00:28
<Hixie>
smaug____: (continuing discussion from yesterday about disabling buttons) no script libraries either
00:29
<Hixie>
smaug____: i just disabled all the addons except firebug and it still happens
00:29
<Hixie>
smaug____: only in firerfox
00:29
<Hixie>
smaug____: and it's a whole bunch of different unrelated buttons, too
00:29
<Hixie>
smaug____: can't seem to work out how to reproduce it though
00:37
<smaug____>
Hixie: Haven't heard of any bug reports about disabled buttons
00:37
<Hixie>
smaug____: yeah me either
00:37
<Hixie>
it has me baffled
00:39
<zewt>
an unusually targetted virus? heh
00:40
<Hixie>
smaug____: seems like it's probably https://bugzilla.mozilla.org/show_bug.cgi?id=592665 and https://bugzilla.mozilla.org/show_bug.cgi?id=654072
00:45
<smaug____>
Well, shift-reload doesn't keep form state
00:46
<smaug____>
Hixie: and you said shift-reload doesn't help
00:46
<smaug____>
so it is something new
00:46
<smaug____>
s/new/else/
00:47
<Hixie>
i might be mistaken about that. it's hard to say since it's intermittent.
00:47
<Hixie>
and i only see it some time after i did theload
00:47
<zewt>
if you're never disabling elements in the first place, 654072 shouldn't be relevant, i think
00:48
<Hixie>
some buttons get disabled, just not the ones that are getting mysteriously disabled
00:48
Hixie
wouldn't consider a button's disabled state to be "form state", btw
00:48
<Hixie>
i recommend not restoring disabled state :-)
00:48
<zewt>
that's pretty much verbatim what i just typed out in a comment on that bug :)
01:09
<GPHemsley>
annevk: Heh. Just noticed you're on the CC list for the bug 716422 I just filed. :)
01:13
<zewt>
gar, firefox console autocomplete is aggravating
01:14
<zewt>
needs to be tricked into *not* autocompleting when unwanted, heh
01:16
GPHemsley
has the same problem with Gmail search. FTW.
02:39
<bencc>
how does a websocket server should respond to a CONNECT request?
08:00
<annevk>
GPHemsley: not sure what that bug is about :)
08:01
<annevk>
GPHemsley: ooh, now I remember again
08:02
<annevk>
mebibytes ftw
10:12
<Ms2ger>
People who hadn't noticed it yet may be interested in http://www.w3.org/community/w3process/
10:49
<zcorpan>
"create small, clean, orthogonal specs"
10:52
<Ms2ger>
Clean specs?
10:52
<Ms2ger>
Let's outsource that to the XHTML2 WG
11:30
<hsivonen>
hmm. annevk isn't here.
11:31
<hsivonen>
it would be nice to make xhr.status say 200 for successful non-HTTP responses unless doing so breaks the Web
11:32
<Ms2ger>
Yeah, that's what I thought :)
11:32
<Ms2ger>
Also, Anne's off the grid from Wednesday
11:33
<hsivonen>
Ms2ger: oh ok. I was hoping to be able to report spec bugs via IRC log.
13:43
<annevk>
hsivonen: what does it report now for non-HTTP responses?
13:44
<annevk>
hsivonen: in general non-HTTP is not really thoroughly considered at the moment
13:45
<annevk>
hsivonen: my view is mostly that for non-HTTP protocols/schemes we should define how they work in an HTTP context because all APIs we have "assume" HTTP
13:45
<hsivonen>
annevk: reportedly, IE10 and Chrome say 200 while Opera and Firefox say 0
13:45
<annevk>
like for disk loads? or are we talking ftp?
13:45
<hsivonen>
annevk: blob: was the scheme that was tested
13:45
<annevk>
aah
13:46
<annevk>
hsivonen: http://dev.w3.org/2006/webapi/FileAPI/#processingModel
13:46
<annevk>
blob URLs already follow the model I want people to use
13:46
<annevk>
as they have been recently added I got that semi-addressed (not entirely pleased with the level of detail on HTTP headers yet)
13:46
<hsivonen>
annevk: and, yet, Firefox and Opera get blob: wrong
13:47
<annevk>
there's only so much I can do :)
13:52
<hsivonen>
annevk: you could add some informative notes to XHR to remind people that non-HTTP protocols should define their mappings to HTTP and give enough information for implementors to fill in the blanks when specs for non-http[s]: schemes fail to do so
13:52
<annevk>
where would be the best place to put that?
