08:41
<Ms2ger>
annevk, <annevk> Ms2ger: remind me Monday and I'll just do it, custom elements is making good progress :-)
08:41
<annevk>
Ms2ger: come on, I'm barely awake
08:41
<annevk>
Ms2ger: but yeah, I'll make some tea and get on that
08:42
<Ms2ger>
No hurry, but you asked to be reminded :)
08:42
<annevk>
Ms2ger: yeah, thanks :-)
08:44
<hemanth>
^_^
08:45
<annevk>
jgraham: could we make it so that "sign in" is "clicked" as some background process?
09:35
<jgraham>
annevk: No
09:36
<jgraham>
annevk: That doesn't work if you don't have a github account or haven't used critic before (you end up on a github login page for some unknown reason)
09:36
<annevk>
jgraham: what if we XHR it?
09:37
<jgraham>
It was the original design and I spent all my time explaining to people that ending up at a github page was expected behaviour
09:37
<annevk>
<iframe>?
09:39
<jgraham>
I guess the problem with that is that a) you won't actually end up logged in until you navigate and b) you don't want to keep sending failed login attempts on every page navigation for non-signed-in users
09:39
<jgraham>
Really the only problem here is that the cookie shouldn't be a session cookie
09:42
<annevk>
jgraham: what's a session cookie?
09:43
<annevk>
jgraham: very low Max-Age?
09:43
<annevk>
Oooh, having neither Expires nor Max-Age
09:44
annevk
learned about the persistent-flag today
10:11
<hemanth>
any work around to avoid a feedback on local audio streaming with webrtc?
10:13
<hemanth>
for example http://jsfiddle.net/gazq9nwy/ gives a horrible feedback, strangely works well on headphones...did try mediaStream.getAudioStreams()[0].enabled = false; wasn't useful, also tried setting the volume level to 0.7, no use...
10:18
<annevk>
hemanth: might just be your OS being crappy
10:19
<hemanth>
annevk, might be i'm on OSX heh heh
10:20
<hemanth>
https://code.google.com/p/webrtc/source/detail?r=5159 should have fixed it...but not really
10:21
<hemanth>
Darwin Kernel Version 14.0.0 to be specific
10:29
<hemanth>
annevk, there is no feedback on your machine?
11:24
<annevk>
hsivonen: bit unclear whether to add ms932 given the lack of web pressure
12:04
<JakeA>
annevk: What's the best error for "You're not allowed to open a window right now", security or invalidState?
12:05
<Ms2ger>
Why are you not allowed?
12:09
<annevk>
JakeA: TypeError all the things :-)
12:13
<JakeA>
Ms2ger: https://html.spec.whatwg.org/multipage/browsers.html#allowed-to-show-a-popup
12:14
<JakeA>
annevk: What's the rationale there? I mean, with the fetch spec it mean you don't have a DOM dependency
12:14
<Ms2ger>
RangeError :)
12:14
<JakeA>
but this is for a spec that already has a DOM dependency
12:15
<annevk>
JakeA: unless there's a use case for branching I fail to see the point to make an effort
12:15
<annevk>
JakeA: and DOMException doesn't really match what how JavaScript designed exceptions
12:16
<annevk>
s/what//
12:16
<JakeA>
annevk: So why TypeError and not just Error?
12:16
<annevk>
JakeA: that's what awb seems to default to
12:16
<annevk>
JakeA: and sometimes RangeError
12:17
<Ms2ger>
Anyway, I guess I'd InvalidAccess
12:18
jgraham
feels like if the answer is "TypeError all the things" the question was probably "what sucks about exceptions in js?"
12:26
<caitp>
there are lots of valid answers to that question though
12:31
<annevk>
Ugh, the people suggesting self-signed certificates are fine are many
13:12
<hsivonen>
annevk: yeah, it isn't clear what the right call on the ms932 thing is. you are the editor. :-) However, if the Encoding Standard doesn't add it, it looks pretty clear that Thunderbird will.
13:13
<hsivonen>
annevk: to the extent email deliberately wants to have extra labels, I think not having them everywhere consistently is sad
13:13
<hsivonen>
email having extra encodings like UTF-7 is another story
13:23
<annevk>
hsivonen: it's weird that per https://wiki.whatwg.org/wiki/Web_Encodings#Firefox Firefox never supported that label
13:24
<annevk>
hsivonen: I guess I'll wait a bit and if there's no negative feedback I'll add it
13:24
<annevk>
Also, www-tag...
13:24
<annevk>
More people arguing for OE
13:25
<annevk>
Ugh
13:32
<gsnedders>
hsivonen: time for an Email Encodings spec? :)
13:40
<hsivonen>
annevk: the comments on the Thunderbird bug suggest that we started doing less autodetection and previously it was caught by the detector
13:40
<hsivonen>
annevk: or something
13:41
<hsivonen>
annevk: more labels is better than more autodetection
13:41
<annevk>
hsivonen: perhaps Thunderbird always used the "Universal" detector?
13:41
<hsivonen>
annevk: IIRC that was only when you attached files--not for incoming email
13:41
<hsivonen>
annevk: but yeah, there was that lurking in the code!
