05:21
<MikeSmith>
Hixie: which API was that? (for creating new custom <input type=> values)
05:32
<Hixie>
MikeSmith: which, the web controls one or the web components one?
05:33
<MikeSmith>
Hixie: I mean the web controls one
05:33
<Hixie>
http://www.whatwg.org/specs/web-controls/current-work/
05:34
<Hixie>
(defunct now, aria basically made it moot. unfortunately, aria solved the problem in a far less compelling way.)
05:34
MikeSmith
looks
05:34
<MikeSmith>
wow
05:35
<MikeSmith>
dunno if I ever saw this spec before
05:35
<MikeSmith>
if I did I forgot
05:36
<Hixie>
it would have solved the custom controls problem and the accessibility problem in one fell swoop while simultaneously preventing the conflicting state problem that ARIA brought us
05:36
<Hixie>
you can tell how old that spec is, it still uses respec-style formatting for interface members :-)
05:36
<MikeSmith>
heh
05:37
<Hixie>
bed time, bb in a few hours :-)
05:38
<MikeSmith>
nn
07:21
<dekiss>
Good morning GGO BUILD THE WEEEEEEEEEEEEB!!!
09:28
<ondras>
Domenic_: ?
09:36
<ondras>
well, probably anyone else can answer
09:36
<ondras>
is "catch" a wise name for a promise's method? is "catch" generally a suitable JS function name, being a reserved word?
12:28
<MikeSmith>
can anybody figure out what https://www.w3.org/Bugs/Public/show_bug.cgi?id=24111 is asking for?
12:34
<jgraham>
MikeSmith: Yeah. They want li {color:green} type rules to also turn the marker green
12:34
<MikeSmith>
ah
12:34
<ondras>
which somewhat makes sense if you ask me
12:34
<ondras>
a pseudoclass would be also reasonable
12:34
<MikeSmith>
that's what he means by "inherit"
12:35
<MikeSmith>
I guess
12:35
<ondras>
yeah
12:36
<jgraham>
It might make sense, but presumably it isn't backward compatible
12:36
<MikeSmith>
oh, "CKEditor Project Lead"
12:36
<MikeSmith>
anyway it's a CSS issue issn't it?
12:36
<ondras>
that's why he mentioned a special property for turning this feature on
12:36
<jgraham>
Yes
12:40
<MikeSmith>
doesn't CKEditor has some problems? like doing some kind of stupid stuff and they refuse to fix it?
12:41
<ondras>
even if they did, is that relevant for considering a feature request? :)
12:43
<MikeSmith>
well I'm not considering it anyway
12:55
<Ms2ger>
Wasn't it CKEditor that used element.all?
12:56
<gsnedders>
Ms2ger: Amongst other things
12:56
<gsnedders>
CKEditor does lots of evil stuff, from memory
12:57
jgraham
cries
12:58
<jgraham>
I hate the web
12:58
<jgraham>
Well specifically
12:58
<jgraham>
I hate that thing where you develop something that works fine in one browser
12:58
<jgraham>
and fails in all others
12:58
<jgraham>
Without so much as an error on the console
13:03
<Ms2ger>
Sounds like the web alright
13:05
<gsnedders>
WEB 3.0!
