02:47
<roc>
anyone have any idea why Safari doesn't supported IndexedDB?
04:25
<MikeSmith>
the definition of "absolute URL" at http://url.spec.whatwg.org/#concept-absolute-url disallows absolute URLs from having a fragment
06:48
<MikeSmith>
http://www.scribd.com/doc/213628421/Exhibit-1871 is pretty interesting
06:49
<MikeSmith>
(by way of roc's blog http://robert.ocallahan.org/2014/03/mozilla-and-silicon-valley-cartel.html)
07:01
<a-ja>
MikeSmith: you do work on validator.nu ?
07:01
<MikeSmith>
a-ja: yup
07:02
<a-ja>
found an inconsistancy between it and w3c's
07:02
<a-ja>
header content on summary element not allowed yet
07:04
<a-ja>
whatwg spec allows it now....fairly recent change (and html5.1 ??? iirc)
07:14
<MikeSmith>
a-ja: I work on the shared backend for validator.nu and the W3C Nu validator but not on the actual http://validator.nu/ or http://html5.validator.nu/ services
07:16
<MikeSmith>
I push changes to http://validator.w3.org/nu/ at least once a week or so but these days http://validator.nu/ and http://html5.validator.nu/ don't get updated with the changes quite as often
07:23
<a-ja>
MikeSmith: related question...if you recall, we chatted a bit about summary role=button and details role=group failing. so removed both, and validator then complained about missing roles (but didn't say which roles they should be). help me out?! :)
07:37
<MikeSmith>
a-ja: if you give me an document or URL I can check
07:38
<a-ja>
MikeSmith: <details id="toc-details" role="group" aria-labelledby="toc-summary" open="open"><summary id="toc-summary" role="button" aria-controls="toc-details" aria-expanded="true" tabindex="0">
07:38
<a-ja>
MikeSmith: vs just: <details id="toc-details" aria-labelledby="toc-summary" open="open"><summary id="toc-summary" aria-controls="toc-details" aria-expanded="true" tabindex="0">
07:39
<MikeSmith>
yeah
07:40
<MikeSmith>
a-ja: so I guess if you use other aria-* attributes on those elements, it's going to say you need the role attribute
07:40
<a-ja>
MikeSmith: perhaps so
07:40
<MikeSmith>
because those particular aria- attributes are only valid for certain roles
07:41
<MikeSmith>
I can't easily fix the error message there
07:41
<MikeSmith>
I wish I could but the fact is the whole aria stuff is so completely baroque and overengineered that we're lucky to have any validator support for it at all
07:42
<MikeSmith>
what's there now has easily cost me more time to work on than any other part of the validator
07:43
<MikeSmith>
and I didn't even do most of the development of the aria support in the validator. hsivonen did. I just worked on top of what he had developed, and it still cost me days and days of time
07:44
<MikeSmith>
so I guess I'll just add role=group to details
07:44
<MikeSmith>
do I need to add role=button to summary also?
07:44
MikeSmith
checks the spec
07:45
<a-ja1>
not sure either of those match spec right now
07:46
<MikeSmith>
role=summary is a role that supports aria-expanded, so it matches the spec close enough
07:46
<a-ja1>
think it probably just got overlooked at the time
07:46
<a-ja1>
yep
07:47
<MikeSmith>
but
07:47
<MikeSmith>
role is not allowed for summary at all
07:47
<MikeSmith>
explicitly
07:47
<MikeSmith>
http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#concept-role-none
07:48
<MikeSmith>
"The entry "no role", when used as a strong native semantic, means that no role other than presentation can be used."
07:48
<MikeSmith>
if you think that's misguided then you need to file a spec bug
07:48
<MikeSmith>
I can't have the validator violate that constraint
07:49
<MikeSmith>
w3c spec says the same thing:
07:49
<MikeSmith>
http://www.w3.org/html/wg/drafts/html/master/dom.html#concept-role-none
07:55
<a-ja1>
yeah...maybe the answer is "do it in script" rather than in the html
07:55
<MikeSmith>
yeah that's a general solution for getting around aria validation problems
07:55
<a-ja1>
polyfills forever! :\
07:56
<MikeSmith>
the other solution is to just ignore the errors
07:56
<a-ja1>
mm hmm
07:56
<MikeSmith>
or filter them out at http://validator.w3.org/nu/ with the message filtering option
08:04
<a-ja1>
i have sympathy for the "fix ARIA, the a11y API's, and the screen readers, and stop messing with clean HTML" argument
08:06
<MikeSmith>
yeah
11:17
<mathiasbynens>
what kinds of security vulnerabilities does `showModalDialog`’s “pause JS execution” behavior cause?
