08:32
<annevk>
FWIW, I informed the IETF of our forking of WebSocket: https://mailarchive.ietf.org/arch/msg/hybi/2OIyLSs5JjDfiFB_I_HGoSinsqc
09:58
<zcorpan>
reflecting as URL seems to have more issues. returns resolved URL when the attribute is absent... <a>.href seems to be better defined. looks like there might be an issue with blob: URLs also. annevk thoughts? should reflect as URL reuse <a>.href machinery?
09:59
<annevk>
zcorpan: I think technically reflecting a URL requires an internal slot for the URL
09:59
<annevk>
zcorpan: this is a problem for most of the reflecting stuff though, that they require an internal slot
10:00
<zcorpan>
http://software.hixie.ch/utilities/js/live-dom-viewer/saved/3994 - difference between gecko and chromium
10:00
<annevk>
zcorpan: I filed a bug on IDL to automatically get internal slots for attributes, unless specified otherwise, but you know how well IDL is maintained
10:03
<zcorpan>
annevk: ok, i think i'll fix some low-hanging fruit today with resolve as url and revisit to do it "properly" some other day
10:06
<annevk>
Yeah, I think we should fix IDL for the proper fix
10:07
<annevk>
It would be nice to standardize [Reflect=URL] or some such too
10:07
<zcorpan>
yeah
10:08
<zcorpan>
oh the spec does handle absent attribute actually
10:13
<zcorpan>
is there any attribute that takes a URL and has a default value? I can't find any
10:20
<annevk>
Seems unlikely
10:20
<annevk>
Don't know for sure though
10:48
<annevk>
jgraham: can the web-platform-tests server be used to transmit a response early?
10:49
<Ms2ger>
What does that mean?
10:50
<annevk>
Ms2ger: transmit a response before the entire request is in
10:50
<Ms2ger>
Hmm
10:50
<annevk>
See https://github.com/whatwg/fetch/issues/229 for context
10:50
<Ms2ger>
I don't know, but I wouldn't be surprised if it couldn't
10:51
<jgraham>
annevk: I doubt it
11:41
<glazou>
zcorpan: ping
11:41
<zcorpan>
glazou: pong
11:41
<glazou>
hi
11:41
<glazou>
I have a question about DOM spec
11:42
<glazou>
https://dom.spec.whatwg.org/#dom-range-clonecontents is supposed to clone a Range but returns a DocumentFragment :-)
11:42
<glazou>
because of https://dom.spec.whatwg.org/#concept-range-clone that itself returns a DocumentFragment
11:44
<Ms2ger>
[NewObject] DocumentFragment cloneContents();
11:44
<Ms2ger>
Why would it return a Range?
11:44
<zcorpan>
Ms2ger was faster than me
11:44
<zcorpan>
:-)
11:45
<glazou>
I really need to stop coffee and go back to beer
11:45
<zcorpan>
glazou: confused with [NewObject] Range cloneRange(); ?
11:45
<Ms2ger>
Now zcorpan beat me :)
11:45
<glazou>
zcorpan: I need vacation, that's all; nm and sorry for annoyance
11:46
<Ms2ger>
Np
11:46
<glazou>
yeah probably a confusion because of "clone" ; a "clone" operation that does not return the same object type... hmmm...
11:47
<zcorpan>
it is confusing, agreed
11:47
<zcorpan>
the concept should maybe be called "clone the contents"
11:48
zcorpan
can PR
11:48
<glazou>
yes that would help
11:50
<benjamingr__>
caitp, littledan__ : hey, I'd love your opinion on https://github.com/nodejs/node/issues/5691#issuecomment-196275629
13:04
<caitp>
benjamingr__: as far as I can tell from api.cc, those api methods just call into the self hosted promise implementation anyways
13:05
<caitp>
so I'm not sure why there would be any difference between them, at least on ToT
13:10
<caitp>
so, Promise::Resolver::Resolve() will enqueue a microtask, so long as the promise wasn't already resolved
13:12
<caitp>
oh I see
13:13
<annevk>
Domenic: https://www.w3.org/Bugs/Public/show_bug.cgi?id=25529 might be of interest (about cloning)
13:27
<annevk>
smaug____: https://www.w3.org/Bugs/Public/show_bug.cgi?id=19962?
13:34
<smaug____>
annevk: yeah, I guess we could add those, given that Gecko has now innerText too and insertAdjacentHTML()
13:35
<smaug____>
hsivonen might have some comment
13:35
<smaug____>
(is he back?)