13:53
<hsivonen>
annevk: in the definition for each affected API
13:53
<annevk>
hmm
13:53
<hsivonen>
annevk: the definition of status should talk about 200 and 500
13:53
<hsivonen>
for example
13:54
<hsivonen>
and GetResponseHeader should talk about Content-Type
13:54
<annevk>
so status / statusText/ getResponseHeader / getAllResponseHeaders ?
13:54
<annevk>
well not just Content-Type, all headers are important
13:57
<annevk>
but this is also relevant to e.g. HTML
13:57
<annevk>
hmm
14:01
<hsivonen>
annevk: adding notes makes it a lot more likely that this gets implemented correctly
14:02
<annevk>
yeah I guess
14:02
<annevk>
it's just that I've tried to avoid redundancy in similar situations
14:05
<bencc>
does a websocket server need to handle HTTP CONNECT requests
14:05
<bencc>
or only a proxy needs to handle it if there is one?
14:07
<annevk>
doesn't the spec say?
14:11
<bencc>
annevk: http://tools.ietf.org/html/rfc6455#section-4.1
14:11
<bencc>
it talks about the CONNECT request but I don't understand who should respond to it and how
14:11
<zewt>
_holy hell_ /that spec/
14:12
<zewt>
what the hell are they thinking?
14:12
<bencc>
?
14:12
<zewt>
To _Establish a WebSocket Connection_
14:12
<zewt>
A client will need to
14:12
<zewt>
supply a /host/, /port/, /resource name/, and a /secure/ flag
14:12
<bencc>
zewt: I don't understand
14:13
<zewt>
... someone might want to let them know that you can actually underline and italic with HTML
14:14
<Philip`>
zewt: Why is that relevant, when they're not writing HTML?
14:14
<zewt>
that's html.
14:14
<zewt>
terrible html, to be sure
14:14
<Philip`>
It's plain text auto-converted to HTML
14:15
<zewt>
i don't care about the authoring path (and I doubt anyone but the editor does)
14:16
<zewt>
(and surely they could convert those)
14:16
<Philip`>
I presume "let them know" was referring to the editor, who cares that they're only able to write plain text
14:17
<bencc>
what about the proxy? :)
14:17
<Philip`>
(as opposed to alerting the IETF people who make the decision to stick with plain text as the canonical spec format)
14:19
<annevk>
zewt: dude, all IETF specs are plain text
14:19
<annevk>
zewt: tools.ietf.org just provides a nicer view of them, the "actual" version is http://www.ietf.org/rfc/rfc6455.txt
14:19
<zewt>
they already have plenty of transforms; it'd pretty much be s/_(.?*)_/<u>\1</u>/g
14:20
<zewt>
(eg. section numbers become links)
14:20
<zewt>
(is that the right syntax for non-greedy matching? i always forget, since I don't use it much)
14:21
<annevk>
bencc: this is what the client sends: http://tools.ietf.org/html/rfc6455#section-1.3
14:21
<annevk>
bencc: so that is what you need to deal with
14:22
<annevk>
zewt: dunno if that would work for all RFCs
14:22
<annevk>
zewt: the syntax in the WebSocket spec is not universal
14:22
<bencc>
annevk: what about this? CONNECT example.com:80 HTTP/1.1
14:22
<annevk>
bencc: I think that's only for proxies
14:22
<zewt>
annevk: i'm just surprised that my already more or less zero confidence in the IETF being able to write readable specs managed to drop yet another notch; but I guess I shouldn't be surprised
14:23
<annevk>
bencc: are you writing client or server software?
14:23
<bencc>
annevk: server
14:23
<bencc>
most of the requests works but recently I'm getting CONNECT requests which I don't know how to handle
14:23
<bencc>
my server directly gets requests with a proxy like haproxy
14:23
<bencc>
but maybe there is a proxy between the client and my server
14:24
<bencc>
I'm using port 8080
14:24
<Philip`>
zewt: .*?