11:26
<annevk>
mathiasbynens: its nested event loop behavior is the only one in the platform, which makes for some tricky coding elsewhere, which might lead to exploits
11:35
<mathiasbynens>
annevk: gotcha, thanks
12:17
<annevk>
"I'm not why why Anne thinks an email from February reflects the current thinking of Mozillians on this topic."
12:18
<annevk>
Well, if you're gonna ask one of us in private what the status quo is and that person is not even implementing, you're gonna have a bad time...
12:19
<annevk>
"but recently selectors were renamed again" :-)
13:16
<MikeSmith>
cool, google came up with a way we don't need cookies any more https://code.google.com/p/chromium/issues/detail?id=355541
13:38
<SteveF>
a-ja1: role=button now allowed on <summary> thanks for the heads up it was something I had meant to fix previously https://www.w3.org/Bugs/Public/show_bug.cgi?id=25135
13:39
<SteveF>
a-ja1: have spoken with mike smith, he will update http://validator.w3.org/nu/
14:02
<Jasper>
I remember a little while ago there were meeting notes saying that :before / :after were mistakes.
14:02
<Ms2ger>
As opposed to ::before?
14:03
<Jasper>
erm, I meant ::before / ::after.
14:03
<Jasper>
I believe they were for Adobe's proposal for having multiple ::before / ::after psuedoelements?
14:04
<astearns>
Jasper: that opinion did come up in that discussion, yes
14:05
<Jasper>
It seems like multiple ::before / ::after elements will probably never happen for those reasons.
14:06
<Jasper>
Would anybody be able to find the meeting notes for me?
14:06
<astearns>
Jasper: the current multiple before/after proposal was dead on arrival
14:07
<Jasper>
Hm
14:07
<astearns>
but more ways of creating boxes in CSS keeps coming up as a need
14:07
<astearns>
so something in that area may eventually happen
14:08
<Jasper>
I remember the issue being mostly related to accessibility, scriptability and poor devtools support.
14:09
<astearns>
IMO, the main issue is scriptability - pseudo-elements will be inadequate until/unless we have better script support for them
14:11
<Jasper>
Aha, found it! http://lists.w3.org/Archives/Public/www-style/2012Aug/0771.html
15:05
<annevk>
https://twitter.com/thijs/status/448064463498133504 o_O
15:07
<annevk>
Oh I'm late to the party
15:16
<ondras>
but it is amusing anyway.
15:20
<annevk>
WHATWG list is pretty close to a <canvas> list these days
15:24
<jcgregorio>
canvas \o/ :-)
15:37
<dglazkov>
good morning, Whatwg!
15:44
<Hixie>
MikeSmith: btw, changing things in scripts silences the validator, but doesn't make them any more valid
15:45
<Hixie>
(why would <summary> get role=button? it's not interactive. am i missing something?)
15:46
<wilhelm>
Toggling its parent is interactive, presumably.
15:47
<Hixie>
well the interactive part of the <details> should be interactive, sure, but it's not in the DOM, so...
15:48
gsnedders
sighs
15:48
<gsnedders>
I feel like I could've saved myself months by showing everything I managed about the tokenizer through a proof by exhaustion...
15:49
<Hixie>
heh
15:49
<Ms2ger>
gsnedders, so, what have you learned? :)
15:49
<Ms2ger>
Why the AAA works?
15:50
<gsnedders>
hah! like I managed to model the tree constructor.
15:50
gsnedders
cries
15:50
<jgraham>
So you are saying that your decision not to use proof by exhaustion led to exhaustion?
15:50
<gsnedders>
Yes, yes entirely.
15:51
<gsnedders>
Does anyone want to try and prove that the tree constructor cannot be implemented by a pushdown automaton?
15:52
<Ms2ger>
No
15:52
<jgraham>
gsnedders: You?
15:52
<Ms2ger>
I would love to see a proof that it can be
15:52
<gsnedders>
I don't have time to before the deadline, sadly.
