06:54
<cortexA9>
hi
07:57
<zcorpan>
jgraham: does trickle(d1) send the headers immediately?
08:16
<Ms2ger`>
So it was true that FB was going to do Presto!
08:16
<Ms2ger`>
https://news.ycombinator.com/item?id=6684318
08:58
<hsivonen>
oh, cool. utf-16le is now the name of the encoding and utf-16 is a label
08:58
<hsivonen>
this has changed recently, right?
08:58
<hsivonen>
anyway, I'm less annoyed now
09:32
<annevk-cloud>
Yeah was changed
11:00
<annevk>
So GPHemsley, should http://wiki.whatwg.org/wiki/New_work become a subpage too then?
11:02
<annevk>
GPHemsley: your spelling of acknowledgments disagrees with all my specs
11:04
<MikeSmith>
fwiw about setting up a way to get RF agreements for whatwg specs, I notice that the Dash Industry Forum is using a relatively lightweight means to get RF agreemenents for code contributions to the dash.js library. see https://github.com/Dash-Industry-Forum/dash.js/wiki/How-to-Contribute#process-for-contributing-code and, e.g., https://groups.google.com/forum/#!topic/dashjs/W4Tmm8sBGX0
11:05
<MikeSmith>
it's for code, not specs, but it seems like the same thing could be used for spec contributions
11:06
<MikeSmith>
in fact, they actually just call it a "Feedback Agreement"
11:06
<MikeSmith>
http://dashif.org/documents/DASH-IF-Feedback-Agreement-5-9-2013.pdf
11:06
<MikeSmith>
"By signing below, you (on behalf of yourself if you are an individual and your company if you are providing Feedback on behalf of the company) grant the companies under all applicable intellectual property rights owned or controlled by you or your company a non-exclusive, non-transferable, worldwide, perpetual, irrevocable, royalty-free license to use, disclose, copy, publish, license, modify, sublicense or otherwise distribute and exploit Feedback yo
11:46
<MikeSmith>
I'm trying to remember if there's data somewhere on how many documents on the Web are actually being served with an XML MIME type
11:47
<MikeSmith>
I mean documents with an XHTML doctype served with an XML MIME type instead of as text/html
12:15
MikeSmith
finds http://dev.opera.com/articles/view/mama-http-headers/#conttype
12:16
<MikeSmith>
'Of the 3,509,180 URLs that MAMA analyzed, the vast majority (~99.9%) used a "text/html" MIME type'
13:21
<annevk>
MikeSmith: I hope you're just doing that because you're curious and not because someone is saying something irrational
13:22
<MikeSmith>
umm, I wish I could claim it was because someone isn't saying something irrational, so let's just go with "curious"
13:28
<annevk>
You know it's bad when you have to revisit a decade old fact-settled debate.
13:31
<hsivonen>
annevk: speaking of those: http://lists.suckless.org/dev/1310/17874.html
13:33
<annevk>
hsivonen: gopher man, how do you find this stuff?
13:33
<hsivonen>
annevk: hendry tweeted it at me
13:33
<annevk>
haha
13:35
<hsivonen>
I wonder if I should add telemetry for those font-only legacy Mac encodings
13:36
<annevk>
To see if the font code path is being hit?
13:36
<hsivonen>
annevk: right
13:36
<hsivonen>
maybe I should add telemetry for legacy stuff that Thunderbird might use, too
13:37
<annevk>
Is there any Mac OS out there that still needs that?
13:37
<annevk>
That Gecko supports?
13:37
<hsivonen>
annevk: I have no idea what fonts people have installed
13:37
<annevk>
Ooh, it's not an OS thing, it's a fonts thing?
13:37
<annevk>
Ew...
13:38
<hsivonen>
annevk: it's for making sense of data in legacy font files
13:39
<annevk>
I guess that does warrant investigation of sorts. Alternatively investigation of the Chrome code base which almost certainly does not do this. (They don't have those encoders/decoders at the Chromium level.)
14:24
<GPHemsley>
annevk: Yeah, there are a few more pages that should probably become subpages.
