00:08
<nessy>
cool :-)
00:08
<nessy>
now I can remove all the webvtt stuff from the html spec, too
05:27
<Hixie>
is there a better way of checking that of three booleans, only one is true, than (not ((a and b) or (a and c) or (b and c)) ?
05:45
<SimonSapin>
Hixie: your expression is also true if they’re all false, is that what you want?
05:46
<jtcranmer>
Hixie: (a xor b) and (b xor c) and (a xor c)?
05:47
<jtcranmer>
hmm
05:47
<jtcranmer>
(a xor b xor c) and not (a and b and c) ?
05:47
<heycam>
Hixie, if it's JS and you know your variables are booleans, you can do `a + b + c == 1`
05:50
<Hixie>
it's perl and they're really integers
05:51
<Hixie>
SimonSapin: they can't be all false in this case, but good point
05:51
<Hixie>
i guess i could do a !! trick
05:51
<Hixie>
!!a + !!b + !!c == 1
08:59
<annevk5>
marcosc: that success/error-callback thing in Midi looks like a future
09:02
<jgraham>
annevk: You have turned into a bot
09:02
<jgraham>
:)
09:02
<annevk>
jgraham: I don't understand the question, "You have turned into a bot"
09:02
<jgraham>
Just need to add another bot that responds to "X looks like a future" with "and futures look like monads"
09:05
<Ms2ger>
And monads look like my head just exploded, again
09:09
<jgraham>
(fwiw http://stackoverflow.com/a/2704795 seems quite nice)
09:10
<Ms2ger>
Maybe I should read it...
09:10
<Ms2ger>
Once I figured out existential types
09:10
<jgraham>
(if one is looking for a practical introduction to Monads, albeit in C#)
09:10
<jgraham>
(Not that I really claim to understand and thing)
09:11
<jgraham>
*anything
09:11
<Ms2ger>
And that's a doctor...
09:11
<annevk>
just like Future is Future<any, any> I guess
09:12
<annevk>
and you can make it Future<MediaStream, DOMError>
09:12
<jgraham>
(but what little I do underestand makes it clear why futures could/should/do form a monad)
09:12
<annevk>
and all the future composition stuff still applies, etc.
09:15
<asmodai>
Do any of you know of a bug with Chrome and refreshing of pages that at some point causes it to lose its styling of said page?
09:28
<jgraham>
darobin: "Any reason you *should* grab it and deploy"? Well maybe if you are insane and like running untested code in production?
09:30
<darobin>
jgraham: there's another way of doing development?
09:30
<jgraham>
So I hear :)
09:30
<darobin>
jgraham: to clarify, my question was because I couldn't tell from your comments how prod-ready you thought it was
09:30
<darobin>
you said untested, but at the same time you seemed to say it was pretty close to being ok
09:30
<Ms2ger>
The usual Knuth quote?
09:31
<Ms2ger>
"Beware of bugs in the above code; I have only proved it correct, not tried it."
09:31
<darobin>
"Who are you and what are you doing in my bedroom?" // that one?
09:31
<jgraham>
It's untested. So it probably has all the usual types of errors that untested python code has e.g. arguments not matching, etc.
09:31
<darobin>
hehe
09:31
<jgraham>
But I don't know of any missing parts to the code
09:31
<darobin>
for reference http://xkcd.com/163/
09:32
<darobin>
ok, so maybe a bit of testing before deployment would be useful
09:32
<jgraham>
I recommend setting up a test repo somewhere, mirroring that somewhere (doesn't even have to be public facing) and seeing if the commands work as expected
09:32
<jgraham>
and it responds in the right way to new PRs and so on
09:33
<jgraham>
I should have done this, but I did something else instead :)
09:33
<jgraham>
No idea what
09:33
<Ms2ger>
Drink?
11:44
<zcorpan>
Ms2ger: here
11:45
<jgraham>
That's some slow-acting magic right there
11:45
<odinho>
glacier
12:14
<Ms2ger>
zcorpan, still here?
12:14
<zcorpan>
Ms2ger: yep
12:14
<Ms2ger>
http://w3c-test.org/web-platform-tests/master/html/webappapis/scripting/processing-model-2/body-onerror-compile-error-data-url.html
12:15
<Ms2ger>
Since you submitted that test, want to explain it? :)
12:18
<zcorpan>
Ms2ger: explain why the arguments should be tainted? or something else?
