00:44
<MikeSmith>
hah :)
01:36
<MikeSmith>
caitp: jochen__ hayato any insight on https://bugs.chromium.org/p/chromium/issues/detail?id=636112
01:36
<MikeSmith>
test failures in blink of the form, class string of PaymentRequest.prototype expected "[object PaymentRequestPrototype]" but got "[object PaymentRequest]"
01:36
<MikeSmith>
for various types
01:37
<MikeSmith>
or see https://bugs.chromium.org/p/chromium/issues/detail?id=635694 because I think my issue there is probably just a duplicate of the (general) problem described tehre
01:43
MikeSmith
bugs yukishiino over on #chromium
02:21
<Domenic>
MikeSmith: this is an intentional change
02:21
<Domenic>
MikeSmith: there is a Web IDL spec bug open on it
02:23
<rniwa>
Domenic: hi
02:23
<rniwa>
Domenic: so it looks like Chrome doesn't throw when calling customElements.define('a-b', HTMLElement)
02:23
<rniwa>
Domenic: but the indent is to throw in that case, right?
02:24
<rniwa>
(per step 2 of https://html.spec.whatwg.org/#dom-customelementsregistry-define)
02:24
<Domenic>
rniwa: yeah definitely. I thought that was being worked on very recently, maybe it hasn't landed...
02:24
<rniwa>
Domenic: okay, it fails as of 54.0.2825.0
02:25
<Domenic>
rniwa: I'll file the bug if there's not one open already. Thanks.
02:25
<rniwa>
Domenic: it would be really useful for all of us to get gather and make sure our implementation of shadow DOM & custom elements are consistent
02:25
<rniwa>
Domenic: there have been enough moving parts in the spec that I'm afraid one of us is going to miss something
02:25
<Domenic>
rniwa: yeah I know dominiccooney is intending to upstream the tests to WPT format; he's been writing a lot in testharness.js style
02:25
<rniwa>
Domenic: cool
02:26
<rniwa>
Domenic: we also have a dozen or so tests that I'm intending to upstream
02:28
<rniwa>
Domenic: so this step 2 is somewhat problematic & ambiguous
02:29
<Domenic>
:(
02:29
<rniwa>
Domenic: how are we supposed to be checking the inheritance?
02:29
<Domenic>
rniwa: via the inherited interfaces concept defined in Web IDL?
02:29
<rniwa>
Domenic: because it's totally possible for author to override prototype/__proto__ of HTMLInputElement
02:29
<rniwa>
Domenic: and make it a getter, right?
02:29
<Domenic>
Right it's a static relationship
02:30
<rniwa>
Domenic: ah, okay
02:30
<Domenic>
Not instanceof
02:47
<rniwa>
Domenic: so that poses a different challenge that I don't think we have any mechanism to check something like that in our engine
02:47
<rniwa>
Domenic: right now, this isn't an issue because none of subclasses of HTMLElement is constructible but
02:48
<Domenic>
rniwa: right, when I was adding that I tried to ask if it was something implementable. I'm open to other strategies as long as we maintain the invariant that you can't have a HTMLButtonElement with a local name that is not button.
02:50
<rniwa>
Domenic: It seems a bit strange to specifically throw an exception on HTMLElement and its subclasses
02:50
<rniwa>
Domenic: when we don't throw on SVGElement, Text, Range, etc...
02:51
<Domenic>
rniwa: trying to page this back in... I believe the problem is that the others will fail later, whereas HTMLElement and its subclasses will not.
02:51
<Domenic>
rniwa: right, the others will fail at https://dom.spec.whatwg.org/#concept-create-element step 6.1.3
02:51
<rniwa>
Domenic: it can certainly fail if we checked new.target
02:52
<Domenic>
Hmm
02:53
<rniwa>
Domenic: it's fairly easy for us to check that something inherits from HTMLElement and then it's not one of DOM classes
02:53
<rniwa>
Domenic: so we can just do that to throw an exception in that case
02:53
<rniwa>
Domenic: since new.target can't be getter, etc... this is basically unobservable from scripts
02:53
<Domenic>
rniwa: can you open an issue if you think we should change this? I want to be sure I have time to go through all the implications and I'm heading to bed soon. I imagine changing it will be possible but I remember investigating this a few weeks ago and thinking that the current spec was the alternative I liked the most, but not strongly.