13:05
<jgraham>
Oh
13:05
<jgraham>
It was all my fault
13:05
<jgraham>
Come back web, all is forgiven
13:20
<zcorpan>
jgraham: that doesn't appear to be what they want. li {color:green} already makes the marker green. They want <li><font color=green>foo</font> to make the marker green
13:20
<odinho>
lol
13:20
<odinho>
crazy
13:22
<jgraham>
zcorpan: Oh
13:25
<annevk>
ondras: catch is no longer reserved
13:34
<jgraham>
http://hoppipolla.co.uk/410/workers.html
13:34
<jgraham>
based on an adaptation of Ms2ger's test runner and a script to create reports from JSON
13:34
<jgraham>
Not finished yet obviously, but better than editing a wiki page I think
13:35
<MikeSmith>
hell yeah
13:35
<MikeSmith>
this is great
13:36
<Ms2ger>
Mine had neat colours ;)
13:37
<jgraham>
Ms2ger: Funny, I wa just going to say that someone should do the visual design
13:37
<jgraham>
Because it turns out that a) I suck at it and b) I get to trying to do it and want to scoop my own eyes out
13:37
<jgraham>
https://github.com/w3c/web-platform-tests/commit/4f68f6e66ab2ec1b62a2d8c035e845f35707346d is the code (jgraham/runner branch in web-platform-tests)
13:38
<Ms2ger>
The one thing that would be nice is having lines between the tests
13:38
<Ms2ger>
s/tests/test files/
13:38
<MikeSmith>
I think it's fine as-is
13:38
<jgraham>
Yeah, sure there are simple things I could do that would make it not obviously terrible
13:38
<MikeSmith>
reading test results should be hard and challenging
13:38
<jgraham>
Heh
13:39
jgraham
will go into the office now
13:40
<Ms2ger>
Isn't it, like, afternoon there?
13:40
<annevk>
he missed mozlunch
13:41
<annevk>
not many people around
13:41
<Ms2ger>
<meta charset="utf8"></meta>
13:41
<annevk>
so close
13:44
<annevk>
Ah, I already analyzed gbk vs gb18030 before!
13:44
<annevk>
hsivonen: http://lists.w3.org/Archives/Public/www-archive/2012Apr/0030.html
13:44
<ondras>
annevk: ah, I guess I need to update my version of google chrome compiler then...
13:44
<annevk>
It seems I just draw conclusions that were wrong for gb18030 UTFness.
13:46
<annevk>
In fact, if I use Firefox' gb18030 table I think that will make gb18030 a UTF and we can still use it for gbk (and just make gbk an alias for gb18030)
13:50
<gsnedders>
Isn't taht what some browser already does?
13:54
<annevk>
Presto had something close to that
13:54
<annevk>
But still had a flag, as seen in the Encoding Standard
13:54
<annevk>
However, in Presto gb18030 was not a UTF
13:55
<gsnedders>
What stops it from being a UTF in Presto/Chrome?
13:56
<annevk>
In Chrome gb18030 is a UTF.
13:56
<annevk>
In Presto a bunch of two-byte sequences does not map to PUA
13:59
<darobin>
given how often people will have to do template.content.cloneNode(true), with the likelihood of getting it wrong, I wonder if we shouldn't have a template.instance() method that just sugars that
14:01
<Ms2ger>
jgraham, want to send me the results for workers so I can test changes?
14:06
<reggna>
So, how do I remove a TextTrack from a TextTrackList?
14:08
<annevk>
reggna: looks like you don't
14:08
<reggna>
Yea, I can't find anything either.
14:09
<reggna>
There's an event though.
14:09
<annevk>
reggna: well if you remove e.g. the corresponding <track> element
14:09
<annevk>
reggna: but if they're all internal to the resource I don't think you can manipulate them
14:19
<reggna>
I'm a bit saddened over this.
14:19
<reggna>
But thanks for the quick answer, annevk. : )
14:19
<annevk>
reggna: if you have a use case, file a bug
14:19
<annevk>
reggna: best way to get a new feature
14:20
<reggna>
Yea, my use case is "this would be very convenient for my test case". : P
14:21
<reggna>
So not that valid, unfortunately.
14:30
<annevk>
hah
14:48
<jgraham>
Ms2ger: Specially ofr you I have amde it really ugly so that someone is forced to improve the style
14:49
<jgraham>
Ms2ger: Do you want the json files?
14:49
<Ms2ger>
jgraham, whatever the input to report.py is
14:52
<jgraham>
Ms2ger: http://hoppipolla.co.uk/410/workers/
14:53
<Ms2ger>
Thanks
15:07
<Ms2ger>
Uh: https://github.com/w3c/web-platform-tests/blob/76e0d1ba75c95f95cb6e24c42f84c15ca18c39ef/workers/WorkerGlobalScope_removeEventListener.htm#L45
15:09
<annevk>
IE10 doesn't support overrideMimeType?!
15:09
<annevk>
Bah
15:11
<Ms2ger>
jgraham, r? https://pastebin.mozilla.org/3789796
15:15
<jgraham>
Ms2ger: Fancy merging that with the change I just pushed?
