06:38
<annevk>
yhirano_: hey, do you want me to rebase and all that?
06:39
<annevk>
yhirano_: cause it looks good now, apart from that
06:46
<yhirano_>
annevk: yeah, I'll do
06:46
<annevk>
yhirano_: thank you
06:57
<annevk>
yhirano_: all good now?
07:04
<yhirano_>
annevk: i'm reading through now, so please wait for a minute.
07:05
<annevk>
sure
07:24
<annevk>
yhirano_: can you remove the "." from the commit title?
07:25
<annevk>
yhirano_: perhaps make the title "Fix #118: add Response.prototype.body"
07:27
<yhirano_>
annevk: got it >> "."
07:27
<yhirano_>
annevk: I plan to add example (https://github.com/yutakahirano/fetch-with-streams/#fetch-api) as a separate commit so i planned to close the issue after that
07:28
<yhirano_>
annevk: though it is possible to include the example change in this commit
07:28
<annevk>
yhirano_: new PR is fine
07:35
<yhirano_>
annevk: fixed the commit title and message
07:41
<mkwst>
annevk, JakeA: I'm around if you two still have concerns about the Fetch handling proposed in the CSP3 draft.
07:41
<mkwst>
Or we can chat later today. Whatever works.
07:43
<annevk>
mkwst: working on something else at the moment, but have you thought through whether it makes sense for the service worker to be able to set the policy of a synthetic response (or perhaps a non-opaque response it got from somewhere) similar to how a server can?
07:45
<mkwst>
annevk: I'd like to get this bit hammered out this week, but no rush to get it done today. If you're busy, we can chat later.
07:47
<mkwst>
annevk: To the specific question, the SW is the network, right? I don't see why it wouldn't be able to set headers on a response. There are particular headers that must take effect before the SW has a chance to screw with them (`clear-site-data` comes to mind), but I guess if the SW can create an entirely synthetic response, there's no real value in
07:47
<mkwst>
restricting its access to the CSP header.
07:48
<annevk>
mkwst: given that you patched HTTP-network fetch, it wouldn't have any effect
07:49
<mkwst>
annevk: Ah, I see the concern. Yeah. It makes sense that we'd need to parse the headers somewhere further down for that case.
07:49
<mkwst>
Not sure how to do that and preserve the option to effect `set-cookie` behavior, but I'll hack something together and ask you to take a look at it later.
07:50
<mkwst>
n:1
07:52
<annevk>
Affecting the set-cookie header sounds intriguing, I guess I should review that
08:02
<mkwst>
annevk: I guess I should write it up. Will eventually live in https://w3c.github.io/webappsec-csp/cookies/, which modifies cookies currently, but also needs to poke at Fetch.
08:06
<annevk>
mkwst: interesting
08:07
<annevk>
mkwst: https://w3c.github.io/webappsec-csp/cookies/#monkey-patching-rfc6264 should be a patch to Fetch afaict
08:08
<annevk>
mkwst: although I guess if you're going to do this on a per-cookie basis... hmm
08:08
<annevk>
mkwst: I guess in that case you'd have to do that
08:08
<annevk>
mkwst: although "none" could still be handled at the Fetch level
08:09
<mkwst>
We need to do that for `document.cookie = 'x=y';`. I guess we could do more granular `set-cookie` handling in Fetch, but if we're already doing it here for the JS interface, there's not much advantage in doing it twice.
08:09
<mkwst>
'none' could certainly be handled in Fetch.
08:11
<mkwst>
annevk: I'll try to spell it out in more detail today and send it over for you to skim.
08:11
<annevk>
mkwst: I would really prefer we just give an API to block document.cookie/domain
08:11
<annevk>
mkwst: HTTP-based API
08:11
<annevk>
mkwst: instead of granular control over what can be set
08:12
<mkwst>
annevk: Why?
08:12
<annevk>
mkwst: even with granular control you can still exhaust the store through XSS
08:13
<annevk>
mkwst: because both are bad APIs and should be avoided
08:13
<mkwst>
annevk: Yes. Both should be avoided. And we should give folks the ability to turn them off (and I plan to).
08:14
<mkwst>
annevk: But apps use them, for better or worse. Allowing them to lock down the usage to their specific needs seems like a way of allowing those apps to harden themselves, and doesn't seem like it adds substantial complexity.