14:43
<GPHemsley>
Any chance we could start encouraging subject prefixes on the mailing list, like CSS and others do? (I know I do it with mimesniff.)
14:45
<zewt>
how about "Subject: "
14:46
<annevk>
GPHemsley: no
14:46
<GPHemsley>
why?
14:46
<annevk>
A lot of topics encompass lots of different specifications. We should not add bureaucracy that isn't needed.
14:47
<GPHemsley>
You have a low bar for what is considered "bureaucracy"
14:47
<annevk>
Yes
14:48
<annevk>
Everything that's a "rule" basically.
14:48
<annevk>
Don't need people that want to contribute to have to go through more rules than strictly needed.
14:49
<annevk>
It's already too hard for a bunch of them.
14:49
<GPHemsley>
On the flipside, it's hard for me to figure out what mail to read ;)
14:50
<Ms2ger>
All of it ;)
14:50
<GPHemsley>
Ms2ger: Do you really what my unqualified opinion on everything? ;)
14:50
<GPHemsley>
s/what/want/
14:51
<Ms2ger>
I want you to read it, not necessarily to reply ;)
14:51
<darobin_>
GPHemsley: I think the hidden message there is that annevk doesn't want to contribute :)
14:51
<jgraham>
Not sure that reading every mail means responding to everything
14:51
<GPHemsley>
jgraham: 'Twas hyperbole ;)
14:51
jgraham
doesn't read every mail, but also doesn't want more rules for the sake of it
14:52
<darobin>
I find that if people use meaningful subjects you don't need a prefix
14:53
<annevk>
GPHemsley: I recommend reading the subjects that interest you
14:53
<annevk>
darobin: indeed
14:53
<darobin>
whereas people tend to get prefixes wrong, especially if there are more than two different ones
14:53
<annevk>
I'd be okay with a "use meaningful subjects" guideline, just like we have a "inline quote" guideline
14:53
<darobin>
that just falls under proper netiquette really
14:53
<annevk>
Yeah
14:54
<darobin>
maybe just training the spam filter to drop emails with poor netiquette would achieve excellent signal to noise
15:11
<SimonSapin>
This is nice Frenglish: "Initialized empty Dépôt git dans /home/simon/projects/specs/html/.git/"
15:12
<Ms2ger>
Why would you use a French OS
15:13
<SimonSapin>
old habits
17:07
<zcorpan>
anyone know about in-browser ITS tools?
17:08
<miketaylr>
ITS?
17:11
zcorpan
-> beijing -> shenzhen
17:13
<Hixie_>
wtf, a bunch of characters in the multipage acks got corrupted
17:14
<Hixie_>
oh i bet the problem is that some parser doesn't supported named references
17:14
<Hixie_>
annnnnnneeeeeeeeeeeeee
17:28
<marcosc>
heh, I was also going to curse Anne for something different :)
17:33
<jorendorff>
Domenic_: are microtasks going to be spec'd in ES6, and is there already draft text for it?
17:36
<Hixie_>
oh, if ES6 is gonna spec them that would make my life easier
17:36
Hixie_
is currently deep in the middle of trying to figure out how to spec them in HTML
17:41
<jsbell>
jorendorff/Hixie: relevant (short) thread is https://mail.mozilla.org/pipermail/es-discuss/2013-October/033940.html
17:42
<jorendorff>
"easier"; "short"
17:42
<jorendorff>
OK, my real question was: how deterministic is the ordering of microtasks?
17:43
<Hixie_>
well i presume they're just going to provide hooks for HTML to use
17:43
<Hixie_>
it's not like they can actually embed the event loop directly into ES6, given how much HTML-specific crap is around the event loop
17:44
<jorendorff>
mmhmm
17:44
<Hixie_>
(e.g. when the event loop starts, etc)
17:44
<Hixie_>
(how it's used for non-script-related stuff)
17:55
<Hixie_>
today's monster is http://www.hixie.ch/tests/adhoc/html/script/callbacks/001.html
18:00
<Hixie_>
anyone got IE11 around?
18:03
<dglazkov>
good morning, Whatwg!