12:18
<Ms2ger>
I guess that's what it comes down to, yes
12:22
<zcorpan>
Ms2ger: it seems the spec has changed since the test was written
12:22
<Ms2ger>
\o/
12:23
<zcorpan>
Ms2ger: when it was written, the "report the error" algorithm compared the origin of the URL of the script with the origin of the script's document to decide whether to mute
12:24
<Ms2ger>
I see
12:24
<zcorpan>
Ms2ger: now the spec supports CORS and the muted-ness of data: has changed
12:25
<Ms2ger>
You don't happen to feel like updating the test? :)
12:26
<zcorpan>
first i want to ponder if the spec change can cause new security problems
12:27
<zcorpan>
the only thing i can think of is something that redirects to a data: URL, but that should be considered cross-origin (known bug)
12:29
<zcorpan>
there should be some CORS test here too
12:31
<zcorpan>
is the process the same for minor bugfixes like this as for any other test submission?
12:32
<Ms2ger>
I think so, yes
12:32
<zcorpan>
ok
12:33
<darobin>
the process is whatever we want to make it, keep that in mind
12:33
jgraham
wants the process to be "darobin reviews all tests"
12:33
<darobin>
as a rule, if it's substantive even if small I think it's better if there's a bit of review
12:33
darobin
thinks jgraham missed the "we" bit :)
12:34
<darobin>
but if it's just a quick and obvious fix I would say you can just push
12:34
<darobin>
or, put differently, the process is "be fucking sensible" :)
12:34
<jgraham>
My view is that getting review for small changes is no big deal because small changes can be reviewed synchronously anyway
12:34
<zcorpan>
my biggest issue is naming the branch :-P
12:34
<zcorpan>
submission/Opera/onerror-data-url-muted-bugfix
12:35
<zcorpan>
what if i want to do another bugfix?
12:35
<jgraham>
But it is OK to just push whitespace changes and typo fixes and similar
12:35
<Ms2ger>
But what if the typo was essential to the test?
12:36
<jgraham>
Then you shouldn't have named a variable "outlinee"
12:36
<jgraham>
But seriously, it's OK to have failure modes
12:36
<jgraham>
I mean review is a long way from perfect
12:37
<Ms2ger>
:D
12:37
<jgraham>
But the goal is a) to reduce the incidence of bad tests and b) to make sure multiple people actually know what tests we have and how they work
12:38
<jgraham>
It isn't to prevent all possible errors
12:38
<jgraham>
(if we wanted that we would, um, well actually I don't know.
12:38
<jgraham>
Have formal proofs and tests in Coq?)
12:39
<odinho>
zcorpan: You don't need to call it that, because you should have the branch in your personal fork anyway.
12:40
Ms2ger
wonders why jgraham enjoys talking about Coq so much)
12:40
<MikeSmith>
cool, when I try to to start my firefox nightly, I just get a window with an error: "XML Parsing Error: undefined entity Location: chrome://browser/content/browser.xul Line Number 506, Column 9"
12:40
<odinho>
zcorpan: You can call the branch bamboo-bowser if you so please, and then do a PR.
12:40
<MikeSmith>
and then I don't get any browser chrome at all so can't actually navigate to anything
12:40
<odinho>
zcorpan: The reason we have the submission branches right now is because we had some stuff already submitted and needed branches for those :-)
12:42
<zcorpan>
odinho: but my branch name will be visible in the PR right?
12:42
<odinho>
zcorpan: Yes. :-) But not really highlighted.
12:43
<odinho>
zcorpan: So don't call your public branch too offensive things.
12:43
<darobin>
are we moving on to making profound jokes about how jgraham likes Coq?
12:43
<jgraham>
FWIW the only thing that's annoying in PR branch names is overuse of "master"
12:43
<jgraham>
git has local branches for a reason, people
12:43
<odinho>
zcorpan: You can, however, call your local ones offensive things, and then just push to a more neutral name for the public repo. :-) git push origin omg_i_hatez_jgraham:shiny_ponies
12:43
<jgraham>
This is not mercurial
12:44
<darobin>
MikeSmith: where we're going, the browser is everything — you don't need chrome. Firefox is just showing you the future there.
12:44
<zcorpan>
submission/odinhos-mom-so-fat-she-would-break-a-branch
12:44
<odinho>
zcorpan: You got it! :D
12:44
<darobin>
haha
12:56
<zcorpan>
Ms2ger: https://github.com/w3c/web-platform-tests/pull/82
12:56
<Ms2ger>
Thanks
13:43
<annevk>
Ms2ger: I put fetch in a separate file for now
13:44
<annevk>
Ms2ger: and put all CORS references towards W3C/TR/cors
13:44
<annevk>
A little unfortunate that rewriting CORS takes the "living" bit out of it, but it was fairly stable anyway.
13:57
<Ms2ger>
annevk, *shrug*
13:57
<Ms2ger>
On another note, Hixie takes bug reports on IRC now?
13:57
<Ms2ger>
(re http://html5.org/tools/web-apps-tracker?from=7832&to=7833)
13:57
<annevk>
Ms2ger: shrug what?
13:57
<annevk>
Ms2ger: me making xref work again?