02:54
<rniwa>
Domenic: sure
02:54
<rniwa>
Domenic: I'm gonna do that on webcomponents issue tracker for now
02:54
<Domenic>
sounds good
03:08
<MikeSmith>
mkwst: I think https://mikewest.github.io/origin-policy/ is pretty clearly already sufficiently viable enough to merit being moved into https://github.com/wicg if you wanted to
03:09
<MikeSmith>
mkwst: though I don’t really know what the WICG criteria are for accepting new repos
03:09
<MikeSmith>
mkwst: nor am I saying it would be necessarily a win to move it
04:43
<annevk>
mkwst is relaxing, try again next week
05:05
<MikeSmith>
annevk: ah OK
05:06
<MikeSmith>
tobie: for WebIDL please consider prioritizing https://www.w3.org/Bugs/Public/show_bug.cgi?id=28244 for resolution
07:09
<annevk>
Opera hasn't hired a new foolip thus far, email address bounces
07:09
annevk
did a reply-all to an old Blink thread
07:23
<annevk>
Sebmaster: did you see https://twitter.com/jasnell/status/763479355196596224 btw?
07:24
<annevk>
I wonder if Sam Ruby has been involved with that since they're both at IBM or if that organization is so large it just sorta happened
07:44
<Ms2ger>
annevk, https://bugzilla.mozilla.org/show_bug.cgi?id=1294100
07:47
<annevk>
Ms2ger: sigh
07:47
<annevk>
Ms2ger: I guess we can overload if it gets bad
07:48
<annevk>
Ms2ger: not the end of the world and certainly not the ugliest API, but why do we always try?
07:49
<annevk>
Ms2ger: left a comment
07:50
<Ms2ger>
Thanks
08:05
<Ms2ger>
One more stupid extension gone: https://hg.mozilla.org/mozilla-central/rev/0ca0282fe48b
08:05
<annevk>
Whoa, Thomas is all over the map
08:25
<foolip>
annevk: :) If you have philipj⊙oc in bugs or email threads, please do replace with something not so bouncy
08:27
<tobie>
MikeSmith: mmm.
08:36
<tobie>
MikeSmith: I'm new to this. What's the strategy here? Toss a coin and beat whomever doesn't accept that resolution until they do?
09:02
<MikeSmith>
tobie: dunno, really. Maybe Domenic has some good suggestion
09:03
<MikeSmith>
tobie: if you actually change the spec maybe bz would reconsider
09:04
<MikeSmith>
would be good to get more details from him or others at mozilla about the compat problems he alludes to
09:06
<MikeSmith>
foolip might also have some insight on https://www.w3.org/Bugs/Public/show_bug.cgi?id=28244
09:17
<foolip>
MikeSmith: ugh, not sure what to make of that
09:20
<annevk>
JakeA: https://twitter.com/jaffathecake/status/763662818277265408 is a great sketch
09:20
<annevk>
JakeA: not a fair comparison though 😊
09:20
<JakeA>
It pops into my head loads
09:22
<JakeA>
Just read an article where the UK divers offered their "reckons" about the green water in Rio. "we think maybe a load of ink has run into the pool"
09:24
<MikeSmith>
foolip: yeah me neither I just want to know what expectations to write into the IDL test harness
09:28
<foolip>
MikeSmith: asked for clarification
09:33
<MikeSmith>
thanks will try to find a test for an interface that all browsers actually implement
09:37
<tobie>
Yeah. The only reasonable thing to spec is that this behavior is implementation specific. :p
09:38
<tobie>
MikeSmith: you're working on idlharness? That's awesome!
09:39
<MikeSmith>
foolip: for now you can look at the results of http://w3c-test.org/webstorage/idlharness.html though that obscures what exactly is being called
09:39
<zcorpan>
hmmmm. old apple docs still has <menuitem><menu>. chromium with experimental web platform features enabled now parses like firefox. i suppose i should give up and spec reality :-|
09:39
<MikeSmith>
tobie: I will try to pull out what actual call will cause it
09:39
<MikeSmith>
s/tobie/foolip/
09:40
<MikeSmith>
tobie: not working on it actively but instead just planning to update it to align with any spec changes or any places where it might not conform to the current spec
09:42
<foolip>
wait, do we have any reason do thing that the current non-interoperable state is required for compat? I thought it was just a case of nobody quite being motivated to care in any engine.
09:47
<annevk>
Hmm, Polymer apparently though it was fine to encourage everyone to use document.createElement(string, string)
09:47
<annevk>
Per the bug Ms2ger referenced earlier, so I guess we'll have to support that
09:54
<smaug____>
uh, this toString() handling
09:55
<smaug____>
I wish I understood why blink wants the behavior they have
09:55
<annevk>
smaug____: v0 of custom elements didn't use a dictionary
09:56
<annevk>
smaug____: and v0 appears to be leaking
09:56
<annevk>
smaug____: despite the promise from Google that changes would be possible due to usage of libraries and such
09:56
<MikeSmith>
foolip: actually as far as the test I guess I am just overcomplicating this
09:57
<MikeSmith>
foolip: you just do, e.g., (new Object()).toString.call(StorageEvent.prototype)
09:57
<MikeSmith>
and from Gecko you will see "[object StorageEventPrototype]" but from Blink you will see "[object StorageEvent]"
09:58
<foolip>
annevk: isn't that typeExtension extra argument in some spec?
