01:52
<MikeSmith>
https://twitter.com/qubyte/status/610968768760651776
01:52
<MikeSmith>
"Linters are like children... Very naïve, often insightful, and worth taking time and effort to explain things to."
01:54
<MikeSmith>
gde33: about your "I was curious if there have been proposals for html to get a nice standard outliner" questions at http://krijnhoetmer.nl/irc-logs/whatwg/20150616#l-762 yeah there have been proposals
01:56
<MikeSmith>
not necessarily for a way to directly render an outline, but at least for a way for Web content to get access to the outline as an object
01:57
<MikeSmith>
a method to get an outline object in script, either for the whole doc or for a subtree
04:02
<MikeSmith>
Hixie: I'm going to disable your W3C bugmail briefly while I mass-change some bugs in an uninteresting way
04:40
<MikeSmith>
Hixie: turned your W3C bugmail back on
04:55
<annevk>
So Safar when seeing a duplicate @ simply returns failure...
04:56
<annevk>
I keep writing Safar instead of Safari...
09:36
<mrflix>
I just ran into this problem: in a web-app UI we want the user to input a timeframe in hours in which to responde to a message. The timeframe should be hours up to 72. We’d love to use input#time but its limited to 24 hours max. Any suggestions where I can suggest a change in the standard?
09:45
<Ms2ger>
People seriously want to put HTML in WebApps
09:45
<Ms2ger>
?
10:05
<jgraham>
Ms2ger: One group to rule them all
10:06
<Ms2ger>
Not one I'll touch with a ten-foot pole
10:16
<annevk>
Rewriting the URL parser is somewhat more work than anticipated
10:49
<MikeSmith>
annevk: you doing the functional thing?
10:49
<annevk>
MikeSmith: no, just the rewrite to support relative URLs
10:49
<MikeSmith>
ah OK
10:49
<MikeSmith>
yeah had forgotten about that
10:50
<annevk>
"just"
10:51
<MikeSmith>
heh
10:51
<MikeSmith>
I wouldn't think it would be so much work
10:51
<MikeSmith>
well, relative to other stuff at least
10:51
<MikeSmith>
(no pun intended)
10:57
<hsivonen>
annevk: yay unit tests. I found a spec bug in the big5 decoder thanks to a unit test added for a Gecko bug you had filed.
10:57
<hsivonen>
annevk: I'll file a spec bug
10:58
<annevk>
yay
11:01
<annevk>
MikeSmith: hehe
11:01
<annevk>
MikeSmith: it's a lot of extra conditionals and states
11:01
<annevk>
MikeSmith: and some reorganizing of existing states
11:10
<JakeA>
annevk: fetch(request) - if request's redirect flag is set to "manual", do you remember why we set it to follow rather than return a redirect (perhaps with the location header opaque)
11:10
<hsivonen>
annevk: https://github.com/whatwg/encoding/issues/5
11:24
<JakeA>
(will create a github issue)
11:30
hsivonen
learns that Gecko's encoding converter API is underdocumented, too
11:31
<hsivonen>
looking forward to seeing a encoding conversion API that isn't underdocumented
11:32
<annevk>
JakeA: because fetch() does not allow setting it to manual
11:38
<JakeA>
annevk: https://github.com/whatwg/fetch/issues/66
12:27
<JakeA>
annevk: "analyzed" - do you mean by a security team?
12:27
<annevk>
JakeA: yeah
12:28
<annevk>
JakeA: last time exposing status was a no-go
12:28
<annevk>
JakeA: and redirects are a constant source of bugs
12:28
<annevk>
JakeA: both security and otherwise
12:28
<JakeA>
annevk: if that's the only objection here, I'll look into it. I thought it was the redirect location that caused the problems rather than just knowing there was a redirect, but I'll dig in
12:29
<annevk>
JakeA: making them more fragile is not something I'm very happy with
12:29
<annevk>
JakeA: well that was a issue we found and why the hole was closed
12:29
<annevk>
JakeA: there might very well be others
12:31
<JakeA>
annevk: thanks, I'll run it past people. I'm not against making it opaque btw.
12:32
<JakeA>
But a common pattern is hit url -> not logged in? -> redirect to other origin for login
12:32
<JakeA>
And we're breaking that
12:32
<annevk>
JakeA: I think one of the problems is what happens if you store a redirect in the cache
12:33
<annevk>
JakeA: what happens if you respondWith that redirect in a different context
12:33
<annevk>
that's all kinds of new vectors
12:33
<JakeA>
annevk: shouldn't be a problem, should it? Would act the same as if you returned Response.redirect(url)
12:33
<annevk>
well, except the latter is not privileged
12:34
<annevk>
and doesn't go to a secret URL
12:35
<annevk>
e.g. I store that redirect
12:35
<annevk>
then I create a DELETE request
12:35
<annevk>
and reply that redirect to that DELETE request
12:35
<annevk>
that'll fail cross-origin, but same-origin could still be problematic
12:36
<annevk>
at least, I hope that'll fail cross-origin...