13:42
<annevk>
hsivonen: however, adding a label has a problem we've seen before: Content-Type:text/html; charset=shift-jis; \n\n <meta charset=utf-8>
13:42
<hsivonen>
annevk: yeah. I don't know what the right call is
13:42
<annevk>
I'll add a comment to the bug, perhaps someone can supply data
13:42
<hsivonen>
also, I'm not happy about how much time pondering about edge cases like this takes
13:45
<annevk>
hsivonen: yeah, see e.g. blink-dev thread on navigator.vendor
13:45
<annevk>
hsivonen: there's some kind of inverse graph for the core primitives of the platform
13:46
<annevk>
hsivonen: which is probably why infrastructure does not get fixed, but papered over
13:56
<duncanw>
hello all, I write a tool that generates websites, for our japanese users I’ve been trying to figure out how to handle hiragana and katakana characters in the URL path part (i.e. the filename produced by my app), I understand browsers request percent escaped UTF8, and I’m pretty sure there’s a unicode normalization step (which makes sense), but don’t know what it is
13:57
<Ms2ger>
Doesn't url.spec.whatwg.org tell you?
13:57
<annevk>
duncanw: what part of the URL are we talking about?
13:57
<duncanw>
Ms2ger: wouldn’t look like it
13:57
<duncanw>
annevk: the path part
13:58
<annevk>
duncanw: there's no Unicode normalization going on there
13:58
<duncanw>
e.g. http://something.jp/ < japanese glyphs >
14:00
<annevk>
duncanw: end of https://url.spec.whatwg.org/#path-state explains what happens to the input
14:01
<duncanw>
annevk: um, I don’t see that fragment, https://url.spec.whatwg.org/#relative-path-state perhaps?
14:02
<annevk>
duncanw: yeah, sorry
14:02
<duncanw>
kk
14:04
<duncanw>
annevk: this explains why unicode normalization is necessary, and I can only assume browsers do it before encoding the path: https://tools.ietf.org/html/rfc5198#section-3
14:05
<annevk>
duncanw: why can you only assume?
14:05
<duncanw>
annevk: but I haven’t found anything official looking or even unofficial… perhaps I should dig in the webkit source
14:05
<annevk>
duncanw: you could just test it and see it's false?
14:06
<annevk>
duncanw: that RFC is immaterial when it comes to URLs or HTTP
14:06
<duncanw>
annevk: the RFC is irrelevant but it shows cases where something the user enters can have different code points for an identical glyph
14:07
<annevk>
duncanw: yes, that can happen
14:07
<duncanw>
annevk: and even unicode terminology considers them canonically equivalent
14:07
<annevk>
duncanw: sure
14:07
<annevk>
duncanw: you would have to implement Unicode normalization yourself though
14:09
<duncanw>
annevk: my app runs on Macs, and Cocoa has pretty good normalization… that’s not my problem, my problem is figuring out which normalization form to use. I was hoping to find something documenting exactly which one it is, if there is one. I assume it is either NFC or NFKC http://unicode.org/reports/tr15/#Norm_Forms
14:09
<annevk>
duncanw: which one what is?
14:10
<duncanw>
annevk: the unicode normalized form used by browsers
14:10
<annevk>
(if you want to avoid being thrown off bridges, use NFC)
14:10
<annevk>
duncanw: I just told you browsers don't do that
14:12
<duncanw>
annevk: well I have had an instance with german where a user entered filename had an u+umlaut, it was properly encoded as UTF8 by the browser but the file couldn’t be loaded from the server… if the browsers don’t do it then perhaps the server filesystem does
14:12
<duncanw>
annevk: I had “fixed” it by removing non-ascii characters, which clearly doesn’t work for CJK text
14:13
<annevk>
duncanw: so, you want to run Unicode normalization on the filename you get on the server side
14:14
<annevk>
duncanw: and you need to be careful since the filename might not be Unicode (if someone is tricking you or using some weird Linux variant)
14:15
<duncanw>
annevk: right, I don’t control the server so I guess I can only do normalization on the filename before sending it to the server...
14:16
<annevk>
duncanw: sure, I recommend implementing NFC or using the normalize method from ES6 (might not be implemented yet though)
14:17
<duncanw>
annevk: ok, thanks a lot for your help
14:22
<gsnedders>
browsers may do normalization on URLs entered by the user, but may just vary based on how the OS hands them the string from the input (which may or may not be normalised)
14:49
<duncanw>
gsnedders: thanks! You’ve seen this or know of browsers that do it?
14:50
<gsnedders>
duncanw: I know OSes do interesting stuff to text input. :)
14:50
<duncanw>
gsnedders: right well I guess I’m just wondering if you have a specific example that I might try to track down
14:51
<gsnedders>
duncanw: IIRC OS X does interesting stuff with composing characters as you type them, with interesting intermediatry values
14:53
<duncanw>
gsnedders: right, I write diacritics on a US keyboard layout by doing the composition (like opt-e plus e to get é), and I see something similar happening when entering japanese text
15:39
<zcorpan_>
annevk: the spec says shift-jis is a known label. is that intentional?
16:10
<annevk>
zcorpan: I confused with euc_jp