18:36
<annevk-cloud>
I can maybe fix bugs from China, otherwise in 10 days.
18:37
<annevk-cloud>
On the tube at the moment with no place to sit. Shit's clogged.
18:39
<annevk-cloud>
ES6 will just have a hook btw
18:40
<Hixie_>
if you're in an iframe and you call parent.setTimeout(parent.open, 0, 'a') -- should the 'a' be resolved relative to your own base URL, or should it be resolved relative to the parent's base URL?
18:41
<Hixie_>
everyone (except IE, which won't accept that construct at all) agrees it should be the parent's base URL, but nobody agrees as to why
18:41
<Hixie_>
in chrome it's because the setTimeout is from the parent
18:42
<Hixie_>
in firefox it's because the argument isn't a script but the argument comes from the parent
18:42
<Hixie_>
and in safari it's because the argument isn't a script but the setTimeout comes from the parent
18:43
<Hixie_>
firefox's answer is inconvenient since it would mean figuring out what global to use when you pass e.g. MessagePort.send to setTimeout
18:43
<Hixie_>
chrome's answer is lame because it means if you do parent.setTimeout(myownfunc, ...), myownfunc is going to resolve relative to the parent instead of relative to itself
18:44
<Hixie_>
seems crazy that you could change how a function works based on who's setTimeout you use
18:44
<Hixie_>
then again, it's already true that if you call a function in the parent frame directly, you change how it behaves relative to it being called by script in that frame...
18:45
<Ms2ger>
How about window.setTimeout.call(parent,...? :)
18:45
<Hixie_>
not sure i want to know
18:46
<Hixie_>
i assume that's the same as parent.setTimeout, at the webidl level
18:46
<Hixie_>
am i wrong?
18:46
<Ms2ger>
I'd hope so
18:47
<Hixie_>
let's assume so for the sake of this discussion
18:49
<Hixie_>
wait, safari's behaviour still requires answering the question above. damnit.
19:42
<Domenic_>
Hixie_: jorendorff: annevk-cloud: yeah my impression is there will be some abstract operation "queue a microtask" which ES uses, and doesn't actually specify how it works, besides maybe some invariants like clean-stack or ordering or something.
20:00
<jorendorff>
clean-stack, definitely; ordering is what i'm after... hmm.
20:00
<jorendorff>
i'll assume it's mostly deterministics
20:16
<annevk-cloud>
It will be deterministic jorendorff
20:17
<jorendorff>
ok!
20:17
<jorendorff>
Domenic_: i was able to remove all direct interaction with microtasks from the module loader stuff by using promises instead
20:18
<jorendorff>
Domenic_: little worried i'm not doing it right -- that spec is impenetrable, to me -- but i guess the bugs will shake out one stuff runs
20:18
<jorendorff>
*once
22:08
<Hixie_>
heycam!
22:08
<Hixie_>
so i've been studying browsers
22:08
<Hixie_>
http://www.hixie.ch/tests/adhoc/html/script/callbacks/001.html has my results and conclusions (see bottom)
22:09
<heycam>
oh my
22:09
<Hixie_>
yeah.
22:10
<Hixie_>
so the only real question that no browsers agrees with is how to determine the entry script (actually, the "settings object", i'll explain what i mean in the bug later) for cases where you call a method with a native method as the callback
22:10
<Hixie_>
as in setTimeout(open, ...)
22:11
<Hixie_>
there are several options:
22:11
<Hixie_>
1. we don't support that (IE9)
22:11
<Hixie_>
2. we use the data from the global of the method you called (setTimeout in this case) (Chrome, Safari)
22:11
<Hixie_>
3. we use the data from the global of the method you passed as a callback (open in this case) (Firefox)
22:12
<Hixie_>
4. we do our own new thing in the interests of sanity
22:12
<heycam>
to me, 3 sounds like it is more sensible
22:12
<Hixie_>
(btw Chrome and Safari are incompatible for the more common case of passing a script-defined function, so don't put too much in the fact that they agree on this question)
22:12
<heycam>
i.e. you look at the thing you will be invoking later
22:12
<Hixie_>
#2 and #3 have a problem: do we have a way to determine what "global" any random method is related to?