08:14
<mkwst>
annevk: I'll write it up one way or the other, and we can argue about whether or not it makes sense. :)
08:15
<annevk>
mkwst: I'd rather always block those two and offer a new asynchronous API where you would get more granular access
08:15
<annevk>
but okay, will wait
08:16
<mkwst>
*shrug* I think it's better to spend our time getting CSP's integrations into shape, and then we can argue about what kind of new stuff to add on top. :)
08:16
<mkwst>
So when I said "today", I think I was lying. :)
08:31
<Ms2ger>
> Symbol.iterator is undefined for document.all
08:31
<Ms2ger>
._.
09:37
<Mikaela>
Hi, I am wondering `Error: Bad value Content-Security-Policy for attribute http-equiv on element meta.` which I get when validating my site. From what I have understood it's w3 standard, so why would it cause validation error? Or was that just invented by caniuse.com or Chrome? http://caniuse.com/#feat=upgradeinsecurerequests Validation results in question:
09:37
<Mikaela>
https://validator.w3.org/nu/?doc=https%3A%2F%2Fmikaela.info%2F
09:39
<Ms2ger>
http-equiv doesn't support all headers
09:39
<Ms2ger>
However, I seem to recall ongoing work to support CSP there
09:40
<Mikaela>
I see, so I can just safely ignore it for now?
09:41
<Ms2ger>
I don't think it'll do anything right now, so you can ignore it unless you actually want CSP to apply
09:42
<Mikaela>
thanks, It does apply on Chrome, no idea on the others
09:43
<Ms2ger>
mkwst, ^
09:43
<mkwst>
Ask MikeSmith.
09:43
<mkwst>
I added it to HTML, but I don't think the validator is updated yet.
09:44
<mkwst>
https://html.spec.whatwg.org/#attr-meta-http-equiv-content-security-policy
09:47
<mkwst>
Mikaela: That directive in particular works in Chrome today, and will work in Firefox 42+.
09:48
<Mikaela>
and is valid?
09:48
<Mikaela>
(that link seems to be loading forever here)
09:49
<mkwst>
Mikaela: Yes. The HTML spec is monstrous. It'll take forever to load, but once it does, you'll see that the `http-equiv` value is indeed specified.
09:49
<Mikaela>
thanks :)
09:49
<mkwst>
Mikaela: The validator hasn't been updated, which is reasonable, since the value just got added to the spec ~weeks ago.
09:50
<Ms2ger>
https://html.spec.whatwg.org/multipage/#attr-meta-http-equiv-content-security-policy is less monstrous
09:50
<MikeSmith>
Mikaela: yeah, sorry I've just not made time to update the sources of the HTML checker yet
09:51
<Ms2ger>
On an unrelated note, is there a way to write a css rule that's overridden by a preshint?
09:51
<Mikaela>
no problem, I was just curious as I was quite sure that it's valid
09:51
<Mikaela>
that less monstrous version works, thanks
09:52
<Ms2ger>
zcorpan, surely you'd know that ^ :)
09:53
<zcorpan>
Ms2ger: write it in your user stylesheet without !important
09:53
<zcorpan>
not very test-friendly
09:54
<raintrap_>
hey guys, any nice WYSIWYG editor you can recommend?
09:55
<Ms2ger>
zcorpan, yeah, no :)
09:55
<Ms2ger>
I don't want this test that badly
09:58
<zcorpan>
preshint vs UA style is not material to interop anyway
09:59
<Ms2ger>
I was trying to test whether ol type=decimal is supported
09:59
<Ms2ger>
But if it only affects people with weird user styles.. Meh
10:00
<annevk>
MikeSmith: http://dev.w3.org/html5/html4-differences/ redirects to a 404
10:00
<Ms2ger>
Is that bad? :)
10:01
<zcorpan>
Ms2ger: you can test <ul><li type=decimal> though
10:01
<annevk>
Ms2ger: someone emailed me and suggest I replace a link from my site to that location with some spam
10:15
<nox>
annevk: Sorry for the useless issue. :)
10:15
<annevk>
it's cool, I think I mistakenly reported that one too once
10:37
<MikeSmith>
annevk: hmm yeah lemme fix that
10:39
<MikeSmith>
ah good after upgrading to El Capitan it seems there's no longer an Apple-provided cvs command
10:39
<MikeSmith>
which is as things should be
10:39
MikeSmith
homebrews is
10:43
<MikeSmith>
annevk: fixed that redirect
10:43
<MikeSmith>
thanks for the heads-up
10:43
<annevk>
MikeSmith: brave new era of useful spam
10:48
<MikeSmith>
hah
10:49
<MikeSmith>
achievement unlocked, found a valid use case for spam
10:52
<MikeSmith>
in similar news, we found somebody actually prefers the IETF for something https://twitter.com/mikewest/status/651342403962601473
10:52
<MikeSmith>
will wonders never cease
10:54
<annevk>
lol, creating a new IANA registry
10:56
<mkwst>
Oh, the exciting life of me.