14:25
<Philip`>
zewt: That'd be terrible when people write underscore_separated_names in their spec text
14:25
<annevk>
bencc: from my brief reading that sounds like a problem that's not your fault
14:26
<zewt>
it'd need adjustment for separators, etc
14:26
<Philip`>
At least if auto-linking makes a mistake, the reader can just ignore the link and it won't affect the interpretation of the text
14:26
<Philip`>
Removing _s and /s would be much more risky
14:26
<zewt>
(not going to try since I'm not actually interested in spending lots of time trying to convince the IETF to do something)
14:26
<bencc>
annevk: what do you mean and how should I handle it?
14:27
<annevk>
bencc: i.e. you should just conform to http://tools.ietf.org/html/rfc6455#section-4.2.1 and CONNECT does not so you should fail the request
14:27
<Velmont>
You don't have to kill the _'s either. can just do <u>_text_</u> and <em>/text</em> it helps with reading.
14:27
Philip`
thinks that would make it even uglier
14:28
<bencc>
annevk: what about 4.1 3?
14:28
<bencc>
annevk: "_Proxy Usage_: If the client is configured to use a proxy when.."
14:29
<zewt>
should you just handle CONNECT the same way a regular HTTP server would? (Apache returns 405 Method Not Allowed)
14:30
<zewt>
(aka I'm Not A Proxy)
14:30
<bencc>
what can cause a client to send me a CONNECT request with Connection Close header?
14:30
<annevk>
bencc: that's how clients are supposed to deal with proxies
14:31
<annevk>
servers != proxies
14:31
<bencc>
annevk: the proxy should handle this request without passing it to my server?
14:31
<annevk>
yes, the proxy is supposed to open a TCP connection to your server and let the client use that to connect to your server
14:31
<annevk>
as I understand it
14:32
<zewt>
yep
14:32
<bencc>
so this is a bug in FF or the proxy server?
14:32
<zewt>
it's probably something goofier than that; a proxy server that didn't swallow the CONNECT header would be totally unusable
14:33
<zewt>
unless FF with proxy+websocket is just broken (that'd be surprising, though)
14:34
<Ms2ger>
zewt, I believe that's a remnant from Hixie's HTML->plain text transformation, actually
14:36
<annevk>
bencc: I suppose it could be either
14:36
<zewt>
i think the IETF just hates people reading their specs and wants to give them headaches :P
14:37
<annevk>
bencc: both browsers and proxies are known to be pretty weird
14:38
<zewt>
i'm somewhat confused that this spec mentions HTTP proxy details at all
14:38
<zewt>
rather than ignoring it as a transparent detail at another layer
14:38
<annevk>
not really sure what the point of that is either
14:39
<annevk>
maybe because in this case only the initial connection is HTTP and the rest is not
14:39
<annevk>
the TLS-only WebSocket design was so much neater
14:42
<hsivonen>
does Web Socket work nicely with Apache and mod_jk these days?
14:42
<annevk>
jk?
14:42
<annevk>
just kidding?
14:43
<hsivonen>
annevk: I believe it's short for Jakarta
14:43
<zewt>
same thing I thought, heh
14:43
<hsivonen>
annevk: for connecting to a servlet container supposedly more efficiently than by proxying HTTP
14:46
<annevk>
hsivonen: thanks for dealing with the document.write bug btw
14:46
<hsivonen>
annevk: you're welcome
14:46
<zewt>
the "i ignore parts of the spec when i know better" one?
14:46
<annevk>
hsivonen: it seemed somewhat reasonable to at least consider it
14:46
<hsivonen>
zewt: that one
14:50
<annevk>
AryehGregor_: where does http://dvcs.w3.org/hg/editing/raw-file/tip/editing.html#selections state when the associated range becomes non-null?
14:52
<AryehGregor>
annevk, 1) addRange(), 2) "The user agent should allow the user to change the selection."
14:53
<AryehGregor>
The idea of "is initially null" is just that <!doctype html><script>alert(getSelection().rangeCount); alerts 0.