15:15
<jgraham>
(We have some of the same bugfixes at least)
15:16
<Ms2ger>
cellspacing=0?
15:16
<jgraham>
:p
15:16
<jgraham>
I was on a train and couldn't remember the right kind of CSS
15:18
<Ms2ger>
Did you also forget to add report.css?
15:19
<jgraham>
Probably :)
15:20
<jgraham>
Added
15:20
<Ms2ger>
Ta
15:27
<Ms2ger>
jgraham, how about now? https://pastebin.mozilla.org/3789871
15:31
<jgraham>
Ms2ger: Would be nice to add the style to the stylesheet.
15:31
<jgraham>
Not sure about the use of multiple <tbody>, but I guess it works
15:32
<jgraham>
Something has crashed so I can't scroll and see the rest of the patch. Or switch applications
15:32
<Ms2ger>
What do you mean about adding style to the stylesheet?
15:33
<jgraham>
Oh, I misread
15:33
<jgraham>
And I can scroll by mouse
15:34
<jgraham>
s/mpouse/keyboard/
15:34
<jgraham>
Sure, that looks fine
15:34
<jgraham>
please push
15:36
Ms2ger
pushes
15:39
<Ms2ger>
Oh, tobie's fellowship is ending?
15:39
<jgraham>
Yup
15:39
<jgraham>
Although I don't know what the cpntext is
15:43
<tobie>
Ms2ger: yup, in two weeks.
15:45
<gsnedders>
tobie: And I presume nobody will properly take over what you were doing? :(
15:48
<tobie>
gsnedders: well, unfortunately, no W3C member(s) stepped up to continue funding this effort further so far, so answer is "no" for the time being. :(
15:50
<Ms2ger>
Someone talk to dbaron? :)
15:56
<annevk>
The W3C should reshuffle its Staff a bit
15:56
<annevk>
plenty of people there...
16:03
<jgraham>
Dammit
16:03
<jgraham>
Now I want to post the example implementation report thing to webapps and public-test-infra
17:10
<MikeSmith>
this window.find() opening up a "find in page" control, I seem to remember reading about it in bugzilla or a mailing list recently but I can't remember where
17:39
<jgraham>
MikeSmith: webapps, I think?
17:45
<MikeSmith>
jgraham: ok
17:48
<MikeSmith>
this whole standards-discussion stuff is so time consuming
17:49
<MikeSmith>
I think we should all skip it and just unilaterally mint stuff in the browser engines that we all ship for iOS
17:51
<annevk>
zewt: but what if there's an IO error
17:51
<annevk>
zewt: if you're saying it should just stop reading at that point, I think we're in agreement
17:52
<annevk>
zewt: which is why I said we need an underlying model
17:52
MikeSmith
re-finds the "browser search api" thread
17:53
<zewt>
annevk: not sure what you mean
17:53
<zewt>
if there's a *real* IO error (user ejects the DVD holding the file the blob points to), yeah, there's an error
17:53
<annevk>
zewt: uhuh
17:53
<zewt>
but I don't think calling close() should cause any new IO errors to be called on fetches already in progress
17:53
<annevk>
zewt: that's what we were talking about
17:54
<annevk>
zewt: we were talking beyond close()
17:54
<zewt>
no, we're talking about close()
17:54
<annevk>
no
17:54
<jgraham>
maybe?
17:55
<zewt>
uh, the whole subthread is about "you can close() a Blob ... shouldn't affect the running operation"
17:55
<annevk>
yeah, but sicking specifically asked about IO errors to see if we can apply that to what should happen for close()
17:55
<annevk>
I said I had no idea what happened for IO errors (crash)
17:56
<zewt>
well I'm saying that close() shouldn't cause IO errors for stuff already in progress before you called close
17:57
<annevk>
-_-
17:57
<zewt>
...
17:57
<annevk>
I mean that much was already established early on
17:57
<annevk>
That's not really the interesting bit
17:58
<annevk>
Anyway, euc-kr
17:58
<zewt>
the only thing that was established early on is "this is underspecified"
17:58
<zewt>
(and also anyway, time to get back to my actual work...)
