00:40
<TabAtkins>
If anyone knows off the top of their head, does the current source selection algorithm for <img> stop when it finds its preferred source, and not try for additional ones if the first one fails?
00:40
<TabAtkins>
(If no one answers in a bit I'll go look for the answer myself.)
07:32
<zcorpan>
TabAtkins: right, img doesn't try further if a source fails to load
07:36
<zcorpan>
hmm, i got a spam email that was kind enough to let me know where it found my email address (the cssom spec)
08:00
<hsivonen>
zcorpan: was it European spam if the spammer thinks that declaring the source of address makes it better?
08:03
<zcorpan>
google translated: "I am a crawler that found email simonp⊙oc on w3.org site exactly http://dev.w3.org/csswg/cssom/ link and I will aim to post your free ad on <something>.ro"
08:13
<hsivonen>
zcorpan: indeed unusual
08:19
<MikeSmith>
sendBeacon makes <a ping> superfluous?
08:20
<Ms2ger>
Right now when we've implemented it?
08:24
<MikeSmith>
Ms2ger: there's an open bug
08:25
<MikeSmith>
https://bugzilla.mozilla.org/show_bug.cgi?id=936340
08:26
<MikeSmith>
and jonas comments at https://bugzilla.mozilla.org/show_bug.cgi?id=951104
08:26
<MikeSmith>
Ms2ger: <a ping> shipped already?
08:27
<MikeSmith>
no just in nightly I guess
08:27
<Ms2ger>
Still preffed off
08:27
<MikeSmith>
ah ok
08:27
<MikeSmith>
ah yeah hence the summary of that bug
08:32
<zcorpan>
<a ping> seems simpler to get right. what if the user middle clicks the link, will the page notice and send the ping?
08:32
<zcorpan>
if implemented with sendBeacon, i mean
08:33
<zcorpan>
what's the problem with having both?
08:33
<zcorpan>
is ping high-cost?
08:38
<MikeSmith>
zcorpan: I guess we need to keep <a ping> regardless anyway
08:38
<MikeSmith>
which is not a problem really
08:39
<MikeSmith>
just if somebody had come up with sendBeacon first maybe we wouldn't have needed <a ping> to begin with
08:42
<zcorpan>
"The User Agent MUST throw the URLMismatchError exception if the URL cannot be resolved."
08:43
<zcorpan>
seems like the wrong exception
08:43
<zcorpan>
"The User Agent MUST throw the SyntaxError is data is null." also seems like the wrong exception
08:46
<Ms2ger>
'is'?
08:49
<zcorpan>
http://lists.w3.org/Archives/Public/public-web-perf/2013Nov/0096.html seems like an interesting proposal. maybe it should allow you to set sync to true but have that be ignored if detach is true
09:29
<darobin>
I'm still not convinced this whole beacon stuff is headed the right way
09:30
<darobin>
it doesn't seem to take into account the use case of e.g. uploading a picture taken in a web app in such a way that you can leave the app immediately but get a notification later upon success
09:31
<darobin>
which is a common thing that can only be done using native apps today
09:59
<jgraham>
darobin: Isn't the difficult part there continuing the upload after you navigate?
09:59
<darobin>
jgraham: well, that's the interesting part :)
10:00
<darobin>
you also need to make it so that notifications can be triggered after navigation (and that clicking on them can reopen a page)
10:00
<jgraham>
Seems like could could solve that with background shared workers, perhaps
10:01
<darobin>
ideally, progress notifications ought to be possible during upload
10:01
<darobin>
yes, maybe
10:01
<jgraham>
At least it seems quite different to the beason stuff
10:01
<jgraham>
*beacon
10:01
<darobin>
that's essentially how it works on Android, apps spawn services that are permanent
10:02
<darobin>
well, if you look at the code examples that developers show when telling what they'd like for fire-and-forget upload, they look a lot like beacon
10:02
<darobin>
or at least some of the beacon proposals
10:02
<darobin>
"give me an XHR that doesn't die"
10:06
<jgraham>
Well, yes if that's the only use case
10:06
<jgraham>
If there are other similar use cases then a different solution is needed
11:03
<zcorpan>
sent http://lists.w3.org/Archives/Public/public-web-perf/2013Dec/0105.html
11:27
<jgraham>
zcorpan: I like how you start with a bunch of small issues and end with "oh and your spec is totally broken"
11:27
<zcorpan>
heh
11:29
<zcorpan>
and i didn't even ask about use cases
11:30
<zcorpan>
step 1: minor bugs. step 2: fundamental brokenness. step 3: possibly wrong solution altogether
11:34
<jgraham>
You should obviously reply to your email again :)
11:36
<zcorpan>
i'll leave step 3 to someone else
12:09
<zcorpan>
actually the brokenness seems to apply to <a ping> also
12:09
<hsivonen>
is one supposed to use a semicolon after a nested function declaration in JS?