09:58
<smaug____>
annevk: that is unrelated to my toString() complains
09:58
<annevk>
smaug____: oh sorry
09:58
<annevk>
smaug____: you're talking about the IDL thing
09:59
<smaug____>
annevk: but yes, we had to fix something in Gecko recently because someone was passing 2nd param to createElement
09:59
<smaug____>
annevk: yes
09:59
<smaug____>
was just reading bugmail
09:59
<smaug____>
and testing blink
09:59
<smaug____>
it makes really no sense
09:59
<annevk>
foolip: it's a dictionary these days
09:59
<smaug____>
I don't care too much, but the behavior what blink has now makes no sense
09:59
<foolip>
annevk: oh. I wonder if anyone considered whether changing that in Blink is actually web compatible
09:59
<annevk>
smaug____: for toStringTag you mean?
09:59
<smaug____>
and I don't understand the reasoning
10:00
<smaug____>
well, toString()
10:00
<annevk>
smaug____: I think the reason is that TC39 introduced @@toStringTag or whatever it's called, exposing toString() in some way
10:01
<annevk>
smaug____: so then the question was how that should work for IDL
10:01
<annevk>
Anyway, I don't really care strongly about it either way
10:02
<smaug____>
annevk: and that other thing... where is Polymer encouraging everyone to use createElement(string, string) ?
10:04
<smaug____>
hmm, I don't even know what happens when string is passed as a param which expects dictionary
10:04
<smaug____>
exception
10:06
<annevk>
smaug____: see the link to the polymer documentation
10:06
<annevk>
smaug____: if you scroll down a bit it's there
10:06
<smaug____>
ah, I need to look at the log
10:11
<smaug____>
annevk: ok, so spec should possibly allow string as param, and do nothing with it?
10:46
<Ms2ger>
Yeah
12:08
<Domenic>
foolip: MikeSmith: Blink cares and would prefer to follow the JavaScript pattern of [object Foo] instead of divergent [object Foo] + [object FooPrototype]
12:09
<Ms2ger>
Blink wanna ship it?
12:09
<Domenic>
We already have for like 5 releases
12:10
<Ms2ger>
Even better
13:11
<Sebmaster>
annevk: yes, saw it live even. Apparently they're not sure how to handle spec updates wrt semver yet, so this is going to be very interesting
13:18
<Domenic>
oh fascinating
13:18
<Sebmaster>
Personally I hope they follow properly, I wouldn't want to keep the slower lib maintained if it's not necessary
13:19
<Sebmaster>
But then browsers don't follow it quite either, so it'd be necessary for shims anyways :(
13:20
<annevk>
Yeah, Gecko is slowly getting there
13:20
<annevk>
It's hard to get anyone from Blink engaged
13:20
<annevk>
WebKit might a few minor tweaks, but nothing major
13:20
<annevk>
And I suspect Edge will be the last
13:21
<annevk>
E.g., someone has been implementing IPv4 parsing and canonicalizing in Gecko, which is nice
13:21
<foolip>
annevk: is this about the @@toStringTag thing, or is there another thing where you can't get any Blink reaction
13:21
<annevk>
foolip: this was about URL parsing
13:22
<annevk>
foolip: there's also something about the Origin header I've been waiting 9 months on or so now
13:22
<annevk>
foolip: it's not a continuous effort on my part though and the Blink side may have forgotten about it
13:22
<foolip>
Oh. I guess URL is hard because you need to figure out how to actually transition.
13:23
<foolip>
Also because the code itself doesn't live in Blink I suppose.
14:10
<zcorpan>
annevk: (and others going to tpac), interested in sharing airbnb?
14:55
<Domenic>
Do the fetch() web platform tests all live in service-workers/?
14:55
<Domenic>
nope, I'm just blind
15:12
<Domenic>
annevk: any ideas on how to test a Response's HTTPS state?
15:16
<annevk>
Domenic: hmm not really
15:16
<Domenic>
What is it used for anyway?