10:56
<mkwst>
The IETF's publication system is so simple. It puts the W3C's to shame.
10:57
<mkwst>
I think I've mentioned this before.
11:00
<MikeSmith>
yeah among other things, the result is great for weeding out the people who might ever want try to read some part a spec on a smartphone
11:01
<MikeSmith>
the lack of ability to include useful internal cross-reference also frees up editors from the nuisance of trying to fill their documents with them
11:13
Ms2ger
remembers when Hixie was told to stop publishing so often, because the publishmonkey couldn't keep up
11:44
<annevk>
Yeah, the W3C is averse to change, but the IETF is worse
12:22
<annevk>
philipj: hmm, now I'm thinking that perhaps navigation doesn't always discard
12:28
<philipj>
annevk: https://html.spec.whatwg.org/#navigate calls "unload a document" which calls "discard the document", but I didn't check for early returns or conditions or anything
12:29
<annevk>
philipj: only when it cannot be salvaged
12:29
<annevk>
philipj: so my patch probably breaks Mozilla's bfcache
12:29
<annevk>
good times
12:30
<jgraham>
Hmm?
12:30
<philipj>
oh, the spec actually doesn't just pretend that every navigation loads the URL from scratch?
12:30
<jgraham>
No
12:31
<jgraham>
Because that's observably different
12:31
jgraham
doesn't know the context btw
12:31
<philipj>
Right, but I have pretty low expectations in the area of navigation :)
12:31
<jgraham>
Well fair enough but it *tries* to be right
12:31
<philipj>
Fascinating
12:32
<jgraham>
So it gets things right like pageshow being different for actual loads vs history navigations
12:32
<philipj>
I guess that means annevk needs to understand how this works, and then I need to pretend to have understood the change :P
12:32
<jgraham>
What is "this"?
12:32
<philipj>
jgraham: when to exit fullscreen during navigation
12:32
<jgraham>
Oh
12:33
<jgraham>
Is that supposed to be always?
12:33
<philipj>
for top-level documents, yes, but not when navigating an iframe
12:34
<philipj>
I'm actually not sure what these restrictions achieve, though.
12:35
<jgraham>
Seems like it might be right to exit fullscreen after firing pagehide if the document is in a tlbc, or something?
12:35
jgraham
hasn't thought about it too much
12:44
<annevk>
jgraham: yeah maybe, though we also need to tackle browsing contexts closing
12:44
<annevk>
jgraham: if you close a browsing context, I guess history is not preserved?
12:47
<jgraham>
Well if you close a child browsing context, the join session history doesn't change. If you close a tlbc I think it all goes through the same steps to fire pagehide+unload+etc.
13:05
<annevk>
philipj: are you sure we shouldn't fire the event btw?
13:06
<annevk>
philipj: wouldn't it be better if the site could still perform animations and such before being navigated
13:07
<izhak>
Hi, guys! What is the status of the multipart responses in http 1.*, http 2.0 and current browsers?
13:07
<izhak>
I mean, how well is it supported in browsers as well as in protocols..&
13:07
<izhak>
?
13:28
<philipj>
annevk: I would think it's too late to do anything at that point, or how long could the animation be?
13:29
<annevk>
philipj: dunno, but there's other events firing at that point
13:30
<philipj>
annevk: an easy way to find out is to just use the incantations that will result in events being fired and see if implementors complain that it doesn't make any sense.
13:30
<philipj>
or more likely, if it doesn't make sense, the events just won't be fired anyway
13:31
<annevk>
hmm
13:31
<annevk>
We should actually know
13:33
<philipj>
Well, for documents that aren't salvageable and are being discarded, it doesn't make much sense to fire any events I think
13:34
<philipj>
but if you can navigate away from a fullscreen page and then back without a full reload, it would be very strange if the page still thinks it's fullscreen because no event has been fired
14:04
annevk
finds "If any event listeners were triggered by the earlier unload event step" and wonders how that is explained
14:06
<Ms2ger>
Not
14:07
<Ms2ger>
annevk, did we ever have dom explain the IDB "if any event listener threw an exception" pattern?
14:07
<annevk>
Ms2ger: nope
14:08
<annevk>
how many fricking hacks did people add?