12:09
<Ms2ger>
I don't think so
12:09
<hsivonen>
Ms2ger: thanks
12:09
<zcorpan>
function declarations don't need semicolon after
12:10
<hsivonen>
zcorpan: ok. thanks
12:10
<zcorpan>
only if you do an assignment like onclick = function() {};
12:10
<hsivonen>
ok
12:11
<jgraham>
Yeah, no semicolon after statements, only after expressions I think
12:15
<hsivonen>
jgraham: I think you mean the other way round, right?
12:16
hsivonen
wasn't sure if a function declaration counted as a statement
12:18
<jgraham>
hsivonen: I don't really mean either, it turns out
12:19
<Ms2ger>
JS is weird
12:19
<jgraham>
http://www.ecma-international.org/ecma-262/5.1/#sec-A.4
12:30
<matjas>
ah, the good ol’ function declaration vs. function expression
13:07
<gsnedders>
hsivonen: Function statements as implemented everywhere don't exist in the spec. :)
13:08
<gsnedders>
hsivonen: Function declarations per spec must be at the top level; only function expressions are allowed to be nested.
13:08
<jgraham>
Wait, what? They still didn't fix that?
13:10
<gsnedders>
Oh, it appears to have been fixed, but the spec isn't up to date with what was agreed at Sept F2F
13:11
<gsnedders>
I missed that entirely.
13:12
<Ms2ger>
What, an up-to-date spec?
13:12
<Ms2ger>
I've missed that quite often
13:12
<zcorpan>
was the bug something like (function() { if (false) function foo() {}; foo(); /* foo is undefined in impls */ })() ?
13:14
<gsnedders>
zcorpan: That simply didn't match the grammar.
13:15
<gsnedders>
zcorpan: the "function" keyword token was a parse error if you only implemented the ES5 grammar
13:16
<hsivonen>
in any case, it looks like my patch didn't make strict mode or whatever checks apply to Firefox UI code to blow up
13:16
<zcorpan>
gsnedders: oh, ok
13:17
<gsnedders>
zcorpan: (the second "function", obv.)
13:32
<zcorpan>
annevk-cloud: does document.URL return HTML's "the document's address"? it says it returns the "URL" which seems like it is a separate concept
13:37
<zcorpan>
annevk-cloud: http://software.hixie.ch/utilities/js/live-dom-viewer/saved/2707 suggests it should return "the document's address"
13:37
<zcorpan>
Hixie: ^
14:19
<Ms2ger>
"The initial value of Symbol.isRegExp is the well known symbol @@isConcatSpreadableisRegExp"
14:19
<Ms2ger>
@@isConcatSpreadableisRegExp, really?
14:21
<jgraham>
"well known"
14:30
<hsivonen>
annevk-cloud: I finally got around to sending email to whatwg and www-international about TLD-based guessing
14:30
<hsivonen>
I hope the message was coherent enough for non-Firefox audience
14:30
<hsivonen>
I adapted it from an older bugzilla comment
14:47
<csmith1>
is this the right place to ask about http://wiki.whatwg.org/wiki/MetaExtensions ?
14:49
<csmith1>
for example wt.* (webtrends) proposal… should these be generic, rather than product-specific for metrics?
14:55
<matjas>
jgraham: “well known” refers to this table https://people.mozilla.org/~jorendorff/es6-draft.html#table-1
14:55
<matjas>
“Table 1— Well-known Symbols”
14:56
<jgraham>
matjas: I know what well known means in this context :) It's just an amusing choice of name
15:04
<Ms2ger>
"Posted to whatwg and www-international but not cross-posted"
15:04
Ms2ger
doesn't follow
15:06
<gsnedders>
Um…
15:11
<jgraham>
Presumably posted to both but each without the other on the to: header
15:12
<jgraham>
i.e. two seperate emails
15:16
<Ms2ger>
Ah
15:18
<jgraham>
Ms2ger: So, why did I get your email 3 times?