14:53
<AryehGregor>
This could probably be made clearer.
14:53
<AryehGregor>
It's supposed to be null until the user actually selects something, or script sets it.
14:54
<AryehGregor>
Basically, what the selection looks like after user interaction is entirely undefined.
14:54
<AryehGregor>
Probably at some point it would be good to define it -- e.g., if the user selects something and then that thing gets deleted, or the user clicks somewhere so the selection is cleared, does the range become null again?
14:54
<AryehGregor>
I'd think so.
14:54
<AryehGregor>
But it's not specified.
14:55
<AryehGregor>
. . . Does that answer your question?
14:56
<annevk>
I'm interested in case 2)
14:56
<annevk>
I think what was not clear to people is that changing the selection makes a Range object appear
14:56
<annevk>
so maybe that should be explicit
14:57
<AryehGregor>
Would you like to file a bug?
14:57
<annevk>
sure
14:57
<AryehGregor>
Right now I'm working on transforms stuff. I'll probably return to the editing spec for a while once I'm done with that.
14:57
<annevk>
k
14:57
<annevk>
https://www.w3.org/Bugs/Public/show_bug.cgi?id=15470
15:00
<AryehGregor>
Is there a Bugzilla component for CSSOM View, or just CSSOM?
15:01
<zcorpan>
MikeSmith: ^
15:01
<MikeSmith>
just for CSSOM
15:01
<MikeSmith>
I can create another one if you want
15:01
<AryehGregor>
I've been told CSSOM View currently doesn't have an editor, so . . .
15:01
<annevk>
I'd first coordinate with the editors
15:02
<zcorpan>
Nobody, is it OK if we create a bugzilla component?
15:03
<annevk>
euhm
15:03
<annevk>
didn't Glenn Adams sign up for both?
15:03
<annevk>
and Shane Stephens?
15:03
AryehGregor
doesn't know, only knows it still says you're the editor but you said you're not
15:03
<AryehGregor>
I assume that if no one has even bothered changing the editor in the spec, it's probably not very actively maintained . . .
15:04
<annevk>
the day I resigned from the CSS WG I was told someone was taking over so I assumed that would be immediate
15:04
<annevk>
often I assume wrongly in new scenarios
15:05
<AryehGregor>
Maybe I'll volunteer to edit CSSOM for a while. CSSOM interop is horrifyingly bad.
15:05
<zcorpan>
maybe they are responsible but haven't made any changes yet so didn't feel inclined to update the "editor" field
15:05
<AryehGregor>
And it's extremely frustrating sometimes.
15:05
<AryehGregor>
I had to spend lots of time hacking around it in my editing commands implementation.
15:05
<annevk>
Glenn explicitly asked me about the editor field weeks ago
15:05
<Ms2ger>
Enjoy dealing with The WG :)
15:06
<AryehGregor>
Is the CSSWG a pain to deal with?
15:06
<annevk>
it wasn't for me
15:07
<AryehGregor>
So far I haven't noticed it being a pain either.
15:07
<annevk>
I meant that in a different way; it seemed most people had different goals
15:08
<annevk>
to really fix the CSSOM, you have to fix the way the CSS model is defined
15:08
<AryehGregor>
How so?
15:08
<annevk>
because it's an API for the model
15:09
<annevk>
in theory defining an API is simple, you just say "return this bit from the model"
15:09
<AryehGregor>
What parts of the model aren't defined adequately?
15:10
<annevk>
but here you have to define some transformation on the model and return that, while not always being able to usefully reference the model
15:10
<annevk>
but maybe I was doing it wrong
15:10
<AryehGregor>
TBH, a lot of the stuff that bugged me was things like style="font-weight: bold" vs. style="font-weight: bold; " or such. That should be fixable without much trouble.
15:10
<annevk>
well e.g. the box model offset*/client/scroll* attributes
15:10
<AryehGregor>
Hmm, right.
15:11
<annevk>
property serialization is indeed something that needs fixing, but also depends on the model
15:11
<annevk>
e.g. is the model that bold becomes 600, or does it stay bold, etc.