13:57
<Ms2ger>
annevk, CORS pointing to TR
13:58
<annevk>
Ms2ger: didn't really see a better way for now
13:58
<Ms2ger>
Yeah
13:59
<Ms2ger>
You're 80% done with killing it anyway, no? ;)
13:59
<annevk>
that's the Q2 goal, 80% is about the level of done-ness CORS has now, so then I can update the references and everything :)
14:00
<annevk>
One of the big problems with Fetch will be that the interface is somewhat different because we've used so many different conventions in the past
14:02
<darobin>
annevk: are you thinking of specific compatibility issues for new HTMLH1Element beyond toString()?
14:03
<darobin>
because I reckon it's not the end of the world if we say it that it always returns "[object HTMLHeadingElement]"
14:04
<darobin>
there's __proto__ but I'd be a bit surprised if that showed up
14:05
<gsnedders>
instanceof is more likely.
14:05
<gsnedders>
Or just constructor directly.
14:05
<darobin>
well instanceof HTMLHeadingElement would still be true
14:05
gsnedders
hasn't read enough context
14:05
<darobin>
the thing is I wonder if people actually access constructor for non-constructable objects
14:06
<jgraham>
I don't think there was any context given
14:06
<annevk>
darobin: no, just "change" is dangerous
14:06
<annevk>
when it comes to compat anyway
14:06
<annevk>
but we should try
14:06
<darobin>
yeah, I know, but we try to guess impact
14:07
<annevk>
jgraham: you're not reading your email synchronously with IRC?
14:07
<darobin>
I guess there could be people doing el.constructor === HTMLHeadingElement instead of instanceof
14:07
<Ms2ger>
annevk, quite the conservative :)
14:08
<annevk>
Ms2ger: can we kill hasAttributes() now?
14:08
<Ms2ger>
I'd be happy to
14:08
<annevk>
Ms2ger: how hard it is to measure its usage?
14:09
<annevk>
Ms2ger: we could at least do the thing BlinKit did, move it to Element
14:09
<Ms2ger>
annevk, I think AryehGregor did telemetry for Range.detach, it's probably feasible to crib from that
14:09
<annevk>
Ms2ger: should I file a bug for you?
14:10
<annevk>
I wonder when we're gonna try Attr exodus
14:10
<Ms2ger>
Sure, I'll add it to the bottom of my todo list :)
14:10
<Ms2ger>
Line 350 or so
14:10
<annevk>
right
14:23
<Ms2ger>
gsnedders, what was the motivation for https://github.com/gsnedders/py-regexp-builder btw?
14:28
<gsnedders>
Ms2ger: Originally? No idea. Now? Partly the fact it had lain around for too long, and partly to build regexp of all "invisible" characters.
14:28
<Ms2ger>
I see
14:43
<gsnedders>
Ms2ger: Too slow to run on each load with CPython (~0.7s to build it, mostly spent iterating over the whole codespace and checking category), but takes ~0.1s with PyPy
14:48
<gsnedders>
(Down to ~0.5s with CPython, no difference with PyPy)
14:54
marcosc
thinks someone needs to W3C meme this: https://github.com/darobin/respec/pull/197#issuecomment-16381951
14:55
<Ms2ger>
People are idiots
14:55
<Ms2ger>
News at 11
14:57
<annevk>
Film at 11, idiot
14:57
<annevk>
Badum-tish!
15:00
<annevk>
marcosc: it's like it's 2004 again with stuff like that
15:01
<Ms2ger>
annevk, :D
15:01
<annevk>
marcosc: so yeah man, last decade didn't happen, stable pointers are much more important than whatever you learned meanwhile
15:02
<jgraham>
2004, or 2014 at W3C
15:03
<marcosc>
:)
15:03
<annevk>
It's like visiting a third world country, technology is about a decade behind. DVDs are still hip and you don't have to look hard to find VHS.
15:05
<marcosc>
I was in Greece the other day, and the airport X-Ray machine assured me that it was safe for diskettes.
15:06
<marcosc>
It has very helpful picture... oh, and film too... but only some types.
15:07
<annevk>
I almost jokingly added France to that sentence. Went to a club there once on vacation, the "pop" music was from about seven years ago.
15:08
<annevk>
To be fair, that was not in Paris, but somewhere middle-of-nowhere.
15:08
<darobin>
you completely missed the point annevk
15:08
<darobin>
the *whole* point of provincial clubs in France is to have dated music
15:08
<darobin>
it's as hip as it gets
15:09
<darobin>
the whole "latest and greatest DJ" parisian routine is so tedious
15:09
<annevk>
darobin: aw man, I guess that's why my French teacher kept making fun of me
15:09
<darobin>
I seriously would only consider going to a club in France if it's outside Pari
15:09
<darobin>
*Paris
15:09
<darobin>
it's all wedding music
15:10
<darobin>
meaning you can booze your head off and dance like a moron
15:10
<marcosc>
you mean, they wouldn't laugh at my dancing there?
