jarek>
shepazu: all HTML5 does is support for inline SVG element, you still have to work with awkward SVG DOM and use SVG namespaces
16:01
<shepazu>
well, SVG2 will solve some of that
16:01
<jarek>
will it be still XML-based?
16:02
<shepazu>
anyway, HTML5 is feature-frozen, so it would be some future HTML spec, and I don't think it's likely
16:02
<shepazu>
it's markup-based
16:02
<shepazu>
later, I have a meeting now
16:03
<smaug____>
shepazu: luckily this channel is about HTML spec which isn't frozen ;)
16:05
<jgraham>
What is this "frozen" of which you speak?
16:12
<jgraham>
smaug____: Your history tests scare me. I look at them, go "oh, surely everyone's interoperable on that" and then get 3 different behaviours in three browsers
16:17
<Ms2ger>
jgraham, could have got 5
16:31
<tantek>
Lachy, odinho re: that input hidden and :enabled issue - was that logged in bugzilla? can we add this testcase/example to it? Test here: http://jsfiddle.net/duHdp/ (from dfltr)
16:32
<jgraham>
Hmm, so if I understand the spec correctly, the behaviour of history.go(-1); history.go(1) depends on whether the previous page is in the fast back cache
16:33
<jgraham>
Um, not quite
16:33
<jgraham>
But something like that
16:34
<jgraham>
Oh, no that is an example
16:34
<jgraham>
Or, maybe I am wrong
16:35
<Ms2ger>
Maybe?
16:35
<jgraham>
Well I am clearly wrong about some things
16:35
<jgraham>
I think this is one of them
16:35
<jgraham>
For some reason I thought that the navigate algorithm cleared out any other history traversal tasks
16:36
<jgraham>
Hmm, maybe it does
16:36
<jgraham>
Depends what calls "update the session history with the new page"
16:38
<jgraham>
Ah, but this is "entry update"
16:38
<jgraham>
Should probably think somewhere other than IRC
17:47
<Hixie>
sicking: do you have a link to the e-mail where you asked for popstate to not fire during load? I thought the spec used to prevent that, and the Mozilla proposal was to switch to firing it always because it was considered really bad to stop links from working while the page was loading some huge image, or something.
17:48
<sicking>
Hixie: i don't have a link no. Define "during load"
17:48
<sicking>
Hixie: we requested multiple change
17:48
<sicking>
s
17:49
<sicking>
Hixie: one was to remove the behavior of "always fire a popstate event when firing a load event"
17:49
<sicking>
Hixie: another was to "allow popstate to fire while the page is loading, I.e. before the load event has fired, when a history traversal happens"
17:52
<sicking>
Hixie: the blog post explains it pretty well, no?
18:02
<Hixie>
sicking: i thought your e-mail just now was requesting that we not fire it while page was loading?
18:02
<Hixie>
sicking: maybe i misread your e-mail
18:02
<sicking>
Hixie: i might have been unclear
18:03
<Hixie>
"it's unfortunate if the spec still calls for popstate to be fired during pageload"
18:03
<Hixie>
i guess that can eb read both ways
18:03
<Hixie>
anyway as far as i know the spec exactly matches the requests you (mozilla) made
18:12
<sicking>
Hixie: yeah, i think i read it the other way
18:12
<sicking>
Hixie: lemme recheck the thread and clarify
18:14
<sicking>
Hixie: i think he's saying that the spec calls for popstate to automatically fire when the initial "load" event fires for a page load. So no state transitions happening other than loading the page
18:15
<sicking>
Hixie: in fact, i'm quite certain that is what he says given the provided examples
18:16
<sicking>
i'll check if his spec quotes are accurate, but if they are the spec doesn't follow the mozilla proposal
18:16
<Hixie>
i'm not aware of 'popstate' being mentioned anywhere near the 'load' event
18:16
<Hixie>
they're in different chapters
18:17
<Hixie>
the only place in the spec that fires popstate is the "traverse the history" algorithm, and then only if /state changed/ is true
18:18
<Hixie>
that's true when the page is first navigated to, but that's _long_ before 'load' fires
18:19
<Hixie>
(in fact in most UAs I'd expect that to happen before you have any chance of hooking an event handler for it)
18:20
<Hixie>
(it is guaranteed to be before any script execution, though i guess not before the <body onpopstate> is parsed, so it could theoretically be detected)
18:21
<sicking>
Hixie: define "when the page is first navigated to"
18:22
<sicking>
Hixie: the mozilla proposal was that popstate only fires when transitioning sessionhistory entries by the current Document doesn't change
18:22
<Hixie>
well you definitely have to fire it if you go from doc A at state 1 to doc B and back to doc A at state 2
18:22
<Hixie>
i guess we could not fire it when you're going to the document for the first time
18:23
<sicking>
how could you go back to doc A at state 2?
