04:16
<MikeSmith>
TabAtkins: yeah I agree it would help for Fetch to have an explainer section. I guess I can make an attempt to write one myself if nobody else beats me to it
07:51
<annevk>
MikeSmith: TabAtkins: I missed the context for this, but I agree Fetch could use more language here
07:52
<annevk>
MikeSmith: TabAtkins: I always find it a little hard to justify working on such things while there's still a ton of incorrect implementer language around
08:14
<annevk>
I have a hard time not reading https://github.com/whatwg/dom/issues/102 as "Make Attr great again"
11:29
<nox>
annevk: That issue makes me very sad. :(
11:58
<zcorpan>
can we also disable features when document.write is used? :-D
12:05
<jgraham>
maybe we could just disable loading webpages when document.write is used :)
12:06
<jgraham>
Although the servo people claim that not supporting document.write is a zero-cost ad-blocker
12:06
<jgraham>
(this is not really true)
14:21
<smaug____>
ok, so I wonder how many specs are now inconsistent with http://w3c.github.io/webappsec-mixed-content/ :/
14:21
<smaug____>
this is a great mess atm
14:21
<smaug____>
oh well
14:21
<smaug____>
such is life :)
14:22
<MikeSmith>
solution: just set Content-Security-Policy: upgrade-insecure-requests and forget it
14:23
<annevk>
smaug____: only WebSocket is
14:23
<annevk>
smaug____: everything else should route through Fetch already which dispatches to that specification
14:24
<smaug____>
one day I'll read Fetch spec
14:24
<annevk>
smaug____: soon WebSocket will also route through Fetch and things will be wonderful
14:24
<annevk>
maybe not wonderful, but slightly better anyway
14:24
<smaug____>
so how is http://w3c.github.io/webappsec-mixed-content/ related to this all then?
14:25
<annevk>
smaug____: Fetch uses that specification to define the Mixed Content check
14:25
<annevk>
smaug____: step 5 of https://fetch.spec.whatwg.org/#main-fetch
14:26
MikeSmith
misunderstood; thought smaug____ was talking about the resources of the spec source
14:28
<smaug____>
MikeSmith: nah, just trying to figure out how many specs I need to review before reviewing a patch.
14:28
<smaug____>
looks like 3
14:41
<MikeSmith>
well that seems manageable
14:47
<annevk>
zcorpan_: if you want I can introduce such a default User-Agent value concept now
14:47
<annevk>
zcorpan_: I mean, I'll do it unless you see a problem
14:47
<zcorpan_>
annevk: sure
14:47
<MikeSmith>
today’s SO question where somebody had a problem that turned out to be due just to them not knowing that Fetch doesn’t send credentials by default http://stackoverflow.com/questions/35868312/django-1-9-ajax-form-csrf-token-403-error-csrf-cookie-not-set/35869947#35869947
14:48
<annevk>
MikeSmith: yeah, this safe default is tripping up many folks
14:48
<annevk>
MikeSmith: maybe I should've used XHR's default... so little feedback during the development of that API
14:50
<MikeSmith>
annevk: I think partly it just takes time for it to enter into the collective developer consciousness
14:51
<MikeSmith>
I mean, there are many more arcane/esoteric things about the platform that developers just sorta know
14:51
<annevk>
MikeSmith: hopefully, I know Mark Miller will be pleased with this default, although not everyone will see that as a good sign
15:41
<Ms2ger>
annevk, what's the "local-URLs-only flag" for?
15:42
<annevk>
Ms2ger: CSS/SVG
15:43
<annevk>
Ms2ger: they haven't really defined their fetching model yet, though, but this is what I think they'll need to define "SVG as image" and "SVG as gradient" and such
15:43
<annevk>
Ms2ger: although maybe they're a little closer now, but nothing to the level of detail I'd like
15:43
<Ms2ger>
I see, thanks
15:45
<Ms2ger>
annevk, is there a reason not to return the network error in step 2 of https://fetch.spec.whatwg.org/#concept-main-fetch ?
15:52
<annevk>
Ms2ger: yeah, async callers wouldn't know what happened
15:53
<Ms2ger>
Still trying to figure the control flow here
15:55
<annevk>
Ms2ger: main fetch returns for recursive and synchronous invocations; for async it ends up queuing stuff in the end
15:56
<annevk>
Ms2ger: it's not the best and I've tried to get input on the design, but as with most things I write I'm kinda left to figure it out on my own
15:56
<Ms2ger>
Yeah
15:56
<annevk>
Ms2ger: not sure if that's due to me or simply a result of the problems I tackle not being super interesting to others
15:57
<Ms2ger>
Or designing stuff like this being Hard(TM) :)
15:58
<annevk>
wanderview: oooh I see
15:58
<annevk>
wanderview: you'd wait all the way until the network
15:58
<wanderview>
yea, that was the idea
15:59
<wanderview>
try to make all the credential stuff work the same
15:59
<annevk>
wanderview: network or cache API?
15:59
<annevk>
wanderview: otherwise you're in trouble offline
15:59
<wanderview>
well, cookies don't go in cache API requests now
15:59
<annevk>
wanderview: no, but they get stored somewhere
16:00
<annevk>
wanderview: I guess these get stored too someplace
16:00
<wanderview>
I would have to think about Cache a bit
16:00
<annevk>
it might be fine, given that there is a credential store
16:00
<wanderview>
annevk: yea, but the cookie connection to the Request is automatic... here script is attaching the Credential object
16:01
<wanderview>
and that association of the Credential to Request should happen at constructor time
16:01
<annevk>
I think you might still want to pass it to request somehow and request might want to identify it will get those things attached
16:01
<annevk>
yeah
16:01
<annevk>
but it sorta makes sense to not do it straight away to avoid a bunch of the pitfalls
16:01
<wanderview>
annevk: I'll write some github issues this afternoon so mikewest can see them
16:01
<annevk>
sounds good
16:43
<Domenic>
annevk: want to wait for another look from aklein or shall I sanity-check compile and merge the structured clone fix?
17:13
<annevk>
Domenic: happy to wait
17:36
<Domenic>
annevk: turned out "create an element" was a really good idea, if only for deduping code between createElement and createElementNS. So yay.
17:37
<annevk>
Yay abstractions
17:40
<wanderview>
annevk: https://github.com/w3c/webappsec-credential-management/issues/11
18:08
<Domenic>
Why are we passing attribute namespace to attributeChangedCallback. It's so useless.
18:30
<smaug____>
Domenic: consistency with MutationObserver?
18:30
<smaug____>
just a guess
18:36
<Domenic>
smaug____: seems legit
19:40
<aklein>
Domenic, annevk: structured clone fix looks good now
20:30
<annevk>
aklein: ta, landed
20:31
<annevk>
aklein: I guess now we can work on the serialize/deserialize rewrite -_-
20:50
wanderview
writes the promise-returning withFrame(url) function a lot too...
22:23
<Domenic>
Starting to feel like the custom elements spec is something we can be proud of...
22:24
<Domenic>
Big remaining issues are nanotasks and parser
22:26
<Domenic>
https://github.com/w3c/webcomponents/issues/412#issuecomment-193935169 is also interesting, should be pretty easy once we figure out what we want though