15:10
<marcosc>
you mean, I could be free like a dancing butterfly?
15:11
<annevk>
darobin: I've found you don't require a club for that
15:11
<darobin>
marcosc: exactly
15:12
<darobin>
annevk: well, I don't require a club for that, but in a parisian club I'd promptly get kicked out :)
15:12
<annevk>
hehe
15:14
<jgraham>
Parisian clubs are for eating cheese?
15:16
<darobin>
they're for cheese eating surrender monkeys for sure
16:19
<zcorpan>
is there an up-to-date table inspector that implements the spec's td-th mapping?
16:20
<zcorpan>
james.html5.org is no more
16:21
<zcorpan>
jgraham: do you have http://james.html5.org/tables/table_inspector.html around somewhere?
16:41
<jgraham>
zcorpan: So I don't know what happened to that code. Maybe it is on some other computer. I will have a look
16:41
<zcorpan>
jgraham: if you find it, please dump it on github or something :-)
16:42
<Hixie>
ms2ger: if i happen to notice it and happen to not have any edits open and happen to be in the mood... which is to say, don't count on it :-)
16:43
<jgraham>
zcorpan: Sure
16:43
<zcorpan>
why does Audio API need the legacy constants?
16:43
<annevk>
oh wtf
16:43
<annevk>
public-svg-wg has a limited mailing list?
16:43
<annevk>
that's super annoying
16:44
<annevk>
(limited as in subscribers only)
16:44
<zcorpan>
or why does Audio API want to use both legacy constants and strings instead of picking one or the other?
16:44
<shepazu>
annevk, it's only for logistics, the technical talk happens on www-svg
16:45
<annevk>
oh, forgot about www-svg :/
16:46
annevk
forwards
16:47
<annevk>
ta shepazu
16:56
<annevk>
I do agree with jgraham. The longer that thread goes on, the more it reads like TC39 should just fold into W3C...
17:10
<MikeSmith>
win 11
17:11
<MikeSmith>
annevk: so we seriously should help do that
17:12
<annevk>
zcorpan: they want to move to new API-style, but don't break compat?
17:12
<zcorpan>
annevk: who?
17:12
<annevk>
zcorpan: Audio
17:14
<zcorpan>
annevk: ah. yeah, i guess. everywhere else we haven't done that, though, we just didn't change the legacy APIs and changed over completely the APIs that were not widely used yet
17:14
<annevk>
I expect we'll get new APIs going forward for various unchanged legacy stuff
17:14
<zcorpan>
annevk: i'm not convinced that supporting both will end up being a win since it means both will be used in the wild and authors will be more confused
17:14
<annevk>
e.g. new APIs for network requests that don't suck as much
17:15
<annevk>
zcorpan: given that not everyone has implemented the Audio API I'm kinda surprised about the compat
17:15
<annevk>
zcorpan: I guess iOS shipped with it...
17:15
<zcorpan>
yeah i would have thought it isn't widely used yet
17:18
<GPHemsley>
TIL <blink> is not in any spec
17:20
<Hixie>
it's in html
17:20
<Hixie>
http://whatwg.org/html#blink
17:20
<annevk>
Hixie: we killed it from Gecko btw
17:20
<annevk>
s/from/in/
17:21
<Hixie>
lame
17:21
<annevk>
*shrug* ;)
17:21
<GPHemsley>
Hixie: https://bugzilla.mozilla.org/show_bug.cgi?id=857820#c57
17:22
<Hixie>
the parser did something with blink?
17:22
<Hixie>
o_O
17:23
<annevk>
looks like the old Gecko parser
17:23
<Hixie>
the only requirement in the html spec is that the ua.css file have blink { text-decoration: blink }
17:23
<GPHemsley>
Hixie: I was referring to this: "And also, <blink> element isn't defined any Web Standard specs."
17:24
<Hixie>
GPHemsley: what about it?
17:25
<Hixie>
people not knowing what's in all the specs is hardly surprising, there's tons of them and they're long. :-)
17:25
<GPHemsley>
Oh, are we splitting hairs about "in" vs. "defined in"?
17:25
Hixie
barely knows what's in the HTML spec, let alone anything else :-)
17:26
annevk
comments
17:28
<zcorpan>
there are surprisingly few uses of `const` in html
17:29
<zcorpan>
maybe the parser thing is about HTMLElement vs HTMLUnknownElement?
17:31
GPHemsley
had no idea <marquee> had so many possibilities.