15:16
<Domenic>
I just know you and mkwst are very conscious to propagate it everywhere :)
15:18
<annevk>
Domenic: it's used to determine whether you get powerful API access
15:18
<annevk>
Domenic: secure contexts, basically
15:18
<Domenic>
Hmm
15:19
<Domenic>
Yeah not sure how that could be tested... I imagine it would only matter for Responses served from a service worker, which will always have a HTTPS settings object
15:19
<Domenic>
Oh maybe I could importScripts from an insecure origin
15:19
<annevk>
Yeah, think so
15:20
<Domenic>
No wait that doesn't work, still the same global
15:21
<Domenic>
annevk: https://github.com/w3c/web-platform-tests/pull/3452 FYI
15:58
<bradleymeck>
if I am reading https://url.spec.whatwg.org/ right `new URL('file://../..') should not normalize away the 2 `..` segments? https://url.spec.whatwg.org/#path-state 2. (noop due to https://url.spec.whatwg.org/#pop-a-urls-path file exception)
15:58
<bradleymeck>
is that correct?
16:00
<Domenic>
seems right, let's test the reference implementation...
16:01
<Domenic>
https://www.irccloud.com/pastebin/wgD0KlaR/
16:01
<Domenic>
That's not promising...
16:01
<Domenic>
(whatwg-url might have a bug, or the spec might be crazy, but that doesn't seem good in any case)
16:03
<Domenic>
bradleymeck: are you sure the predicate at the top of path state step 1 is true?
16:04
<Domenic>
I'd debug through the whatwg-url implementation to see what's going on
16:04
<bradleymeck>
i think `c is EOF code point or "/"` should be true since it consumes a `/`
16:05
<Domenic>
Hmm
16:05
<Domenic>
I thought c would be . here
16:06
<Domenic>
But yeah I can't really trust myself to get this right at a glance, and don't have time to dig into it in detail, so I dunno. It does seem possible the result is wrong. But as for what's going on I'd debug through whatwg-url to confirm.
16:07
<bradleymeck>
i can take a look because trying to figure out what import of a url would look like for node, thanks
16:09
<Domenic>
Hmm I thought you'd do `new URL(importSpecifier, "file://" + __filename)` for that
16:16
<bradleymeck>
node modules are not directly mapped to files though
16:16
<bradleymeck>
importSpecifier for "module" and "@scope/module" don't work I think
16:17
<Domenic>
Yeah true
16:18
<Domenic>
Anything that doesn't start with ./, ../, or / would be handled specially I think
16:20
<bradleymeck>
trying to figure out the exact mechanics of that but yea
16:22
<bradleymeck>
the idea in #node-dev was to resolve using a `new URL(new URL(importSpecifier, "node://"), "file://" + __filename)` basically
16:22
<bradleymeck>
but saw that importSpecifier was auto-normalizing without any way to turn it off
16:31
<annevk>
I did not know about https://news.ycombinator.com/item?id=11175258
16:31
<annevk>
Seems very exciting if it works out
16:32
<annevk>
Hundreds of FPS is what VR needs
16:35
<annevk>
bradleymeck: I think in HTML the model is that you first do syntax checks on the identifier
16:35
<annevk>
bradleymeck: afterwards you hand it to the parser
16:36
<bradleymeck>
they do prefix checking, which we will need to do
16:36
<bradleymeck>
but after that we are in new territory
16:37
<bradleymeck>
prefix of ./ ../ or / will definitely need to resolve against current script, modules need to use node_modules (suggested node:// scheme)
16:38
<bradleymeck>
but noticed that any non-file url in spec should auto-normalize which is a pretty no-go, so reading spec and saw this maybe bug?
16:56
<Domenic>
annevk: any reason sendBeacon spec does not use Fetch's keep-alive flag?
17:00
<annevk>
Domenic: prolly bug
17:01
<annevk>
bradleymeck: I don't really follow
17:03
<bradleymeck>
Domenic: https://github.com/jsdom/whatwg-url/pull/58 looked like a bug in whatwg-url, but unsure about that test that it makes fail, just gonna hand off here
17:04
<Domenic>
bradleymeck: wow nice find... the definition of "pop" is different than normal, that's awesome :-/
17:07
<annevk>
Domenic: ah, I guess I can replace pop with shorten
17:07
<annevk>
Domenic: it used to match, but then I had to add a special case, but I kept the term
17:07
<Domenic>
that would be nice
17:08
<annevk>
https://github.com/whatwg/url/issues/140
17:09
<annevk>
Domenic: https://aturon.github.io/blog/2016/08/11/futures/ might be of interest
17:10
<annevk>
Domenic: with streams they don't talk about backpressure at all, I wonder if that's a problem
17:15
<annevk>
Pffff https://github.com/Polymer/docs/issues/1705
23:16
<MikeSmith>
annevk: http://stackoverflow.com/questions/38907385/should-non-2xx-status-code-responses-include-cors-specific-headers