18:23
<Hixie>
(first navigated = when the Document has no session history entry that it has ever been traversed to)
18:23
<sicking>
if you left doc A at state 1 to doc B, and then go back, wouldn't you end back at state 1 in doc A?
18:23
<Hixie>
not if you go forward or back in the session history by more than one step at a time
18:24
<sicking>
ah, yes
18:25
<sicking>
so if you have three entries in SH: <A, 1> <A, 2> <B, 1>. And then go from first to third, no popstate fires. but if you then go from third to second, popstate does need to fire
18:25
<Hixie>
right
18:25
<Hixie>
currently, if you click on a link to <C, 1>, it also fires, as soon as the page is shown
18:25
<Hixie>
we can change that though
18:25
<sicking>
but if you have <A, 1> <A, 2> and the user is at the second state and clicks a link to document B then no popstate should fire at any time
18:26
<Hixie>
(there's no risk of compat issues since you can barely ever detect it anyway)
18:26
<sicking>
yeah, i think we should change that
18:26
<Hixie>
currently the spec treats that as if you are going from <B, null> to <B, 1>, which is why it fires the event, fwiw
18:27
<sicking>
it's unclear what the compat risk is since i don't know how much of the page can be parsed at that time. But gecko already does this, so i'm not too worried about compat risk
18:27
<sicking>
especially for Gecko since we wouldn't change :-)
18:27
<Hixie>
and the other browsers ahven't updated to the new model anyway, right?
18:27
<sicking>
i don't think Webkit does what the spec currently or Gecko does though
18:28
<sicking>
i thought webkit had implemented the gecko model a long time ago, but apparently they have bugs
18:28
<sicking>
this stuff really really needs a test suite
18:28
<sicking>
i'm quite worried that IE gets it totally wrong
18:28
<Hixie>
filed https://www.w3.org/Bugs/Public/show_bug.cgi?id=18605
18:30
<sicking>
thanks
19:00
<nvartolomei>
.
19:37
<Yuhong>
Just posted this: http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2012-August/036933.html
20:52
<jgraham>
sicking: Hmm we have tests for the history state stuff. Did we not release them?
20:52
<jgraham>
Oh we did
20:52
<jgraham>
http://w3c-test.org/html/tests/submission/Opera/historyinterface/
20:53
<sicking>
jgraham: They apparently missed this case though, since I believe Gecko passes them
20:53
<jgraham>
sicking: Not the gecko I have here at least
20:54
<jgraham>
They might have missed this case of course, we don't claim to write perfect tests :)
20:54
<sicking>
hmm.. yeah, you're right. I seem to recall being told we pass, but i'm clearly wrong
20:54
<jgraham>
Patches welcome of course
21:33
<zewt>
(opera history api seemed pretty broken when I was testing recently; clicking <a href=#foo> didn't fire onpopstate)
21:36
<jgraham>
zewt: Write a test and we will fix it
21:41
<zewt>
i really wish there was a web app I could just write, run and submit tests from
21:44
<jgraham>
Yes, several people have suggested that
21:44
<jgraham>
But really, even a live DOM viewer page that demonstrates the problem and can be turned into a real test would be fine
21:45
<zewt>
well, it's pretty straightforward, eg. https://zewt.org/~glenn/test-same-hash.html clicking the links does nothing
21:46
<zewt>
fortunately I can polyfill it pretty easily
21:53
<zewt>
hmm, is it intentional that when postMessage(stuff, [transfer map]) throws DataCloneError, sometimes the objects are neutered and sometimes they're not (depending on when the exception is thrown)?
22:48
<rniwa>
how do I write a nullable method using WebIDL?
22:49
<zewt>
nullable method?
22:49
<rniwa>
yeah
22:49
<rniwa>
that may or may not exist.
22:59
<TabAtkins_>
Don't think you can.
23:00
<zewt>
doesn't really mesh with prototype chains
23:01
<TabAtkins_>
Yeah, you'd have to override the method on the instance to be undefined.