16:03
<gsnedders>
Ms2ger: I simultaneously suspect it is and that it is not. :(
16:06
<Ms2ger>
gsnedders, so your belief is in a state of superposition?
16:06
<gsnedders>
Actually, does the parsing of <p><table><p> not show that HTML cannot be context-free?
16:14
<SteveF>
hixie: implementations appear to disagree with you e.g. in blink summary is focusable control
16:15
<Hixie>
yeah, that's a known bug in blink
16:15
<SteveF>
bug number?
16:16
<Hixie>
no idea
16:30
<annevk>
Hixie: if an exception is thrown from a compound microtask subtask, do you handle that? or should I be catching exceptions?
16:31
<annevk>
gsnedders: what does context-free mean?
16:31
<Hixie>
interesting question
16:32
<Hixie>
a subtask is just a set of steps
16:32
<Hixie>
so i guess it depends on how you're invoking the script
16:32
<annevk>
I was looking at https://www.w3.org/Bugs/Public/show_bug.cgi?id=17713#c16 and wondering if what bz said under 11) is still the case
16:32
<annevk>
I just say "invoke callback with x and y"
16:33
<Hixie>
how does webidl define "invoke callback"?
16:34
<annevk>
Hixie: http://heycam.github.io/webidl/#es-invoking-callback-functions it just propagates
16:34
<annevk>
Hixie: unless it's part of a promise returning thing
16:38
<annevk>
Hixie: is http://www.whatwg.org/specs/web-apps/current-work/#report-the-error what I should be invoking?
16:39
<Hixie>
we should just have a single function that invokes a callback, catches exceptions and reports them
16:39
<Hixie>
but yes, i think so
16:39
<annevk>
hmm yeah, if you could provide that shorthand...
16:40
<annevk>
Hixie: https://www.w3.org/Bugs/Public/show_bug.cgi?id=25138
16:47
<Hixie>
annevk: will do
16:48
<Hixie>
(i'm mostly afk this week btw, won't be getting much done; sorry)
16:48
<Hixie>
annevk: should i wait for that bug to be assigned to me once heycam has seen it?
16:49
<annevk>
Hixie: either way I guess, I'd still like a <dfn> in IDL
16:49
<annevk>
Hixie: I could file a separate one on you
16:50
<Hixie>
either way. i commented on that idl bug.
16:52
<TabAtkins>
annevk: https://en.wikipedia.org/wiki/Context-free_grammar
16:52
<TabAtkins>
annevk: A grammar that doesn't take the surrounding context into account when expanding things.
16:52
<annevk>
Hixie: ta
17:49
<Hixie>
i can't believe how inefficient mediawiki and wordpress are
17:49
<Hixie>
i host literally dozens of sites on this one machine, including multiple long-lived websocket servers and so on, and all the cpu and ram goes to the php scripts running wiki.whatwg.org and blog.whatwg.org.
17:50
<Ms2ger>
Blame php :)
17:51
<annevk>
Hixie: oooh, so me hammering svn.whatwg.org is fine? :p
17:51
<annevk>
Hixie: I tried to get foolip to rewrite the setup, he put on his list of things to do before he dies
17:58
<foolip>
Hixie: is there any particular reason you haven't switched to git, other than not having a particular reason to do it?
17:59
<Ms2ger>
Eww, git
17:59
<Hixie>
not having any reason to do it is the main reason i don't do most things :-)
17:59
<Ms2ger>
Why would you ever use git if you weren't forced to?
17:59
<foolip>
:)
18:00
<Hixie>
annevk: svn stuff seems to all fall under apache, so i've no way to tell if it's svn or just regular web traffic
18:00
<jgraham>
I mainly use it to wind up Ms2ger
18:00
<Ms2ger>
:D
18:00
<Hixie>
annevk: (apache usage is highish, but i expect that on a machine with 60+ domains hosted)
18:00
<foolip>
Ms2ger: I use Git wherever possible, like if I need to diff two files I add them to git to get the pretty colors :)
18:00
<Hixie>
(it's nowhere near as high as i'd expect, really)
18:01
<Hixie>
foolip: i use emacs to get pretty colours
18:01
<foolip>
Hixie: I suspected you used Emacs from the way the source is wrapped :)
18:02
<tantek>
Hixie, is it PHP or is it MySQL?