15:18
<jgraham>
:)
15:18
<Ms2ger>
You're on all three lists I sent it to? :)
15:18
<Ms2ger>
Sorry, should also have cc'd you :)
15:20
<jgraham>
Interesting
15:20
<jgraham>
Thunderbird claims they are all To: public-test-infra, but doing Reply-List on all three will reply to three different lists.
15:25
<Ms2ger>
Yeah
15:26
<Ms2ger>
I only got one copy of Art's email, and reply-to-list went to p-w-t
17:09
<MikeSmith>
what constitutes a declarative API?
17:12
<jgraham>
Either a) one that you want but aren't getting or b) one that someone else wants but you don't want to give them
17:13
<jgraham>
But I think typically people mean one that isn't implemented in script
17:13
<MikeSmith>
I like your first answer better
17:15
<MikeSmith>
so then next part of this quiz is what's declarative about approach B) in http://lists.w3.org/Archives/Public/public-webapps/2013OctDec/1096.html
17:15
<jgraham>
(but I think people can use "declarative API" to mean "high level API that allows implementing common tasks without worrying about the underlying mechanics", hence my first answer)
17:16
<MikeSmith>
ah
17:17
<MikeSmith>
oh man http://wiki.apidesign.org/wiki/Declarative_Programming is fun
17:18
<MikeSmith>
it reads like it was writteng by horsejs
17:18
<MikeSmith>
"JavaScript is just too powerful language for declarations."
17:19
<MikeSmith>
"That is why, if one wishes to worship declarative programming inside existing imperative languages, it seems very important to make the declarative syntax less powerful than Turing Machine."
17:20
<jgraham>
All hail our declarative overlords
17:24
<MikeSmith>
J̥̫͗̅ͥ̄͛͡ȧ̛͎͈̬̺̳̤̘̔͗ͣ́ͅv̵͉͕̯̗͓̠̫͈͑̐̊̔ͪ̃͠a͍͇̼̙ͭ̀ͅS̨͕̘̔͛̈̽̑̑ͭ̏ͤc̵̪̳̼̲̲̘̟ͮ͊r̵͉̼̲͍̱̱̋̇̉̂͟i̸͓̺͍͎͔̱̭͆͊ͬ̑̾͐̊p̋̽̎̑͒̅͑̎҉͖̞̥t͙̺̠̻͖͕̒͐̓͜ ̷͍̜̝͂̈̄͛ï̸̧̛̼̼͔̭̤̥̪̋ͥͭ͑̒͂̀s̴̢̤̯̰̗̲͙̬͗͛ͬͪ̅̀̽͢ ̫͚͆ͨͬ͗͛̌́̀͠j̀͑ͬͧ̍͗͏̯̙̳̀͡u̙̻̘ͪ̓s͖̬̮̯͈̱ͬͨ̀̂̿̚͟͡t̷̻̝͙̂̆̃̃ͭͮ͌̅ ͦͩ҉͓̳�
17:52
<annevk-cloud>
zcorpan: they match
17:59
<bholley>
Hixie: any last-ditch ideas on how to get input from Microsoft and Apple before we meet tomorrow?
18:23
<MikeSmith>
bholley: ping hober
18:23
<bholley>
hober: ^
18:26
<MikeSmith>
from Microsoft you need Travis I guess
18:26
<MikeSmith>
somebody could e-mail him if you haven't already
18:32
<TabAtkins>
MikeSmith: "Declarative" and "imperative" are obviously spectrums, but declarative APIs are more rule-based and "let the UA figure it out" rather than the author explicitly doing things themselves.
18:32
<MikeSmith>
ok
18:33
<TabAtkins>
So, though that's a large email that I'm not going to do any more than skim, it does look like the "caches" approach is slightly more declarative than option A.
18:33
<MikeSmith>
TabAtkins: that sounds like what people usually call a high-level API vs a low-level one though
18:33
<TabAtkins>
Yes, high-level APIs are usually "high-level" because they're more declarative.