18:24
<Domenic_>
So <dialog> is still show()/close()?
18:25
<annevk>
Domenic_: yes
18:26
<Domenic_>
despite the web dev feedback that that was crazy-sauce?
18:30
<annevk>
Domenic_: feedback where? I remember you discussed it here with Hixie and he ended up reaching the conclusion that there was no good alternative
18:30
<Domenic_>
annevk: http://updates.html5rocks.com/2013/09/dialog-element-Modals-made-easy#disqus_thread
18:31
<jgraham>
Domenic_: There is an open bug on the issue
18:32
<jgraham>
Why would you assume it will be ignored?
18:32
<annevk>
Domenic_: see end of http://lists.w3.org/Archives/Public/public-whatwg-archive/2013Sep/0230.html
18:32
<Domenic_>
jgraham: I guess because Chrome keeps on chugging away at implementation.
18:32
<annevk>
jgraham: well it's filed against W3C HTML
18:32
<annevk>
jgraham: the one filed against WHATWG HTML is marked WFM
18:32
<annevk>
with a reference to that email I just mentioned
18:32
<jgraham>
annevk: Fair point. Sucks to have them fork the spec
18:33
<Domenic_>
jgraham: also because of the thread annevk just linked to
18:34
<jgraham>
Domenic_: That is a reasonable answer
18:34
<annevk>
Domenic_: that does not mean ignored though
18:34
<jgraham>
Anyway I think Hixie is wrong and it should be show/hide
18:35
<jgraham>
But this is way down the list of problems on the platform
18:35
<jgraham>
(well actually I think it should be open/close, but it seemed like there might be a more reasonable reason to avoid "open")
18:36
<annevk>
open="" and open() don't go together
18:42
<annevk>
Domenic_: as in, it seems to have been considered and no real viable alternative was proposed that was obviously better
19:03
<Hixie>
i definitely considered the open/close/show/hide thing. did a ton of research for it to figure out what to change it to. was surprised by the result, which was that the spec already matched most platforms.
21:02
<rdebeasi>
From the perspective of an author, many existing JS frameworks implement show/hide rather than show/close for similar behavior.
21:45
<Hixie>
rdebeasi: i actually looked for data to support that, but couldn't find any that isn't listed in that e-mail. if you do have data i missed, please do reply to that e-mail giving the new data :-)
21:45
<Hixie>
(though it might be too late by now)
21:48
<rdebeasi>
Sure, will do! It certainly couldn't hurt.
21:59
<annevk>
https://twitter.com/WTFMarketing/status/412616169850281984 :-)
22:44
<zcorpan>
now i have 100 tests on \u00E5 in the query of a url
22:45
<zcorpan>
someone: pls review
22:50
<zcorpan>
hmm, in 5 different encodings, so i guess 500 tests
22:55
<annevk>
zcorpan: should this really be in the HTML directory?
22:55
<zcorpan>
annevk: http://www.whatwg.org/specs/web-apps/current-work/multipage/urls.html#resolving-urls is in the html spec
22:56
<zcorpan>
annevk: i guess the css tests don't belong, but oh well
22:57
<annevk>
not a big fan of how Hixie deals with URLs
22:57
<annevk>
hmm
22:58
annevk
files https://www.w3.org/Bugs/Public/show_bug.cgi?id=24119
22:59
Hixie
isn't a big fan of it either
23:02
<annevk>
Sorry, no concrete suggestions yet
23:02
<Hixie>
did the parse error thing change?
23:02
<annevk>
Maybe in half a decade or so
23:02
<Hixie>
i seem to recall asking you about it when i wrote it
23:02
<annevk>
No, that's always been this way
23:03
<Hixie>
huh
23:03
<Hixie>
k
23:03
<Hixie>
btw you should say in the paragraph that dfns "URL parser" that it could return failure
23:04
<zcorpan>
annevk: btw some things don't use utf-8 that maybe could, like workers, EventSource
23:04
<zcorpan>
i also noticed that blink doesn't do XMLDocument#load
23:05
<annevk>
you mean new Worker?