18:02
<Ablu>
foolip: you do not need to add files to diff them
18:02
<Hixie>
foolip: the first line wasn't a giveaway? (-*- mode: Text; fill-column: 100 -*-)
18:02
<Ablu>
git diff <file1> <file2> works
18:02
<Hixie>
tantek: php
18:02
<Ablu>
even if they are not in git
18:02
<Hixie>
tantek: the mysql stuff is on a different box entirely
18:02
<foolip>
Hixie: well, and the comment about emacs lines in the WebVTT spec
18:02
<Hixie>
heh
18:03
<Hixie>
i have "f6" in emacs bound to "open the html spec" :-)
18:03
<foolip>
Ablu: cool, I didn't know that :)
18:03
<foolip>
but I'll probably keep using Git where not needed to spite Ms2ger
18:04
<Ms2ger>
foolip, that's probably a better reason than most to use git
18:04
<foolip>
Ms2ger: dare I ask what you would rather use?
18:04
<Ms2ger>
Take a guess
18:04
<Hixie>
git seems reasonable to use if you're working on a colaborative distributed project, like linux
18:05
<foolip>
hg
18:05
<Hixie>
if you're one person editing one file, git seems kinda pointless.
18:05
<Ms2ger>
Correct
18:05
<jgraham>
Hixie: It seems significantly better than svn for either case
18:05
<Hixie>
how is it better than svn for solo use?
18:05
<Ms2ger>
Hixie, actually, I found one person editing to be the only reasonable way to use git :)
18:06
<foolip>
Hixie: I guess, unless you're already used to Git for all other projects
18:06
<Hixie>
well sure, if you're already used to it
18:06
<Hixie>
in my case, when i had to pick a version control system, all i had experienced so far was cvs
18:06
<Ablu>
why not to use git: http://vimeo.com/74377782
18:06
<Hixie>
and git didn't exist
18:07
<TabAtkins>
I definitely vastly prefer git over svn for personal projects as well.
18:07
Ms2ger
gets off Hixie's lawn
18:07
<jgraham>
Ah, hg, the vcs that's so broken it needs a whole different system to allow you to work locally. A system that itself is so broken it requires a whole second repository to make it not astonishingly likely to lose your work
18:07
<TabAtkins>
The distributed-ness of it is nice for multi-user projects, sure, but most of the benefits are in personal use.
18:08
<Hixie>
i'm open to being convinced, but i'm failing to see anything that i need from a version control system that i don't already have :-)
18:08
<Hixie>
mostly i want commit, and blame.
18:08
<Ms2ger>
jgraham, hey, you were already winding me up :)
18:08
<Hixie>
and diff
18:08
<Hixie>
and monotonically increasing version numbers
18:08
<Hixie>
svn gives me all that
18:08
<foolip>
Hixie: I asked mostly to find out if it was because of dependencies like web-apps-tracker, in which case I would have prioritized fixing that, but if you're happy the way things are then all is well
18:08
<Ms2ger>
git gives... blame, I guess
18:08
<Hixie>
oh and also the ability to quickly graft different parts of a repo into my directory tree
18:08
<Hixie>
svn supports that too, luckily
18:09
<Hixie>
foolip: oh, no. my build system already supports committing to multiple repos anyway, so it wouldn't be an issue for me to change the canonical one.
18:10
<Hixie>
foolip: without breaking back compat.
18:10
<foolip>
oh, ok
18:11
<foolip>
well, the mirror works for me, except when annevk makes commits and it broke :)
18:11
<foolip>
(fixed now)
18:11
<Hixie>
anne committed to the mirror? o_O
18:12
<foolip>
no, r8556 and r8557 in your repo was some test commits
18:12
<foolip>
my script broke because it only had an entry in AUTHORS for you
18:13
<Hixie>
oh hey
18:13
<Hixie>
i didn't even know anne had access to do that!
18:14
<foolip>
well, he broke your 8 year streak of commits :)
18:15
<Hixie>
hey look at that, hsivonen has access too
18:15
<Hixie>
i wonder when i gave them access :-)
18:16
<SamB>
hmm, the spoiler tag in http://codegolf.stackexchange.com/questions/24401/so-obviously-p-np acts a bit funny WRT the hyperlink :-)
18:17
<Hixie>
looks like it was just two empty commits
18:17
<foolip>
a test.txt was added and removed. no commit message though
18:18
<Hixie>
weird
18:18
<Hixie>
says something about dreamhost tech support
18:18
<Hixie>
annevk: any idea what the story is with the test.txt file?