15:11
<annevk>
I pointed this out a few times, but it was basically up to me alone to fix the model
15:12
<annevk>
and then push that through the WG somehow
15:12
<annevk>
I didn't really want to get into that
15:12
<annevk>
I was hoping more people would be on board and see the problem, but it didn't work out that way
15:13
<annevk>
as far as dealing with the CSS WG goes, I guess the issue is mostly that you can only get people's attention at a F2F meeting and the discussion is not always focused there
15:13
hsivonen
learns about https://developer.mozilla.org/index.php?title=en/application%2F%2Fhttp-index-format_specification
15:14
<hsivonen>
https://developer.mozilla.org/en/application%2F%2Fhttp-index-format_specification rather
15:24
<zcorpan>
another application of the html parser?
15:54
<matjas>
annevk: “doing that just that”
15:55
<matjas>
s/honers/honors/
15:55
<annevk>
yay for matjas
15:56
<annevk>
new weekly everyone: http://blog.whatwg.org/happy-2012
16:02
<annevk>
ooh, now 7,129 subscribers according to Google Reader
16:10
<AryehGregor>
Yay, I found a point where there's a notable lack of interop.
16:11
<AryehGregor>
For transform and transform-origin, percentages are supposed to be relative to the border box.
16:11
<AryehGregor>
Everyone does the actual transforms that way, but for getComputedStyle(), IE treats percentages as relative to the padding box, and Chrome 17 dev relative to the content box.
16:14
<annevk>
if you've been working on this for a while that's surprising
16:14
<AryehGregor>
Yeah, interop is quite good.
16:14
<AryehGregor>
I've heard 3D transforms are worse.
16:15
<AryehGregor>
Of course, transitions and animations are probably a mess, but those are going to be a real pain to test anyway . . .
16:37
<dglazkov>
good morning, Whatwg!
16:38
<AryehGregor>
So, so far there's perfect interop on the visual results of 2D transforms, except that Opera sometimes slightly exceeds my 1.5px error margin for rotations.
16:38
<AryehGregor>
The only lack of interop so far is computed values.
16:38
<GPHemsley>
annevk: Agreed.
16:39
<AryehGregor>
But I'm only written a 487-line test file so far.
16:39
<AryehGregor>
I'm sure I'll find something.
16:39
<annevk>
GPHemsley: ah, we're back to conversing about mebibytes :)
16:40
annevk
has been trying to write something coherent about XML5
16:40
<GPHemsley>
annevk: Yup. Gotta keep you on your toes. :)
16:40
<annevk>
Introduction: XML is hard.
17:04
<AryehGregor>
Now I found an Opera bug, but it's really table-related, not transform-related: height on a table seems to control the height of the border box, not the content box.
17:07
<jgraham>
AryehGregor: Seems to be already bugged
17:08
<AryehGregor>
k.
17:13
<gsnedders>
I presume WebKit doesn't do that anymore, then?
17:13
<gsnedders>
WebKit used the quirks box-model for tables for years
17:15
<AryehGregor>
Doesn't seem to.
17:16
<gsnedders>
That bug goes back to when I did WebKit stuff, so yeah.
17:16
<gsnedders>
Though it goes all the way back to KHTML.
18:45
<Ms2ger>
Why's this forums guy going on about Mosaic?
20:22
<annevk>
http://www.reddit.com/r/videos/comments/o98ty/pow_blinking_in_morse_code_to_spell_torture/c3ff18w reading about history on reddit is so much more fun
22:59
<Hixie>
of all the features i've specced over the years, none has had as many people trying to use it for unexpected purposes than appcache
22:59
<Hixie>
on www-html, someone is asking for how to embed a manifest inside an HTML file because they want to make their app a one-file app.
22:59
<Hixie>
...what?
23:10
<Hixie>
is Smylers ever on IRC?
23:48
<Hixie>
ooo, there's a CSS Product now in bugzilla
23:48
<Hixie>
is this new?
23:48
<Hixie>
looks like it's new as of a few months ago
23:49
<Hixie>
good to know
23:59
<MikeSmith>
Hixie: it was there for a while but we added some more components to it
23:59
<MikeSmith>
if you want any other components added, lemme know