06:32
SamB
wonders why you can't use an SSH key to authenticate with an http[s] server ...
08:40
<MikeSmith>
U+000C isn't whitespace in CSS?
08:41
<Ms2ger>
What is whitespace, really, if you think about it
08:56
<caitp>
how often are those control characters used these days in web documents?
09:08
<MikeSmith>
caitp: probably not at all. But the problem is, the HTML spec defines U+000C as a space character but CSS doesn't, and I'm writing code to parse CSS in HTML "sizes" and "media" attributes but I can't reuse the existing space-processing code I have for HTML
09:08
<caitp>
mm
16:29
<Ms2ger>
MikeSmith, is irc.w3.org still down?
16:29
<gsnedders>
down for me
16:38
<Ms2ger>
Hm, all of w3.org seems to be down
16:52
<Ms2ger>
Back up
16:53
<Ms2ger>
(IRC, at least)
20:00
<zcorpan>
MikeSmith: i thought it was whitespace in css also
21:04
<MikeSmith>
zcorpan: http://drafts.csswg.org/css-syntax/#whitespace-diagram and http://drafts.csswg.org/css-syntax/#whitespace
21:06
<zcorpan>
MikeSmith: drafts.csswg.org/css-syntax/#input-preprocessing
21:06
MikeSmith
looks
21:07
<MikeSmith>
zcorpan: ah OK good
21:07
<MikeSmith>
FF -> LF I see
22:04
<SimonSapin>
"User agents are encouraged to expose parse errors somehow." Does anyone care about this in practice, outside of validators? http://url.spec.whatwg.org/#parse-error
22:05
<SamB>
SimonSapin: you mean, like, the Console?
22:05
<SimonSapin>
SamB: yeah, for example, does any browser report non-fatal URL parse errors in their Console?
22:06
<SamB>
SimonSapin: I'm not sure; at the time when it would be doing that, the console usually gets cleared anyway
22:42
<zcorpan>
SimonSapin: i think several browsers log HTML parse errors
22:43
<SimonSapin>
zcorpan: There is at least one URL parse error that neither Chromium nor Firefox logs (this is how much testing I’ve done so far)
22:43
<SimonSapin>
(one URL parse error in the spec)
22:44
<zcorpan>
SimonSapin: the URL so spec isn't implemented yet is it?
22:44
<zcorpan>
-so
22:44
<SimonSapin>
zcorpan: I don’t know, maybe not
22:45
<SimonSapin>
I’m implementing it in Rust, but it seems so far that only annevk cares about non-fatal URL parse errors
22:46
<SimonSapin>
detecting parse errors slightly increases the code complexity
22:46
<SimonSapin>
but then I have no good way to log them, for a library that way be used outside of a browser
22:47
<SimonSapin>
Does html5lib log parse errors?
22:47
<SimonSapin>
gsnedders: ^
22:47
<zcorpan>
yes
22:48
<SimonSapin>
where do they go?
22:50
<gsnedders>
SimonSapin: it can
22:50
<gsnedders>
SimonSapin: I think it always does
22:50
<gsnedders>
SimonSapin: there's some API from the parser, maybe errors or something?
22:50
<gsnedders>
idk
22:51
<SimonSapin>
looks like HTMLParser().errors is a list of tuples
22:52
<SimonSapin>
gsnedders: any opinion on URL parse errors?
22:54
<gsnedders>
SimonSapin: what sort of opinions? :P
22:54
<SimonSapin>
should they be logged, or does nobody care?
22:54
<gsnedders>
out of scope
22:54
<gsnedders>
much like we don't report CSS parse errors
22:55
<SimonSapin>
I mean for a URL parsing library
22:55
<gsnedders>
It'd be nice to log them, at least optionally
23:06
<gsnedders>
SimonSapin: Also, note that people said nobody cared about exposing parse errors until HTML5 parsers suddenly started doing so
23:08
<SamB>
gsnedders: how else are they supposed to test that they had correctly *detected* them?
23:08
<gsnedders>
there's no requirement to detect them
23:08
<gsnedders>
just handle them correctly
23:09
<gsnedders>
and handling can be tested by checking the output DOM
23:09
<SamB>
yes, but it makes it harder to tell whether or not you keeping your branches straight if you don't report them ;-)
23:09
<SamB>
+'re
23:10
gsnedders
only very briefly had anything to do with any browser impl
23:10
<gsnedders>
very very briefly
23:28
<SimonSapin>
SamB: I’m not very familiar with the HTML parser, but I believe it’s similar to URL: the spec defines what implementations should do in all cases. Additionally, in some cases, it says "this is a parse error". Which in itself does not affect the implementation (except if it chooses to log such errors), it only expresses author requirements
23:28
<SimonSapin>
a validator would check author requirements
23:28
<gsnedders>
SimonSapin: browsers log parse errors when the console is open
23:29
<gsnedders>
SimonSapin: or at least WebKit, Blink, and Presto do. I think Trident too.
23:29
<gsnedders>
I'm pretty sure Gecko logs them, um, somewhere.
23:29
<SamB>
SimonSapin: yes, well, I assume it's a lot easier to verify that the parser logic has been accurately translated into code if you can see those parse errors ;-)
23:30
<SamB>
or at least to sanity-check it
23:32
<SimonSapin>
hum, I wouldn’t say "a lot easier"