23:05
<zcorpan>
right
23:05
<annevk>
workers themselves should already be utf-8-only
23:05
<Hixie>
(fixed, btw)
23:05
<annevk>
ta
23:05
<Hixie>
annevk: any idea on https://www.w3.org/Bugs/Public/show_bug.cgi?id=23892 ?
23:05
<annevk>
Hixie: I would not expect me to convince TC39 anytime soon
23:06
<annevk>
Hixie: the way they operate would make it largely my task to get the work done since nobody else is really keen on doing it
23:06
<annevk>
Hixie: and given other issues I can't say it's much of a priority
23:06
<zcorpan>
annevk: https://www.w3.org/Bugs/Public/show_bug.cgi?id=23822 https://www.w3.org/Bugs/Public/show_bug.cgi?id=23823
23:07
<annevk>
Whereas I think getting Map/Set support there would be good
23:08
<annevk>
zcorpan: the output of http://damowmow.com/playground/tests/urls/001.html is weird
23:08
<annevk>
zcorpan: in Gecko anyway
23:08
<annevk>
zcorpan: seems buggy for ws:
23:08
<Hixie>
annevk: should i just define it in HTML then?
23:08
<annevk>
Hixie: I guess for now :/
23:09
<Hixie>
k
23:09
<Hixie>
any idea what it should say?
23:09
<zcorpan>
annevk: ws.url in gecko just returns what was passed to teh constructor
23:09
<annevk>
ouch
23:10
<annevk>
Hixie: I think you create a new Map and then copy [[MapData]] or some such
23:10
<annevk>
Hixie: prolly similar for Set
23:10
<Hixie>
k
23:10
<annevk>
Hixie: [[SetData]]
23:10
<zcorpan>
annevk: the smile isn't encode-able in the document's encoding so gecko uses utf-8 while blink uses the <form> way and i guess IE would replace it with a ? and the spec requires %3F
23:11
<zcorpan>
iirc
23:11
<hober>
Hixie annevk: note https://bugs.webkit.org/show_bug.cgi?id=120654
23:12
<annevk>
omg no prefix used by olliej
23:12
<annevk>
(see #jslang for context)
23:12
<annevk>
(see also es-discuss for context)
23:13
<annevk>
zcorpan: hmm
23:13
<Hixie>
hober: thanks
23:16
<annevk>
zcorpan: yeah, need to look at that shit again at some point and see if we can always use the <form> way rather than the <form> way and ?
23:16
<annevk>
zcorpan: I reopened the bugs on the grounds that encoding override is for certain legacy APIs only, not even XMLHttpRequest uses it
23:17
<zcorpan>
annevk: ok
23:17
<Hixie>
boo, giving me more work
23:18
<Hixie>
doesn't this all mean that if you take a <Script src=""> and stick it into new Worker() you'll get a different URL?
23:18
<Hixie>
that seems bad
23:20
<annevk>
not using utf-8 is bad
23:20
<Hixie>
zcorpan: i fixed https://www.w3.org/Bugs/Public/show_bug.cgi?id=24088 (script error in worker); let me know if it's not compatible enough
23:20
<zcorpan>
you usually can't take an existing .js file and stick it into new Worker anyway
23:20
<annevk>
especially not if it's not utf-8
23:20
<Hixie>
(not that i can really see how someone would be depending on a worker failing to run in a particular way, but you never know)
23:21
<annevk>
I think the problem is that currently you need to override to use utf-8 with your setup
23:21
<annevk>
whereas that could have been the default
23:21
<annevk>
and therefore less work
23:22
<zcorpan>
Hixie: LGTM, thanks
23:24
<Hixie>
dglazkov: ping https://www.w3.org/Bugs/Public/show_bug.cgi?id=24106
23:25
<Hixie>
annevk: so i don't really understand why we'd want one place to not do this when literally dozens of other places in HTML do do it
23:26
<Hixie>
annevk: i understand that we don't like it, but being inconsistent seems far worse than having this sketchy behaviour
23:26
<annevk>
we are already inconsistent
23:26
<annevk>
and every new API should not do it
23:26
<Hixie>
why?