22:13
<heycam>
yes
22:13
<Hixie_>
oh excellent
22:13
<heycam>
Web IDL says every "initial object" is associated with a particular "global environment"
22:13
<Hixie_>
initial object is like the method?
22:13
<Hixie_>
what's a "global environment"?
22:13
<heycam>
yeah, it's basically all of the objects that exist before you run any scripts
22:13
<heycam>
(apart from the ES-defined ones)
22:13
<heycam>
"global environment" is something a bit fuzzy
22:13
<Hixie_>
so what if someone passes an ES-defined one?
22:14
<Hixie_>
like, toString
22:14
<heycam>
I guess my definition won't help there… maybe there is one in the ES spec?
22:14
<heycam>
not sure if the spec has begun to talk about realms yet or not
22:14
<Hixie_>
last i checked, the ES spec didn't acknowledge multiple globals
22:14
<Hixie_>
but i may not have checked for a while
22:15
<heycam>
http://people.mozilla.org/~jorendorff/es6-draft.html#sec-code-realms
22:15
<heycam>
ok so you can go from the toString object to the realm
22:15
<Hixie_>
what does the realm give you?
22:15
<heycam>
that table of things
22:15
<heycam>
including the global object
22:15
<Hixie_>
a global, we can use that
22:16
<Hixie_>
well if you can get from any random method (script-defined, ES-defined, or WebIDL-defined) to a global, i can do the rest
22:17
<heycam>
I can
22:17
<Hixie_>
ok, cool
22:17
<heycam>
whew! finally we can use style=""
22:18
<TabAtkins>
heycam: Hm?
22:19
<Hixie_>
hmmm... i need to move script origins to the settings object also
22:19
<heycam>
TabAtkins, sorry, just being facetious about http://www.w3.org/TR/2013/REC-css-style-attr-20131107/ becoming a Rec
22:19
<TabAtkins>
Ah, indeed.
22:29
<Ms2ger>
Time to deprecate style=""
22:39
<Hixie_>
i tried
22:39
<Hixie_>
people presented... compelling arguments
22:39
<Jasper>
i'm sure they did
22:54
<Hixie_>
TabAtkins: dunno if you're the right one to ask, but, are you on board with javascript: being killed in CSS url()s?
22:54
<TabAtkins>
I'm surprised those even work. Do they?
22:54
<TabAtkins>
Kill them.
22:54
<Hixie_>
probably not
22:54
<Hixie_>
ok
22:55
<Hixie_>
the plan is to move javascript: to just be a quirk of the navigate algorithm
22:55
<Hixie_>
i'm about to make the first move towards that, by removing the concept of scripts having origins, the bulk of which is concerned with defining the origin of javascript: scripts
22:56
<Hixie_>
christ, 1993 matches for "script" in the spec
22:56
<Hixie_>
this could take a while
23:10
<jorendorff>
Domenic_: interesting situation... the Loader's use of Promises is supposed to be kind of an implementation detail, unobservable
23:10
<jorendorff>
Domenic_: but at various points user scripts can screw us over by modifying the builtins
23:11
<jorendorff>
Domenic_: for example, CastToPromise calls GetDeferred which calls C.[[Construct]] which calls C[@@create], which is configurable
23:11
<jorendorff>
big ufn
23:20
<TabAtkins>
jorendorff: We're dealing with that too. The solution seems to be to have some way to provide clean built-ins to browser JS.
23:21
<Hixie_>
that's basically what we do for all the other things that are defined in terms of author-visible features
23:21
<jorendorff>
TabAtkins: oh, dherman has designed an API for this: new Realm
23:21
<jorendorff>
but all the methods are probably configurable
23:21
<Hixie_>
(though i've tended to just define things in terms of algorithms and then invoke the algorithms rather than the native features)
23:22
<Hixie_>
s/native/author-accessible/
23:23
<jorendorff>
that's what I'm used to (algorithms), but Promises are high-level enough that it starts to break down