17:35
<zcorpan>
i wish browsers would do a better job at animating the <marquee> instead of doing 10 pixel jumps like it's 1994
17:36
<Hixie>
ah yeah maybe <blink> is now HTMLUnknownElement
17:36
<Hixie>
spec does say it should be HTMLElement
17:36
<Hixie>
but i can add it to the list of Unknowns if nobody does that
17:38
<zcorpan>
http://software.hixie.ch/utilities/js/live-dom-viewer/saved/2214 - opera/gecko HTMLElement, safari/chrome HTMLUnknownElement
17:40
<Hixie>
in other news, how should we handle changing the start/end times on a cue
17:40
<Hixie>
essentially remove the cue and re-add it?
17:40
<Hixie>
even if that moves it around?
17:40
<Hixie>
or should we try hard not to move it
17:40
<Hixie>
also we still don't show cues when they are added if they should be active, iirc
17:41
<zcorpan>
in IE10 it's "HTMLPhraseElement"
17:41
<Hixie>
wtf is HTMLPhraseElement
17:42
<zcorpan>
i asked about this a few years ago. apparently they had these interfaces internally all along, but didn't expose them until IE8 or some such, when catching up with the specs
17:42
<zcorpan>
and they didn't prioritize matching the spec exactly, or something
17:42
<Hixie>
...ok
17:43
<GPHemsley>
"Implements the document object model (DOM) representation of the phrase elements such as em, strong, dfn, code, samp, kbd, var, cite, abbr, and acronym." http://msdn.microsoft.com/en-us/library/ie/hh869728%28v=vs.85%29.aspx
17:44
<Hixie>
hmm... we do support _removing_ cues dynamically
17:45
<zcorpan>
Hixie: i think moving cues around when changing the times isn't a problem
17:45
<Hixie>
well if all you're doing is extending the end time
17:45
<Hixie>
it'd be sad if it moved
17:46
<Hixie>
i think in practice we don't need to move
17:46
<Hixie>
i just need to check if the active flag should be set, and repaint the cues if it changed
17:46
<Hixie>
the question is more about events
17:46
<Hixie>
i'm thinking we don't fire events if you manually change the time and active changes
17:46
<zcorpan>
oh moving it can change the rendering huh... didn't consider that. i was just thinking of "move" as in placement in the list of cues
17:47
<Hixie>
right
17:47
<Hixie>
e.g. if the cue was placed in a suboptimal place, then all the other cues went away
17:48
<zcorpan>
then i think i agree
17:49
<Hixie>
or... _or_... we could run the "When the current playback position of a media element changes" steps
17:49
<Hixie>
and that would fire all the events magically! and update the active flag and everything!
17:49
<Hixie>
oh this is so the right solution
17:50
<Hixie>
yes yes yes, this is the way to do it.
17:51
<Hixie>
ok i need a name for this algorithm
17:51
<Hixie>
"the 'time moves on' steps"?
17:52
<Hixie>
ok unless someone has a better idea, i'm calling these the /time marches on/ steps.
18:05
<Hixie>
hm, can you have a texttrack without a media element...
18:06
<Hixie>
i guess so, orphan <track> for one
18:06
<Hixie>
hmm
18:17
<Hixie>
are there any APIs (e.g. XHR maybe?) that do the blob: delayed preservation magic needed for https://www.w3.org/Bugs/Public/show_bug.cgi?id=17765 ?
18:19
<annevk>
Hixie: there's no blob URL special casing anywhere currently
18:20
<annevk>
Hixie: we want to lift the same-origin restrictions on blob URLs (make them work similar to data URLs, which when pasted in somewhere, work)
18:20
<Hixie>
the origin thing seems orthogonal
18:21
Hixie
doesn't really want to be the first one to do this special casing for blob:
18:21
<annevk>
Hixie: except that we track the revoking thing on a per-global basis
18:21
<Hixie>
and i find it hard to believe that all the specs are going to be updated for this
18:21
<Hixie>
like, is CSSOM going to say that when you set 'border-image-source', you have to do something special?
18:21
<annevk>
Hixie: implementers have to decide what to do here I think
18:22
<annevk>
Hixie: zewt suggested adding this to the URL parser, but I'm not sure if that actually solves the issue as that can be delayed in CSS too
18:23
<Hixie>
jesus wept, what happened on https://www.w3.org/Bugs/Public/show_bug.cgi?id=17842
18:24
<Hixie>
53 book-length comments
18:24
<Hixie>
well the actual semantics here are imho very weird, urls are supposed to be idempotent and this breaks that invariant
18:24
<annevk>
Hixie: basically, I think the whole autoRevoke thing is kind of a joke
18:25
<Hixie>
not a joke, but yeah, i'm rather skeptical of it
18:25
<Hixie>
but i don't have a better proposal
18:25
<annevk>
has it been implemented?