12:38
<hsivonen>
annevk: It looks like I have push access to whatwg repos. to submit a PR to encoding, should I push a feature branch to the main repo or should I first create a fork under my own account?
12:39
<annevk>
hsivonen: main repo should be fine
12:39
<hsivonen>
annevk: ok. thanks
12:39
<annevk>
hsivonen: that used to trigger the Twitter bot, but it now filters for the master branch
12:55
<hsivonen>
annevk: OK. I created a PR.
13:00
<annevk>
hsivonen: thanks!
13:00
<annevk>
hsivonen: merged and stuff
13:07
<hsivonen>
annevk: thanks
14:51
<wanderview>
JakeA: do you have any thoughts on this one? https://github.com/slightlyoff/ServiceWorker/issues/709
14:51
<wanderview>
just wondering if I should try to get it in
14:56
<JoWie>
is that enforce https for new features still a thing?
15:03
<annevk>
JoWie: yes
15:08
<JakeA>
wanderview: replied!
15:09
<wanderview>
thanks!
15:09
<JoWie>
so would i need a localhost certificate ;)
15:10
<wanderview>
I'll try to get something implemented soon even though its not completely spec'd
15:10
<wanderview>
I think localhost should be exempted
15:10
<MikeSmith>
I think it is already
15:11
<wanderview>
for SW it is, but this is about adding a new restriction... so nothing is really specified... I don't think this mentions localhost in general: https://w3c.github.io/webappsec/specs/powerfulfeatures/#settings-secure
15:11
<MikeSmith>
「Certain origins are always potentially trustworthy. In particular, UAs MUST treat file URLs and URLs with hostnames names equivalent to "localhost" as potentially trustworthy.」
15:12
<MikeSmith>
https://w3c.github.io/webappsec/specs/powerfulfeatures/#is-origin-trustworthy
15:15
<wanderview>
ah, cool... thanks
15:15
<annevk>
JoWie: localhost is whitelisted
15:15
<annevk>
JoWie: did you not read the threads?
15:15
<JoWie>
i read it a long while back
15:15
<annevk>
wanderview++
15:15
<JoWie>
but i never heard from it afterwards
15:16
<annevk>
Better start ordering some certificates, or wait until September 14 I guess
15:17
<JoWie>
i rarely use https because you have to renew them
15:18
<JoWie>
i just hope i do not accidentally use one of the newer features and then only find out alter i need a cert
15:19
<JoWie>
or just bug people to modify their whitelist :p
15:21
<annevk>
JoWie: I recommend doing some reading then, something as basic as HTTP/2 requires it
15:23
<JoWie>
i thought http 2 had opportunistic ssl
15:34
<annevk>
Firefox implements that, unfortunately, but fortunately, nobody else does
15:48
<wanderview>
MikeSmith: annevk: does this address the case of http doc creates https iframe which creates https SharedWorker? seems SharedWorker doesn't have a strict parent to trace back through: https://w3c.github.io/webappsec/specs/powerfulfeatures/#settings-secure
15:50
<annevk>
wanderview: SharedWorker can have multiple parents...
15:51
<annevk>
wanderview: I guess using the initial parent or all parents would be an okay solution...
15:51
<annevk>
wanderview: I'd somewhat prefer all if that's not a major perf issue
15:51
<annevk>
wanderview: although I guess all is not really an option as that would make it change over time
15:51
<annevk>
wanderview: probably the one that created the SharedWorker then
15:51
<wanderview>
annevk: I think the issue with that is it could oscillate between trusted to untrusted to trusted again as docs attach/detach
15:52
<wanderview>
annevk: I think sicking has requested that SharedWorker not base anything on doc at all, since the docs can change over time
15:52
<annevk>
so you determine securedness on creation and store that
15:53
<annevk>
the initial doc can't
15:53
<wanderview>
annevk: so if something else for that origin has made SharedWorker already... then http doc->https iframe->SharedWorker works... but if nothing else has made the worker yet, then you get the restriction? that will probably be a bit magical for people to understand
15:53
<annevk>
and I'm okay with docs changing
15:53
<annevk>
I'm just saying that we set this bit on creation and keep it around until the sharedworker is destroyed
15:54
<wanderview>
annevk: I understand that... but from the call site in the http doc->https iframe the behavior will be inconsistent
15:54
<wanderview>
and I guess also lets the GC of SharedWorkers be observed
15:55
<wanderview>
annevk: I guess I can open an issue in the powerfeatures repo
15:55
<annevk>
wanderview: maybe sharedworker should simply not work in such a mixed environment?
