00:09
<Domenic>
I believe the proper term is Web Hypertext Application Technologies Un-Plenary
00:09
<Domenic>
wait no it was Unconference Plenary
00:33
<tantek>
Domenic++
00:33
<TabAtkins>
+1
05:55
<annevk>
rniwa++ for tests
06:03
<rniwa>
annevk: :)
06:03
<rniwa>
annevk: i've got a few dozens of tests for shadow DOM and custom elements now:
06:04
<rniwa>
https://github.com/WebKit/webkit/tree/master/LayoutTests/fast/shadow-dom
06:04
<rniwa>
https://github.com/WebKit/webkit/tree/master/LayoutTests/fast/custom-elements
06:35
<philipj_>
annevk: I don't like having prefixes to the commit messages, it uses up valuable space. So I'm also not a fan of "Fix #100:" and like to put this in the description
09:46
<rniwa>
Domenic: yt?
09:53
<annevk>
rniwa: it's nearly 5AM if he isn't traveling
10:03
<zcorpan>
mutation events :-| ... https://www.chromestatus.com/metrics/feature/timeline/popularity/144 at least it's going down
10:05
<annevk>
I'm hoping esprehn can change the timing of mutation events
10:05
<annevk>
If that's feasible we should just add them to the standard and forget about them
10:14
<annevk>
Looks like I'm going to standardize port banning today ๐Ÿ˜ƒ
10:14
<zcorpan>
https://www.chromestatus.com/metrics/feature/timeline/popularity/146 DOMNodeRemovedFromDocument is at 0.0086% now
10:29
<smaug____>
Hmm, there was some discussion about hitregions in some github bug recently... anyone happen to recall which bug it was
10:29
<smaug____>
not https://github.com/whatwg/html/issues/849 but some other
10:29
<smaug____>
(and could we move issue tracking back to some bugzilla instance)
10:31
<zcorpan>
miketaylr: fyi https://lists.w3.org/Archives/Public/www-style/2016Feb/0169.html
10:32
<smaug____>
ok, found it. https://github.com/whatwg/html/issues/548
11:19
<annevk>
smaug____: we cannot, we migrated away from Bugzilla
11:25
<annevk>
smaug____: so your suggestion for hit regions is to make the hit region object (is there such a thing?) inherit from EventTarget?
11:26
<annevk>
smaug____: seems somewhat reasonable
11:26
<annevk>
smaug____: would its "get the parent" return the <canvas> it's associated with?
11:28
<smaug____>
I don't think I suggested that
11:31
<smaug____>
annevk: I'm talking more about these steps: "Let control be the control represented by region for this canvas element, if any." ... "If there is a control, then target the event object at control instead of the canvas element."
11:32
<smaug____>
so we already dispatch the event on the "hit region control"
11:32
<smaug____>
and event.region == event.target.id
11:32
<smaug____>
or something like that
11:33
<annevk>
Seems indeed redundant then
11:34
<smaug____>
but Touch handling is something different
11:36
<smaug____>
but I think mostly the whole .region stuff is busted.
11:36
<smaug____>
whoever designed the API could explain why it is the way it is
11:37
<annevk>
I think Hixie_ designed it
13:05
<zcorpan>
annevk: random thought: it seems slightly weird to have "should" in a defined term
13:12
<Ms2ger>
mix and csp do the same
13:37
<zcorpan>
i see Sec-WebSocket-Extensions: permessage-deflate; client_max_window_bits as a request header in opera
13:39
<zcorpan>
and .extensions returns "permessage-deflate"
14:08
<annevk>
zcorpan: that is weird
14:08
<annevk>
zcorpan: ugh
14:09
<annevk>
zcorpan: file an issue? That requires a bunch of changes to various drafts
14:09
<zcorpan>
ok
14:41
<annevk>
zcorpan: did you work on WebSocket tests?
14:41
<annevk>
zcorpan: the README doesn't really say where to find server code
14:41
<wanderview>
annevk: step 29 here seems to allow HEAD no-cors requests... https://fetch.spec.whatwg.org/#dom-request
14:42
<zcorpan>
annevk: i did, yeah. the server code is pywebsocket i believe
14:42
<Ms2ger>
Yep
14:42
<annevk>
zcorpan: did you test protocols not matching?
14:42
<annevk>
zcorpan: and where do I find that code in relation to the tests?