18:25
<Hixie>
no idea
18:26
<annevk>
basically currently it seems it cannot work as starting a fetch will do that fetch async
18:26
<annevk>
except for sync XHR
18:26
<annevk>
so if UAs implemented it, they'd break sites
18:27
<annevk>
because autoRevoke was not the default before
18:28
<Hixie>
hm?
18:29
<Hixie>
in other news, i think it's possible some people think that the more they talk about a feature, the more likely it is that it'll get added to the spec
18:29
<Hixie>
but since i deal with bugs on a least-recently-touched basis, it actually has the opposite effect...
18:32
<annevk>
Hixie: createObjectURL() in the spec defaults to autoRevoke true; that's not implemented atm afaik
18:33
<annevk>
Hixie: also, <img>.src = blobURL will always load async, at which point blobURL is already revoked
18:44
<marcosc>
Hixie, re: Bug 17842 ... it's not the case that people are just bla bla-ing to make it seem important. It's actually quite controversial what is being asked in the bug... because of how radios on mobiles work.
18:45
<annevk>
Hixie: I'm emailing webapps
18:45
<marcosc>
Hixie: in summary, we want a "lazy" attribute :) Lazy === hey browser, load it when you want.
18:46
<zcorpan>
marcosc: Hixie's just saying that if you want him to look at the bug, you need to stop adding more comments
18:46
<marcosc>
zcorpan: I know, totally get that.
18:47
<marcosc>
zcorpan: what I was trying to say was that the original bug had made some incorrect assumptions, so lots of people had opinions about it.
18:47
<marcosc>
I was actually going to jump in the other day to tell them to stop proposing solutions and just wait for Hixie
18:48
<marcosc>
But some good new material was actually brought forward
18:48
<marcosc>
so it was better to just leave the conversation going
18:48
<zcorpan>
also, in my experience saying "stop talking" doesn't work but just adds more noise :-P
18:49
<marcosc>
heh, rock and a hard place
18:49
<marcosc>
I think we reached consensus in the bug, so either Ilya or Yoav should be able to provide Hixie with a summary
18:50
<zcorpan>
if you know of someone wanting to implement it right now, you can ask Hixie to fast track the bug
18:53
<marcosc>
zcorpan: yeah, I don't know of any actual interest from browser folks.
18:53
<zcorpan>
ok
18:54
<marcosc>
I mean, there a few people form Google, at least one from Opera, ... but I guess until there is actual bugs filed in the right places...
18:55
<Hixie>
annevk: the idea zewt was putting forward was that <img>.src = blobURL should cause the load to be sync so that the data was preserved for that <img>, iiuc
18:55
<Hixie>
marcosc: oh i wasn't suggesting it was just bla bla, sorry if that was the implication!
18:56
<annevk>
Hixie: hmm maybe we could make http://fetch.spec.whatwg.org/#concept-fetch work that way
18:56
<annevk>
Hixie: even so, that doesn't solve it for CSS I think
18:57
<Hixie>
annevk: yeah, i dunno. arun and zewt are the right people to figure out the solution, i'm just hoping another spec fixes it first because i don't want to be the trailblazer here when i'm skeptical anyone else will follow.
18:57
<annevk>
Hixie: I don't think you should fix it
18:57
<annevk>
Hixie: it should either happen in fetch or URL
18:58
<annevk>
Hixie: and everyone will have to invoke fetch synchronously I suppose, but that's okay
18:58
<Hixie>
well that would be even better, sure. right now that doesn't fit the proposa.
18:58
<Hixie>
proposal.
18:59
<Hixie>
in other news: does anyone recall/know if the difference in <a>'s activation behaviour and <area>'s activation behaviour, with respect to how to handle target="" in sandbox=""ed iframes, is intentional or not? :-)
19:01
<Hixie>
looks like i just forgot <area> when fixing http://www.w3.org/Bugs/Public/show_bug.cgi?id=13894
19:02
<annevk>
Hixie: http://lists.w3.org/Archives/Public/public-webapps/2013AprJun/0242.html (re blob URLs)
19:04
<Hixie>
well we can't always invoke fetch
19:04
<zcorpan>
hsivonen: MikeSmith: is there a story for automated tests for v.nu yet?
19:04
<Hixie>
e.g. <img> might want to delay forever if we add this new defer=""
19:04
<Hixie>
and <embed> has to fetch async for compat reasons
19:04
<Hixie>
(base url has to be resolved after setting)
19:04
<Hixie>
(after settings source url, that is)
19:04
<Hixie>
typos galore!
19:05
<annevk>
re-resolved?
19:05
<annevk>
Hixie: you can invoke fetch sync and then still wait
19:06
<Hixie>
how does that work?