13:50
<smaug____>
we need a bot to file browser engine bugs
13:51
<annevk>
smaug____: I didn't know he was gone
13:53
<smaug____>
Looks like still away this week
14:45
<annevk>
Ms2ger: FYI: https://github.com/w3c/DOM-Parsing/issues/4
14:45
<annevk>
smaug____: harder to spec than I thought
14:46
<annevk>
philipj_: I found a bug in Blink, it says insertAdjacentElement returns "Element" in your IDL, but it's actually "Element?" per impl (and now spec)
14:46
<Ms2ger>
I didn't fix issues with that spec when I edited it, not going to start now :)
14:47
<zcorpan>
hmmm... http://software.hixie.ch/utilities/js/live-dom-viewer/saved/3998 the spec says absent.href should return http://example.org/ afaict but webkit/chromium/gecko don't do that
14:47
<zcorpan>
can someone check Edge?
14:47
<annevk>
Ms2ger: you didn't write that part? I thought you did
14:47
<annevk>
Ms2ger: anyway, never mind then
14:47
<Ms2ger>
I probably did write it
14:48
<annevk>
I should probably fold that whole spec into DOM at some point
14:48
<Ms2ger>
No time to maintain, though
14:48
<Ms2ger>
Of course it's in licensing hell now
14:49
<annevk>
Ms2ger: hmm yeah, I guess the additions are
14:49
<annevk>
ugh
14:53
annevk
was about to add CDATASection back but then figured out he isn't 30 yet
14:56
tantek
lol annevk
14:59
<zcorpan>
philipj_: still not excited to remove CDATASection? still have a few months :-) https://www.w3.org/Bugs/Public/show_bug.cgi?id=27386
15:00
<zcorpan>
https://www.chromestatus.com/metrics/feature/timeline/popularity/113 0.0048%
15:51
<annevk>
So low and yet too high
16:04
<KiChjang>
i'm currently writing a white paper on the evolution of HTTP, and am wondering what i can include in there
16:06
<annevk>
KiChjang: that's a very generic question, anything specific?
16:08
<KiChjang>
annevk: i'm currently considering how every subsequent versions is trying to fix or improve upon previous versions
16:08
<KiChjang>
like how HTTP/0.9 is a really simple protocol designed to transfer HTML documents as-is without any metadata attached to responses and requests
16:09
<annevk>
KiChjang: I guess you could say that HTTP/1.0 obviated the need for <plaintext>
16:09
<annevk>
but we still have it because HTTP/0.9 was a real thing
16:09
<KiChjang>
annevk: obviated?
16:09
<KiChjang>
as in, it assumes plaintext is used?
16:10
<annevk>
KiChjang: removed the need for
16:10
<annevk>
KiChjang: HTTP/0.9 did content sniffing, afaik
16:30
<annevk>
smaug____: perhaps you know what we should do with https://www.w3.org/Bugs/Public/show_bug.cgi?id=28097?
16:31
<annevk>
Domenic: I thought your custom element work resulted in an abstract operation for creating all elements
16:31
<annevk>
Domenic: which could then have some kind of "element created steps" associated with it, if specifications needed a hook
16:43
<Domenic>
annevk: I guess it could. What does that help with though?
16:43
<Domenic>
annevk: looking through HTML's existing "when the element is created" it's basically just specifying what the UA-callable-only constructor does
16:45
<annevk>
Domenic: I think the main problem Hixie_ had was with attributes, where browsers assume that elements created by the parser are created with attributes, so the question I guess is what happens for cloning
16:46
<Domenic>
annevk: https://w3c.github.io/webcomponents/spec/custom/#clone-modifications-to-dom should hopefully cover it?
16:46
<annevk>
Domenic: I don't immediately know what the observables would be though
16:46
<smaug____>
annevk: looking
16:47
<annevk>
Domenic: yeah, shouldn't typeExtension be an optional argument as well?
16:47
<Domenic>
annevk: my intention was you always supply null. I guess I messed that up in the cloning section.