14:43
<annevk>
wanderview: hmm, I wonder if that's accurate
14:43
<zcorpan>
annevk: trying to find a test that tests the removed protocols thing, but little luck so far
14:43
<annevk>
wanderview: afaict there's not a single feature on the web that can do HEAD across origins
14:43
<wanderview>
annevk: why is it bad? if it ends up going to a same-origni you can see status... if you end up going cross-origin you see 0 status
14:43
<zcorpan>
annevk: the handlers are in websockets/handlers/
14:43
<annevk>
wanderview: even if we did allow HEAD though, that doesn't mean cache should be able to modify opaque responses
14:44
<wanderview>
annevk: at Request constructor time you don't know if you will get an opaque response or not if you fetch() it
14:45
<wanderview>
annevk: where is the failure for doing the HEAD request cross-origin?
14:45
<zcorpan>
annevk: hmm here we have something ./constructor/012.html: var ws = new WebSocket(SCHEME_DOMAIN_PORT+'/handshake_no_protocol', 'foobar');
14:46
<annevk>
wanderview: I'm saying there might be a bug in Fetch that we allow that, or maybe we decided on allowing it since nobody thought it could do harm
14:46
<annevk>
wanderview: I'm vaguely remembering it might be the latter
14:46
<annevk>
wanderview: but that still doesn't mean you should be able to modify an opaque response
14:47
<wanderview>
annevk: the browser can clearly modify opaque responses and does whenever it sets the bodyUsed by internally draining
14:48
<wanderview>
annevk: the cache spec is generally problematic here, though, because it does not serialize and deserialize the Response... so it looks like it just has a list of Response objects... its really creating a new one each time it returns
14:48
<annevk>
zcorpan: contents of handshake_no_protocol_wsh.py are a little weird
14:48
<zcorpan>
annevk: yeah, it's a hack due to limitations of pywebsocket (or limited understanding) at the time the tests were written
14:54
<annevk>
zcorpan: thanks
14:54
<wanderview>
annevk: so we don't think we will ever want to send a password credential without cookies?
14:55
<annevk>
wanderview: it doesn't seem great to let developers modify opaque responses, although I'd have to page everything in again as to why that might be problematic
14:55
<annevk>
wanderview: well, that's what mkwst thinks
14:55
<wanderview>
oh, that's mkwst's irc handle! I kept looking for him here
14:55
<annevk>
wanderview: I raised that point in the PR
14:56
<wanderview>
yea
15:00
<annevk>
Domenic: https://github.com/whatwg/html/pull/840 all good now or still wary?
15:00
<wanderview>
HTML seems particularly hostile to diff reviewing for some reason
15:01
<zcorpan>
annevk: just added a comment
15:33
<annevk>
zcorpan: thanks, so how about I add a check in Fetch
15:33
<zcorpan>
annevk: wfm
15:34
<annevk>
zcorpan: "If _protocols_ is not the empty list and _response_'s header list does not contain a header named Set-WebSocket-Protocol, then fail the WebSocket connection"
15:37
<zcorpan>
annevk: also check if the header is present but doesn't contain any value
15:38
<zcorpan>
(havne't tested that but it seems it should be equivalent to omitting the header)
15:48
<annevk>
zcorpan: https://github.com/whatwg/fetch/pull/242
15:53
<zcorpan>
r+; need to leave now
16:21
<mkwst>
wanderview: Hi, sorry. Wasn't looking at IRC.
16:21
<wanderview>
not a problem
16:21
<mkwst>
wanderview: No, I don't think it makes sense to separate this kind of credential from any of the other kinds of credentials Fetch already deals with.
16:22
<wanderview>
yea, sorry... I finally found the discussion in the PR
16:22
<mkwst>
That is, if I'm trying to authenticate a session, I'm going to want to manage HTTP state along with the authentication information I'm sending over.
16:23
<mkwst>
(and sorry about the IRC handle. It seemed like a BRILLIANT idea in the heady, vowelless days of Web 2.0.)
16:35
<Ms2ger>
Pff, even I have a vowel
16:48
Ms2ger
realizes he's had this name for over a decade now
17:59
<gabr13l>
o/ WHATWG crew
18:00
gabr13l
used to be cgcardona FWIW
18:05
<annevk>
gabr13l: good evening
18:06
<gabr13l>
hi annevk
18:30
<wanderview>
mkwst: btw, I'm still concerned about Matt's feedback here: https://groups.google.com/a/chromium.org/d/msg/blink-dev/7ouLjWzcjb0/s7aZHGnlAwAJ
18:30
<wanderview>
mkwst: I'm asking the person interested in implementing the spec in gecko to post to dev-platform so we can come to some concensus within mozilla
18:52
<MikeSmith>
hey gabr13l
18:52
<gabr13l>
hi MikeSmith
18:56
<smaug____>
annevk: trying to understand why people are against the 4th param version of addEventListener
18:57
<smaug____>
any new arguments ?