19:06
<annevk>
Hixie: the only thing invoking fetch would do is take a reference to the object and then return early
19:06
<MikeSmith>
zcorpan: from W3C we have a GSOC proposal for a student to do some work related to that, and we have a great candidate who's submitted an GSOC application for it already
19:06
<Hixie>
yes my point is we don't even want to resolve the url in some cases
19:06
<Hixie>
e.g. <embed>
19:06
<annevk>
Hixie: or <a href> I suppose
19:06
<Hixie>
and in some cases we don't know what url to resolve, e.g. <img srcset>
19:06
<MikeSmith>
zcorpan: http://www.w3.org/2013/03/gsoc2013#guidelines
19:07
<Hixie>
especially <img srcset defer>
19:07
<Hixie>
or whatever we call it
19:07
<MikeSmith>
zcorpan: http://www.w3.org/2013/03/gsoc2013#validator-nu-testing
19:07
<annevk>
mkay
19:08
<Hixie>
like, what should we do if someone sets an img srcset to "blob:a 1x blob:b 2x blob:c 3x" ?
19:08
<Hixie>
er, with commas
19:08
<Hixie>
you get the idea
19:08
<annevk>
you figure out which to fetch and do it?
19:09
<Hixie>
and you don't revoke the other two?
19:09
<Hixie>
what if the user zooms in and out, or moves the window across to a low res screen then a high res screen?
19:09
<Hixie>
do you still have the images around? when do you forget them?
19:11
<annevk>
when you run the autorevoke stuff
19:12
<annevk>
not sure there's much winning to be had here
19:18
<Hixie>
MikeSmith: re https://www.w3.org/Bugs/Public/show_bug.cgi?id=21553#c9, by '"the main content" of each <article> (currently marked up as <li> with class="postitem".)', i meant the parenthetical to apply to "<article>", not to "the main content of each <article>".
19:18
<Hixie>
sorry for the ambiguity
19:24
<MikeSmith>
Hixie: ah, OK
19:32
<zcorpan>
MikeSmith: nice. keep me updated on that :-)
19:34
<zcorpan>
MikeSmith: it would be nice if tests could be written to be validator-agnostic and live on web-platform-tests or so
19:39
<zcorpan>
though maybe people following web-platform-tests don't want validator tests there
19:39
<jgraham>
Well
19:39
<jgraham>
I don't plan to review them at least
19:39
<jgraham>
But with critic filters I don't have to!
21:06
<annevk>
'[The "or" operator is not yet supported by ReSpec.]'
21:06
<annevk>
I don't even
21:06
<jsbell>
annevk: context?
21:06
<annevk>
http://lists.w3.org/Archives/Public/public-device-apis/2013Apr/0029.html
21:07
<Hixie>
o_O
21:08
<annevk>
ReSpec makes specs both way harder to read and then it does not even let you write a spec in the right way
21:08
<jsbell>
Seems to "work" for me, although in (foo or bar) foo and bar don't end up linked correctly. That said, linking to webidl default types seems broken in respec anyway at the moment (reported to robin)
21:52
<Hixie>
TabAtkins: ping https://www.w3.org/Bugs/Public/show_bug.cgi?id=18026 comment 8 paragraph 2
22:01
<rillian>
did the websec-mime-sniff draft become http://mimesniff.spec.whatwg.org or is one a fork of the other?
22:06
<Hixie>
rillian: yeah, iirc ietf basically didn't like it existing
22:06
<Hixie>
abarth would know more
22:07
<TabAtkins>
Hixie: No, getting a selector to match elements whose contents are numeric is not realistic in the short term. It's basically equivalent in difficulty to getting one which matches a regex against text contents, which has been shot down for performance reasons for a long time.
22:08
<Hixie>
TabAtkins: can you comment to that effect on the bug? would be much appreciated. (also if you have any other ideas for addressing that use case, cos i'm all out of ideas on that front)
22:08
<rillian>
abarth: IIRC we were trying to limit things to the first 500 bytes and/or only mask/compare
22:08
<TabAtkins>
Yes, I can comment. I think the right thing to do is simply having the column selectors/combinator that Selectors 4 has, which is sufficient.
22:08
<rillian>
which is why the mp3 and webm sniffing doesn't really work. Does that match your recollection?
22:09
<rillian>
Hixie: thanks
22:09
<Hixie>
TabAtkins: cool, thanks
22:10
<TabAtkins>
Hixie: Though... Hm, testing an attribute value for numeric-ness would probably work.
22:10
<TabAtkins>
It's roughly equivalent in difficulty to the existing attr operations.
22:10
<TabAtkins>
(Fundamental problem with testing text content is what to do with non-child text nodes.)
22:13
<Hixie>
i once proposed [attr>4] and related selectors
22:13
<Hixie>
fwiw
22:13
<TabAtkins>
Yeah, I wanna put those into either Selectors 4 or 5.
22:13
<TabAtkins>
We're trying to shut down 4, so it'll probably make 5.