18:38
<MikeSmith>
bholley remind me whats that bug
18:38
<bholley>
MikeSmith: I already emailed Travis
18:38
<bholley>
MikeSmith: https://www.w3.org/Bugs/Public/show_bug.cgi?id=20701
18:39
<MikeSmith>
ok
18:40
<MikeSmith>
ah great I follow that link and it launches firefox nightly I just finished building and I get XML Parsing Error: undefined entity Location: chrome://browser/content/browser.xul
18:44
<MikeSmith>
hmm well http://developer.chrome.com/extensions/events.html#declarative is interesting
18:45
<MikeSmith>
var rule = { conditions: [ /* my conditions */ ], actions: [ /* my actions */ ]
18:45
<MikeSmith>
};
18:53
<TabAtkins>
MikeSmith: Ooh, that is rather interesting.
18:54
<MikeSmith>
something cooked up just for extensions I guess
19:07
<TabAtkins>
Okay, now that I'm reading Mellor's background sync thread in Gmail rather than the archives, it's much more understandable.
19:07
<TabAtkins>
One of the few times I've ever experienced in mailing lists where HTML emails are more readable than plain text.
19:10
<TabAtkins>
(Though I suspect that if the email were plain-text Markdown instead, it would be just as readable in the archives.)
19:19
<MikeSmith>
the plain text version seems pretty readable as is in my mua
19:20
<TabAtkins>
It's not terrible, it's just wall-of-texty, and difficult to distinguish the cases.
19:20
<TabAtkins>
A bit of bolding in the HTML version helps a ton for skimming.
19:20
<MikeSmith>
ah he's got some boldface of stuff in the HTML version though, that's nice
19:20
<MikeSmith>
in the code examples I mean
19:20
<TabAtkins>
Yeah, that's very helpful as well.
19:21
<TabAtkins>
Now I'm finally motivated to go read the ServiceWorker spec. It looks quite good!
19:22
<MikeSmith>
TabAtkins: yeah now only if there were actually a ServiceWorker spec that exists
19:22
<MikeSmith>
you'd be all set
19:22
<TabAtkins>
Yeah, the explainer documents are really nice, though.
19:22
<MikeSmith>
yeah
19:22
<TabAtkins>
I might ping Alex about working on that for real.
19:22
<MikeSmith>
I thought Jake said he was planning to work on it very soon
19:23
<TabAtkins>
Ah, cool.
19:39
<gsnedders>
This is so boring. Why did I decide to essentially rewrite the parser in process algebra as a uni project?
19:54
<gsnedders>
Just, implementing the state machine is so dull.
19:55
<odinho>
lol :)
19:57
<odinho>
I really liked reading the serviceworker docs as well. meant for a wide audience i guess :)
20:16
<TabAtkins>
Hahaha, I like the paragraph in the Explainer under "Offline & Fallback Content" that's just a single italic "Indeed".
20:16
<TabAtkins>
It's fun when you can tell someone's speaking mannerisms are bleeding into their writing.
20:35
<zcorpan>
annevk-cloud: where is it defined that they match?
21:17
<Ms2ger>
tobie, I don't follow the "This would be very detrimental to vendors running these test suites." comment
21:22
<tobie>
Ms2ger: did I misunderstand the mail?
21:22
<tobie>
I read it quickly.
21:23
<tobie>
I thought he was suggesting splitting the repository.
21:23
<Ms2ger>
Oh, I hope not
21:23
<Ms2ger>
I think that was about bugzilla components
21:24
<Ms2ger>
If we agree Bugzilla should be used to report all test case bugs: 1) should we have an agreed way for a test suite in WPT to point to Bugzilla (although Bugzilla has bug reports for websockets and workers tests, that link is missing from the test suites); 2) should we continue to lump all of the tests in a single component or create per test suite components (e.g. tests-workers, workers-tests, ...).
21:35
<zcorpan>
i don't want to hold a presentation at test the web forward going through the steps to create a bugzilla account and report a test bug
21:36
<zcorpan>
so, i agree with jgraham
23:48
<deadowl>
Why isn't the time element married to the Date objects of javascript? Anyone working on that?
23:49
<deadowl>
p.s. time, date, and datetime ought to be synonymous