16:53
<annevk>
philipj_: FWIW, I might actually add CDATASection back before August
16:53
<Domenic>
:(
16:53
<Domenic>
I wish someone would make the parser change
16:53
<caitp>
do you think it would make sense for browserland SVG to be specified in a living standard? all vendors seem to differ in at least a few ways from the snapshot editions
16:53
<Domenic>
It seems plausible that using Text nodes would work
16:54
<annevk>
philipj_: ooh, maybe not, I was thinking it was required for Shadow DOM, but it's not, whatever Shadow DOM adds to Text ends up automatically on CDATASection
16:54
<Domenic>
caitp: I think it would make sense for everything to be specified in a living standard. But, it's up to the editors who have the bandwidth to work on those technologies, and I don't think the SVGWG is that excited about doing so.
16:54
<annevk>
Domenic: I think philipj_ tried and it broke tests at least, e.g., some serialization expectations
16:54
<Domenic>
caitp: in the meantime we have https://html.spec.whatwg.org/#svg-0
16:54
<Domenic>
annevk: yeah, i meant try for real
16:54
<Ms2ger>
A long time ago (when vendor prefixes were still in fashion), I implemented mozSerializeAsCDATA
16:55
<Ms2ger>
(Never landed)
16:55
<annevk>
I have pondered about defining SVG, but it's quite a bit of work
16:55
<annevk>
As is it's quite a mess though, with nothing really well defined 😟
16:56
<caitp>
I just meant some basic ways, was trying to figure out why we add @@iterator to SVGxxList interfaces, blink and geck give it a "length" attribute, but apparently webkit doesn't
16:56
<caitp>
gecko*
16:56
<caitp>
there are probably other things
16:56
<annevk>
caitp: I was thinking basics too, e.g., how <svg:script> works
16:56
<Domenic>
Ooh caitp you're setting yourself up for a PhiStucK email with that "no feature dashboard" entry. Just wait for it...
16:57
<caitp>
there are enough chromestatus pieces for real features that people consciously care about :x
16:57
<annevk>
caitp: at some point the complex stuff is hopefully all abstracted through CSS, and then the remaining complex stuff can be abstracted through HTML
17:09
<jsbell>
Domenic: only if joe doesn't beat phistuck to it...
17:10
<annevk>
Domenic: hayato: I think tomorrow I'm going to try to add attachShadow(), ShadowRoot and the ShadowRootOrDocument mixin to DOM, and update the various algorithms that need to account for them
17:11
<annevk>
Domenic: hayato: and then as exercise I'll try to update Fullscreen, if that's feasible without the flat tree
17:11
<Domenic>
annevk: sounds ambitious, awesome!
17:12
<annevk>
Just the basic building blocks, but yeah, might end up being more than anticipated, we'll see
17:13
<annevk>
The alternative is probably working on outstanding Fullscreen/Storage issues
17:50
<annevk>
jsbell: https://github.com/whatwg/encoding/issues/18, do it?
17:58
<jsbell>
annevk: yep.
18:07
<annevk>
jsbell: so exciting to get rid of that
18:08
<annevk>
jsbell: makes the constructor argument are a bit useless, but okay
18:08
<annevk>
s/are//
18:14
<Domenic>
annevk: can't you just remove the constructor argument?
18:15
<annevk>
Domenic: I guess we could, though it's probably impossible to introduce a different argument at a later stage
18:16
<annevk>
Domenic: but I can make sure that doesn't happen by leaving some kind of note in the source
18:16
<Domenic>
yeah that makes sense
18:16
<Domenic>
can also remove the encoding internal slot
18:16
<annevk>
Actually, should probably keep the first argument, so that it keeps throwing for invalid values
18:18
<annevk>
But yeah, can simplify a couple of things
18:18
<Domenic>
why though
18:18
<annevk>
Next big change after this will probably be streams integration
18:18
<Domenic>
don't throw just ignore it
18:19
<Domenic>
annevk: I don't think anything else on the platform validates an argument it does nothing with.
18:19
<annevk>
Domenic: I guess that's true, I'm mostly worried what happens if we want to add that argument back, but maybe that's not going to be a thing
18:19
<Domenic>
annevk: yeah I don't think it will be. It'll all be fine...
18:25
<KiChjang>
what's HTTPbis?
18:26
<gsnedders>
KiChjang: the set of RFCs that obsoleted RFC 2616
18:26
<gsnedders>
KiChjang: the revision to HTTP/1.1 specs
18:27
<KiChjang>
i thought RFC2616 was the latest HTTP/1.1 spec
18:27
<KiChjang>
so there are others
18:27
<gsnedders>
KiChjang: RFC 7230–7235
18:28
<KiChjang>
gsnedders, ah, so SPDY stuff?