22:18
<TabAtkins>
I wonder what kind of syntax would make sense to test if an attr was numeric.
22:18
<TabAtkins>
I mean, you could definitely do :matches([foo>0], [foo<=0]), but that's a bit verbose.
22:20
<Hixie>
TabAtkins: yeah
22:21
<Hixie>
TabAtkins: also, you should clearly switch to an unversioned development model so you could add things whenever :-)
22:22
<TabAtkins>
Hixie: As soon as we get CR of a spec, we push the unversioned url to a new version, so we can work on whatever we were waiting on.
22:22
<TabAtkins>
The only time versioning gets in our way now is the period between "ok, time to stabilize and finish what we've got" and "CR published!".
22:23
<TabAtkins>
Hmm, since we already have a type flag for case-insensitive matches, maybe we can just use that. [foo n] matches if the foo attribute's value is numeric.
22:24
<TabAtkins>
Or just make a new comparison. [foo <=> 0] tests for numericness. ^_^
22:25
<Hixie>
there's an unversioned url?
22:26
<TabAtkins>
Now there is! Just drop the number (or dash-number) from any dev.w3.org css url, and you'll get the latest draft.
22:26
<Hixie>
ooo
22:26
<TabAtkins>
dev.w3.org/csswg/css-flexbox-1 and css-flexbox both point to the same spec, for example.
22:26
<TabAtkins>
We're planning to apply the same policy to our /TR drafts.
22:27
<TabAtkins>
All the old css3-foo links still work, because we added redirects, but the preferred url form now is "css-foo", or "css-foo-N" if you want a specific version.
23:08
<Hixie>
abarth: ping https://www.w3.org/Bugs/Public/show_bug.cgi?id=19662#c7
23:09
<abarth>
Hixie: hi
23:09
<abarth>
Hixie: looking
23:12
<abarth>
Hixie: too complicated for me to think about right now, sorry. I've added it to my asana for this week
23:14
<Hixie>
thanks!
23:16
<Hixie>
abarth: while i have you here, are you still working on web origin?
23:16
<Hixie>
abarth: in particular, it seems a bit ambiguous what "host" means when the origin has an ipv6 address in it
23:16
<Hixie>
abarth: whether it has the []s or not
23:16
<abarth>
yeah, we never resolved that issue
23:17
<abarth>
do you know which way it ought to be?
23:19
<Hixie>
i think square brackets included?
23:19
<Hixie>
but i'm not really sure
23:19
<Hixie>
it would match url.spec.whateg.org
23:19
<Hixie>
whatwg even
23:20
<Hixie>
abarth: but really my question isn't so much about the serialisation, but about whether "the host part of the tuple" contains the square brackets or not
23:20
<Hixie>
abarth: and that i've really no idea if it should or not
23:21
<abarth>
this gets into how much browsers canonicalize IP addresses
23:21
<abarth>
As I recall, IE and Chrome do a lot of canonicalization
23:21
<abarth>
whereas Firefox and Safari do very little canonicalization
23:22
<abarth>
I don't remember the details about this particular bit of canonicalization
23:22
<Hixie>
url.spec.whatwg.org requires a specific form
23:22
<abarth>
for my money, i'd err on the side of more canonicalization
23:22
<abarth>
so, if you supply and IPv4 address in octal, it will end up as a dotted quad?
23:23
<Hixie>
no idea if he supports octal
23:24
<Hixie>
oh, ipv4 parsing isn't yet defined
23:24
<Hixie>
indeed host parsing isn't defined other than ipv6
23:24
<abarth>
i see
23:25
<abarth>
if folks like the general approach of canonicalizing IP addresses, then we should canonicalize IPv6 address to include the [ ]
23:25
<Hixie>
k
23:26
<Hixie>
so I should assume the 'host' part of an origin tuple includes the brackets?
23:26
<abarth>
yes
23:26
<Hixie>
is there somewhere i can file a bug on web origin, or does that go on your list too? :-)
23:32
<abarth>
Hixie: you can file an issue in this github project: https://github.com/abarth/websec
23:33
<Hixie>
k
23:35
<gsnedders>
hober: I presume you don't mind if I base the new html5lib documentation on http://edward.oconnor.cx/2009/08/djangosd-html5lib? :)
23:46
<gsnedders>
jgraham: We (html5lib) got mentioned in PEP-411!
23:50
<hober>
gsnedders: fine by me! i think most or all of my site is cc by sa, iirc
23:50
<TabAtkins>
Ah, you crazy viral copylefters.
23:50
<gsnedders>
hober: Yeah, it says so. But would be nice to have docs same license as everything else, and given I'm basically just ripping off your structure :P
23:51
<hober>
:)
23:51
<hober>
i forgot all about that talk
23:51
gsnedders
was just Googling for html5lib :)