18:57
<smaug____>
other than "I don't like it"
18:57
<smaug____>
the API happens to be so commonly used that we need to be _very_ careful with all the compatibility issues
18:58
<smaug____>
and 4th param approach should be safer than 3rd param approach
18:58
<smaug____>
(and I'd really prefer if we spent less time on this and more time on designing some new way to add listeners)
18:59
<smaug____>
(since I assume that will take several months to get right)
19:56
<TabAtkins>
smaug____: If we can *actually commit* to doing that, and not just (as I expect) say "Yeah, we'll fix this in a new API" and then never develop it, then I'm okay with the stupid 4th-arg approach.
19:56
<TabAtkins>
I'm just pretty sure that no one will actually spend the time to design a new API, or if they do, it'll get shot down as not worth the implementation effort for a minor usability benefit.
19:57
<TabAtkins>
See: what has happened with createElement for the last several years.
19:58
<smaug____>
what has happened with createElement?
19:58
<smaug____>
it hasn't really changed
19:58
<smaug____>
this way radically
20:00
<TabAtkins>
People keep asking for a *usable* version of it, that lets you set attributes/properties/children.
20:00
<TabAtkins>
What every single DOM wrapper library provides, because the current DOM API is really terrible in usability.
20:00
<TabAtkins>
(I have one in Bikeshed - E.h1({attrs}, children...), for example.)
20:01
<TabAtkins>
But even tho the ideas get well-developed, no impls ever bite, because it doesn't actually provide anything *new*.
20:01
<TabAtkins>
And afaik, the same is true of a better-designed event-listener API - it's just "more usable", not "new functionality".
20:02
TabAtkins
is heading to lunch
20:36
<rbyers>
TabAtkins: if you or annevk or someone can commit to getting a new API into the spec, I'm happy to commit to implementing and shipping it in blink.
20:41
<Domenic>
rbyers: TabAtkins: annevk: https://github.com/straker/html-tagged-template/ looks like the best bet for me
20:41
<Domenic>
(modulo https://github.com/straker/html-tagged-template/issues/6 )
20:45
<annevk>
Domenic: I think he meant a better event listener API
20:46
<Domenic>
ah, right, that would make sense :P
20:46
<annevk>
smaug____: there is no backcompat issue
20:47
<rbyers>
annevk: I talked a bit with smaug offline, it's all the Firefox ESR users (9-12 months old) that he's worried about - so forwards compat, but not back compat.
20:48
<rbyers>
.. that said, it's still possible there's a back compat issue - we just haven't found any concrete examples yet (and it's now in Chrome stable, so if there are a non-trivial number we should hear soon).
20:50
<annevk>
Those users will be in trouble anyway with each attractive new feature
20:50
<smaug____>
addEventListener is very different thing than some new attractive feature
20:51
<annevk>
This API will be around much longer than twelve months, it's important to do right by it
20:52
<smaug____>
and hacking 3rd param isn't more correct than 4th param approach. The 4th param approach is even less hacky since it doesn't add this bizarre boolean or dictionary thing
20:52
<smaug____>
this is of course very subjective
20:55
smaug____
wonders what cloneRecord.[[TransferList]] means
20:55
<smaug____>
in the spec
21:11
<Domenic>
Am I not following enough links or is XHTML parsing rather under-defined
21:11
<Domenic>
Like where does it say for XHTML that <a> tags -> HTMLAnchorElement and such
21:16
<zcorpan>
Domenic: it's under-defined. there's no spec for how to turn an XML byte stream into a DOM at all.
21:17
Domenic
flips some tables
21:19
<zcorpan>
iirc there's Infoset that defines a mapping for the Infoset, but there's no explicit mapping between Infoset and DOM
21:22
<zcorpan>
fun fact: https://www.w3.org/TR/xml-stylesheet/ is defined in terms of Infoset
21:26
<gsnedders>
zcorpan: as is XPath and XSLT
22:52
<JonathanNeal>
Should a <picture> element display like an <img>? In that itโ€™s an inline element that behaves like an inline-block element?
22:56
<zcorpan>
JonathanNeal: no <picture> displays like a <span>
22:57
<zcorpan>
it's the child <img> that shows the image
22:59
<JonathanNeal>
@zcorpan, this is the expected result? http://codepen.io/anon/pen/KzMmyX
22:59
<zcorpan>
JonathanNeal: yes
23:01
<zcorpan>
JonathanNeal: you can style the <img> as usual
23:01
<zcorpan>
HTH
23:01
zcorpan
-> bed