18:21
<TabAtkins>
SamB: <blockquote class="spoiler"> does a 'color' transition to/from transparent over .5s. But hyperlinks already have a transition rule for 'color' going over .2s, intended for a more subtle color shift when hovering. So the blockquote's 'color' rule overrides, but the a's transition wins, and you get the weird extra-fast transition.
18:22
<SamB>
oh, bonus, an explanation
18:22
<Hixie>
foolip: (did your script get invoked?)
18:25
<TabAtkins>
SamB: That's why you should do silly tricks like a color transition to 'transparent'. Use 'opacity' on the block element, like God intended, even if it means you need a wrapper element inside the spoiler element.
18:26
<TabAtkins>
s/you should/you shouldn't/
18:26
<foolip>
Hixie: not by your ping, the last one was 2014-03-19T22:02:24.506Z
18:27
<foolip>
but I also have it as a daily cron job in case of network failures or whatever
18:40
<SamB>
TabAtkins: sigh, wrapper elements :-(
18:42
SamB
didn't even style this site, actually
18:43
<annevk-cloud>
Hixie: DH support
18:56
<Hixie>
foolip: ah, ok
18:57
<Hixie>
annevk-cloud: ?
19:08
<gsnedders>
annevk-cloud: Importantly, pushdown automata and context-free grammars are equivilient.
19:09
<gsnedders>
annevk-cloud: So if you can show no context-free grammar can express the language, it is implied there is no pushdown automaton
19:11
<SamB>
hmm, U+231B HOURGLASS is nice
19:14
<annevk>
Hixie: I contacted DreamHost support about svn timing out, I assume they ran a test...
19:14
<Hixie>
ah
19:14
<annevk>
Hixie: not timing out, not having enough space allocated to calculate a diff or some such
19:14
<Hixie>
ah, yeah, i'd already told them /tmp was full
19:14
<annevk>
Hixie: k
19:19
<SamB>
what do you have: a vhost, or a full VM?
19:22
<annevk>
Hixie: seems weird that I have commit access btw, you might want to revoke that
19:22
<annevk>
Hixie: I have no idea how DH got hold of say, the credentials...
19:22
<Hixie>
yeah...
19:23
<Hixie>
k, i revoked your account. let me know if you find a reason you had it, so i can add it back.
19:36
<annevk>
cool
21:28
<SamB>
what was the original motivation for this behaviour: http://updates.html5rocks.com/2012/09/Stacking-Changes-Coming-to-position-fixed-elements
21:35
SamB
mumbles about no decent anchors for e.g. text/plain ...
21:54
<TabAtkins>
SamB: Motivation is that fixpos is crazytimes, and allowing things in the page to sandwich between descendants of a fixpos is even more crazytimes.
21:54
<TabAtkins>
Particularly on mobile, where (a) fixpos is *extra* crazy, due to the virtual viewport, and (b) gpu resources are smaller.
21:54
<SamB>
ah, yeah, virtual viewport, right
21:55
<TabAtkins>
(b) is important, because we put fixpos things on their own gpu layer so we can scroll decently, and if things can sandwich between, we have to split the fixpos into 2 or more layers.
21:55
<SamB>
I was kind of expecting an efficiency-related answer
21:55
<TabAtkins>
Making it always a stackign context lets us always use a single layer.
21:55
<SamB>
like "saves rendering resources"
21:55
<TabAtkins>
Yup, that's basically it.
21:55
<SamB>
wasn't quite sure if that was going to be time or space though
21:56
<SamB>
same reason I don't like those background images that don't scroll
21:56
<SamB>
I want to kill them dead
21:57
<SamB>
they literally tear
22:04
<Hixie>
isn't fix pos always a stacking context per CSS2.1:E ?
22:06
<TabAtkins>
Hixie: Nope, not if z-index:auto
22:06
<Hixie>
huh
22:07
<Hixie>
well we should fix that
23:07
<cabanier>
TabAtkins: stacking context does not always create a gpu layer
23:27
<TabAtkins>
cabanier: No, but fixposes do.
23:28
<TabAtkins>
Hixie: It is fixed, now.