18:28
<gsnedders>
KiChjang: no
18:28
<gsnedders>
KiChjang: this is a simple revision of the HTTP/1.1 spec
18:29
<gsnedders>
KiChjang: SPDY became HTTP/2.0 which is RFC 7540
18:40
<KiChjang>
is HTTP/2 stateful?
18:40
<KiChjang>
it looks like it is, since it has this concept of a stream
18:53
<smaug____>
annevk: finally looked at https://www.w3.org/Bugs/Public/show_bug.cgi?id=28097 but couldn't immediately understand what it even is about. Why any of this stuff has anything to do with host-inclusiveness ?
18:55
<smaug____>
oh, hmm, is this about the cycle?
18:56
<smaug____>
which isn't there...
18:56
<smaug____>
yeah, don't really understand the bug
18:58
<smaug____>
pre-insertion validity check shouldn't probably use "host-including inclusive ancestor "
21:19
<gsnedders>
KiChjang: what sort of statefulness do you mean?
21:20
<gsnedders>
KiChjang: there's no statefulness exposed at the request/response layer beyond what HTTP/1.1 has as the messaging semantics are unchanged
21:48
<Domenic>
TabAtkins: where did "Document Object Model" come from for https://storage.spec.whatwg.org/#normative ?
21:48
<Domenic>
TabAtkins: nevermind, I guess that hasn't been recompiled in a long time
21:51
<TabAtkins>
Domenic: Yeah, it used to be that SpecRef had the W3C version squatting on "DOM", and Shepherd named their ref (which points to the WHATWG copy) dom-ls, so you reffed that. I've since fixed things - the Shepherd ref is now just "DOM", and Bikeshed has an explicit fixup for the squatted WHATWG specs, so biblio-reffing "DOM" will grab all the right data from
21:51
<TabAtkins>
SpecRef's WHATWG-DOM reference.
21:51
<Domenic>
TabAtkins: but it will still show [WHATWG-DOM] as the name, IIRC? Kind of sad.
21:52
<TabAtkins>
It's the result of me applying a lowest-cost fix, until SpecRef fixes itself properly.
21:52
<Domenic>
I have little confidence in SpecRef fixes sadly
21:52
<TabAtkins>
Eh, I believe he'll fix it eventually.
21:53
<TabAtkins>
In the meantime, it only shows as [WHATWG-DOM] in the biblio index; any explicit `[[DOM]]` tags in your source stay that way.
21:54
<Domenic>
Which honestly just seems like a bug? A link text [DOM] going to a [WHATWG-DOM] in the index?
21:54
<Domenic>
It's like, no, I clicked on [DOM], not [WHATWG-DOM]... now I need to scroll up to find where [DOM] is defined... argh it isn't there, where did it go?
22:05
<TabAtkins>
Like I said, lowest-effort fix. Removing this when SpecRef is fixed will be a 2-line deletion commit.
23:14
<tobie>
Domenic: Fwiw, I'm happy to accept a PR in Specref that fixes the problem. I just don't have time to look into it myself right now.
23:17
<tobie>
I agree that this is a problem and want to fix it.
23:18
<tobie>
But I also want to make sure it doesn't get overwritten with the next update and that we don't loose references to previous specs in the process.
23:45
<rniwa>
Domenic: is anyone working on new custom elements API in Blink?
23:49
<Domenic>
rniwa: kochi is
23:49
<rniwa>
Domenic: cool. is there an issue that I can follow?
23:51
<Domenic>
rniwa: I will ask him
23:51
<rniwa>
Domenic: thanks
23:51
<rniwa>
Domenic: between late March and early April, which one do you prefer for a telecon?
23:52
<Domenic>
rniwa: probably early April; TC39 is the last week of March.
23:52
<rniwa>
Domenic: okay
23:55
<rniwa>
Domenic: alright, I just asked for one on public-webapps
23:55
<rniwa>
Domenic: btw, we really appreciate your work for the spec :)
23:56
<rniwa>
Domenic: so are you guys going to be fine with having an explicit method for upgrading custom elements?
23:56
<rniwa>
Domenic: and not upgrading any elements by default when there is no definition at the time of element construction?
23:56
<rniwa>
Domenic: or do you still want the HTML parser to set the flag for auto-upgrades?
23:57
<Domenic>
rniwa: yeah we're having a meeting tomorrow to make sure everyone is on the same page but it looks like maybe a move to in-document-only upgrades is on the table
23:57
<Domenic>
like you originally proposed