15:55
<wanderview>
yea
15:56
<annevk>
wanderview: that would be the best way out here I think
15:56
<annevk>
wanderview: require a same-scheme parent chain
15:59
<wanderview>
annevk: I wrote this: https://github.com/w3c/webappsec/issues/406
16:04
<annevk>
https://blog.mozilla.org/luke/2015/06/17/webassembly/
16:48
<JonathanNeal>
https://github.com/jonathantneal/postcss-nesting @TabAtkins I hope I got it right. Getting the parent reference to work was a pain. Would you be able to dumb down how @media works for me? The spec doesn’t give an example.
17:29
<annevk>
"As another example, the mobile Web depends on your browser downloading font resources without checking the CORS headers (as per spec) — we’ve had to change to WebKit behavior in our mobile browsers to make the mobile Web work."
17:29
<annevk>
http://blogs.windows.com/msedgedev/2015/06/17/building-a-more-interoperable-web-with-microsoft-edge/
17:29
<annevk>
Well that's interesting
17:42
<annevk>
hsivonen: your latest big5 patch still mentions the GitHub issue
17:42
<annevk>
hsivonen: not a big deal I guess, but you might want to remove that reference locally
17:43
<Domenic>
meanwhile i feel like there was movement in chrome to try to enforce the cors header. not sure it went anywhere though.
17:45
<hsivonen>
annevk: I'll remove the reference if I respin the patch or, alternatively, in the encoder patch
17:49
<annevk>
Domenic: yeah, TabAtkins mentioned that at some point
17:53
<annevk>
MikeSmith: more URL tests: https://github.com/w3c/web-platform-tests/pull/1909
17:53
<annevk>
Sebmaster: ^
17:54
<annevk>
The URL Standard just had a major change landed. If you encounter bugs please tell me. It's likely...
17:55
<Sebmaster>
annevk: havent had time to integrate the changes yet and its kinda late
17:55
<Sebmaster>
but ill get to it tomorrow
17:55
<annevk>
Sebmaster: no rush
18:17
<TabAtkins>
annevk, Domenic: I'll ping Kenji about that again today.
18:18
<TabAtkins>
JonathanNeal: What needs explaining?
18:21
<JonathanNeal>
TabAtkins: I’m unsure what happens when @media is nested.
18:23
<TabAtkins>
The spec doesn't add any new @media nesting abilities
18:25
<JonathanNeal>
.foo { { @media (min-width: 768px) { /* so this is just gonna fail, right? */ } }
18:26
<TabAtkins>
Under current specs + Nesting, yeah.
18:27
<TabAtkins>
You don't need Nesting Block for at-rules anyway.
18:27
<TabAtkins>
(I'm still considering getting rid of the Nesting Block, and using plain & at the start, or a new @nest rule when it's not at the start.)
18:28
<TabAtkins>
So for most cases it'd be the same as Sass.
18:29
<TabAtkins>
Technically, we only need *anything* special to handle the case of a selector starting with a type selector.
18:30
<TabAtkins>
But I think that's too fine of a distinction to capture directly. The "starts with &" criteria is easier.
19:14
<JonathanNeal>
TabAtkins: are you thinking of refreshing the syntax?
19:20
<TabAtkins>
Yeah
19:37
<JonathanNeal>
How does the new version work, it requires & somewhere before the {?
21:02
<David17Obbatala>
hola
21:02
<botie>
niihau, David17Obbatala
21:05
<David17Obbatala>
niihau botie
23:42
<MikeSmith>
is this true? https://twitter.com/w3be/status/611306366758727681 (that browsers fall back to UTF-8 for unknown/supported values of charset; e.g., charset=utfX or charset=utf9)
23:42
MikeSmith
tries Chris Weber's tool at http://www.lookout.net/test/charsets/iana-charset-support/ to see
23:44
<MikeSmith>
hmm yeah, "utf16" -> UTF-8
23:45
<Domenic>
No way...
23:47
<MikeSmith>
far as I can see
23:48
<MikeSmith>
hyphens are important :)
23:56
<Domenic>
I wonder about the header vs. the meta tag