14:08
<annevk>
IDB no less, I though there was no real theoretical reason to dislike that API
14:09
<annevk>
time for 😓
14:09
<annevk>
or maybe 😭 since it seems the floodgates on event listener hacks are open
14:11
<Ms2ger>
> Call for Review: W3C DOM4 Proposed Recommendation Published
14:11
<Ms2ger>
lol
14:16
<annevk>
wat
14:16
<annevk>
"This branch is 209 commits ahead, 120 commits behind whatwg:master."
14:16
<gsnedders>
So, uh, how many tests did they have to get to PR?
14:17
<annevk>
Ms2ger: another interesting data point for https://github.com/w3c/charter-html/issues/112#issuecomment-141880890 though I doubt anyone really cares
14:19
<annevk>
plh is on his solo mission to replicate all things WHATWG
14:20
<annevk>
but not doing a great job of it for the couple of things I checked
14:21
<wanderview>
can stylesheet @imports be cross-origin?
14:23
<ondras>
wanderview: that opens a potential xss vector, as you can verify whether a certain css rule was applied and thus if that rule has been defined, right?
14:23
<wanderview>
ondras: I don't know... I would be happy to hear its spec'd as same-origin only
14:26
<wanderview>
ah, here it is: "The origin of an imported style sheet is the same as the origin of the style sheet that imported it."
14:26
<wanderview>
although maybe that means its just changing the origin of the imported stylesheet? hmm
14:27
<wanderview>
oh, thats some css origin concept
14:28
<ondras>
http://jsfiddle.net/m4drt18k/
14:28
<ondras>
loading cross origin css apparently works
14:30
<ondras>
and the followup... http://jsfiddle.net/m4drt18k/1/
14:34
<annevk>
wanderview: yes they can
14:34
<wanderview>
yea... nice way to start the morning :-\
14:34
<annevk>
wanderview: CSS is full of SOP fail
14:35
<wanderview>
my easy bug fix is turning into a month-long endeavor
14:53
<caitp>
isn't that just the worst?
15:05
<gsnedders>
is it deliberate the Acid tests don't have any clear license on them?
16:43
<smaug____>
hmm, stable links to html spec would be nice
17:22
<annevk>
smaug____: what link broke?
17:27
<smaug____>
annevk: https://html.spec.whatwg.org/multipage/webappapis.html#processing-model-9:run-the-animation-frame-callbacks
17:30
<Domenic>
smaug____: so, I do have plans for this
17:31
<Domenic>
You'll be able to press "y" on your keyboard and your URL will mutate to something like https://streams.spec.whatwg.org/commit-snapshots/02e0a6431731da258d0dcad8b813c216c9f16d7c/#queuing-strategies (but for HTML of course)
17:32
<smaug____>
well, what I'd like to see is to avoid removing fragment identifiers from the spec if just possible
17:33
<smaug____>
but of course that isn't always possible
17:33
<Domenic>
hmm
17:55
<annevk>
smaug____: I'm guessing that fragment identifier was generated by dfn.js
17:55
<annevk>
or maybe Wattsi, not sure what's up
18:18
<Domenic>
yay URL!
19:37
<Domenic>
Can you not CC people by email on W3C bugzilla
19:41
<smaug____>
one needs to have a bugzilla account
19:41
<smaug____>
or perhaps you meant something else
19:42
<Domenic>
yeah
19:42
<Domenic>
that is annoying
19:43
<smaug____>
Domenic: how is that different to github or chromium's bug tracker?
19:43
<Domenic>
smaug____: neither of those have CC fields which list email addresses, inviting you to simply add another email address
19:44
<smaug____>
oh. Never seen it that way
19:45
smaug____
is annoyed by chromium's bug tracker requiring a Google account
19:45
<Domenic>
that is pretty bullshit :-/
19:46
<smaug____>
Domenic: well, it in practice reduces the likelihood I'd file blink bugs
19:47
<smaug____>
I do have google accounts but try to not use them
19:47
<smaug____>
(since google tends to track me so well)
19:47
<smaug____>
oh well
19:48
<Domenic>
I heard there was a header you could send </troll>
19:54
<annevk>
heycam|away: we need (undefined or type) stuff in IDL
19:55
<annevk>
heycam|away: on the input side it exists as optional type, but there's nothing for the return side
19:55
<annevk>
(if someone wants to file a bug on that, be my guest, I'll get to it eventually otherwise)
20:01
<annevk>
smaug____: feel free to file them through IRC proxy
20:03
<smaug____>
annevk: blink bugs?