23:27
<annevk>
meaning there's a few exceptions in the end, such as <a> and location
23:27
<Hixie>
it's not a "few exceptions"
23:27
<Hixie>
there's literally dozens of them
23:27
<Hixie>
just in HTML
23:27
<annevk>
there's nothing outside of HTML that needs this; note that this is only about the query parameter
23:27
<annevk>
and only if you're not using utf-8 or utf-16
23:28
<annevk>
as I explained in an email somewhere it's already a very large footgun to not use utf-8
23:29
<annevk>
on top of all that the query parameter handling is a bug and inconsistent with how URLs are used elsewhere, and given that we're already inconsistent I don't see a reason to propagate it further
23:30
<annevk>
if all our APIs did this, mkay, but XMLHttpRequest does not, CSS does not, SVG does not
23:30
<annevk>
so making the query encoding override thing the exception is better, because it means we have less places in the long to test for this weirdness
23:30
<Hixie>
html. base. link. style. script. a href="" and ping="". cite="". img src and srcset. iframe. frame. embed. object. applet. video. track. source. src="" on <input type=image>, icon="". microdata's use of the aforementioned. window.open. location.*. pushState() and replaceState(). showModalDialog(). register*Handler(). external.AddSearchProvider(). drag and drop's use of some of the aforementioned.
23:31
<Hixie>
that's just the ones i could find in the HTML spec, i'm sure i missed some.
23:32
<Hixie>
video poster=""
23:32
<Hixie>
audio.
23:33
<zcorpan>
hmm i haven't tested window.open, location, pushState, replaceState
23:33
<Hixie>
WebSocket is a new protocol, so using UTF-8 there makes sense. CSS is a separate language, so using UTF-8 there makes sense (is that really implemented?). SVG does everything different anyway (and does it really do this differently?).
23:33
<Hixie>
and similarly inside Workers it makes sense to say the context is utf-8.
23:34
<Hixie>
but the others, i don't see why we'd be inconsistent.
23:34
<Hixie>
e.g. EventSource and Worker constructors.
23:34
<annevk>
I would expect those to work similar to XHR
23:34
<Hixie>
yeah, XHR being different makes no sense to me
23:35
<annevk>
also, many of the new contexts above could use utf-8
23:35
<Hixie>
i think inconsistency is far more of a problem than poor design decisions
23:35
<Hixie>
with a consistent quirky interface, you learn the quirk and then you laugh about it while working around it
23:36
<Hixie>
with an inconsistent interface, you spend all your time cursing that you can never guess what's going to happen
23:37
<annevk>
that's what happens when you don't use utf-8
23:37
<annevk>
not just for urls
23:37
<annevk>
i don't really see the big deal, non-utf-8 is already broken
23:38
<Hixie>
indeed, i don't really see the big deal either, non-utf-8 is already broken :-)
23:39
<Hixie>
so let's at least keep it consistent :-)
23:40
<annevk>
it's already not consistent and given we'll forever add more APIs that take URLs, it seems better to keep it contained
23:40
<annevk>
about consistent quirks...
23:40
<annevk>
<script id=x>w(document.querySelector("#X") !== document.getElementById("X"))</script>
23:40
<zcorpan>
blink uses document's encoding for XHR. gecko uses utf-8 for EventSource
23:41
<Hixie>
annevk: "contained" = "inconsistent". it's better to keep it uniform than have a small subset of things that are different.
23:41
<Hixie>
annevk: especially when the ones that are different are the most used ones, like <a href>.
23:42
<Hixie>
not sure what the relevance of that id=x example is; the presence of things in the web platform that are inconsistent isn't an argument for adding more.
23:44
<annevk>
Hixie: I guess if you think we can still get it uniform I might be up for trying that
23:45
<annevk>
Hixie: would be interesting to hear what others have to say about it
23:47
<zcorpan>
seems like it should be possible
23:50
<annevk>
though if we decide that e.g. CSS should be sane, you get these silly crossover APIs... oh well
23:50
<annevk>
it'll be messy either way
23:57
<zcorpan>
if we want consistent we can make css use the style sheet's encoding