00:28
<abarth>
i think the main thing we want is the autosizing
00:28
<abarth>
which I guess is what you mean by "shrink wrapping"
00:28
<Hixie>
right
00:28
<abarth>
disabling borders by default makes sense too
00:28
<Hixie>
do you know of anyone outside chrome who wants to implement this?
00:29
<abarth>
(one sec)
00:31
<abarth>
ok, back
00:32
<abarth>
there are definitely other folks at Google who would like us to ship seamless
00:32
<abarth>
and to have crossorigin support
00:32
<abarth>
I don't know where it is on the priority list for other browser vendors
00:34
<abarth>
the list you have above looks reasonable if you assume that you already have a fully implementation of seamless
00:35
<abarth>
if you're looking to lower the activiation energy for implementors, you might want to focus on just the core behaviors
00:36
<Hixie>
is there an activation energy problem on this front?
00:43
<Hixie>
implementors - if you see anything on this that you want implemented, please do fill in the table or let me know: http://wiki.whatwg.org/wiki/New_Features_Awaiting_Implementation_Interest
00:54
<othermaciej>
Hixie: I'm surprised to learn that Dean Jackson works for Google
00:54
<othermaciej>
how did he manage that and get <dino⊙ac> as his email?!?
00:56
<othermaciej>
(fixed btw)
01:00
<Hixie>
othermaciej: oops, i should proof-read the table :-P
01:26
<nessy>
Hixie: I'm really happy about that new proposal for cue positioning!
01:41
<GPHemsley>
It doesn't make sense to set up a selector for credit card companies.
01:41
<GPHemsley>
There are too many and the names could change.
01:48
<GPHemsley>
I'm not sure that deleting wiki pages is the best course of action for things that are no longer applicable.
01:50
<smaug____>
What is FormData in http://wiki.whatwg.org/wiki/New_Features_Awaiting_Implementation_Interest ?
01:50
<smaug____>
Something else than http://xhr.spec.whatwg.org/#interface-formdata ?
01:51
<smaug____>
I guess just additional features
06:11
<zcorpan>
Hixie: my recollection from the FOMS conference was that it's more OK to change the width than to change the position
06:11
<Hixie>
k
06:11
<zcorpan>
Hixie: so at x=50%, widht width=50%?
06:11
<Hixie>
does the proposal in the bug seem ok?
06:11
<zcorpan>
s/?//
06:11
<zcorpan>
i'll have a look
06:12
<heycam>
zcorpan, thanks for your scoped style test suggestions yesterday
06:12
<zcorpan>
heycam: np
06:23
<zcorpan>
Hixie: hmm. my recollection might have been backwards. a different case - align:middle position:0% size:50% - was expected to respect the width and change the position (or possibly have position 0% mean the anchor point is at 0% of the box)
06:23
<Hixie>
align:middle position:0% size:50% would presumably have text centered in a box covering 0% to 50% from the left, no?
06:24
<zcorpan>
yes, that's the expected result, but not the result of the current spec, iirc
06:24
<Hixie>
forget the current spec :-)
06:24
<Hixie>
look at the proposal in the bug from today
06:24
<zcorpan>
yeah, haven't got that far yet
06:24
<Hixie>
:-)
06:24
<zcorpan>
long bug is long
06:25
<Hixie>
20037?
06:25
<Hixie>
oh yikes, it has lots of comments
06:27
<zcorpan>
i read a dup first to refresh my memory
06:27
<zcorpan>
(18501)
06:29
Hixie
hates the "reply" link in bugzilla
06:29
<Hixie>
in other news, why are facebook sending me updates about their policy when I DON'T EVEN HAVE AN ACCOUNT
06:29
<zcorpan>
align:right position:10%
06:29
<zcorpan>
should be right-aligned with the right edge 90% across from the left
06:30
<zcorpan>
Hixie: what's the reasoning for that one?
06:30
<Hixie>
what would you expect?
06:31
<zcorpan>
i think i'd expect the left edge be 10% from the left
06:31
<Hixie>
why?
06:32
<Hixie>
(i don't disagree, just curious if you have a reason)
06:32
<zcorpan>
because position should do the same thing regardless of alignment
06:32
<Hixie>
hm, not sure i agree that that should be an invariant
06:33
<Hixie>
one reason to design it the way that 0% starts at the alignment edge (though not a reason to expect align:right position:10% on the right) is that it means that align:right size:10% will be on the right, which people seem to expect
06:33
<Hixie>
(if we default position:0%)
06:36
<zcorpan>
position could have an 'auto' default that makes right-aligned cues by default have the right edge at 100%, or some such
06:38
<Hixie>
auto values are just a pain
06:42
<zcorpan>
Hixie: i thought what was considered confusing was that the different settings depend on each other. the new proposal still does that
06:43
<Hixie>
different people have different ideas of what is confusing
06:43
<Hixie>
but i don't really see how the values can't depend on each other
06:43
<Hixie>
at least, not if we want it to have a reasonable number of settings and not have to specify them all all the time
06:47
<zcorpan>
maybe we should have 'left' and 'right' settings instead of 'position'
06:47
<zcorpan>
both default to 0
06:47
<zcorpan>
and work like absolute positioning in css
06:48
<zcorpan>
(or 'start' and 'end' if it should be flipped for rtl)
07:00
<Hixie>
zcorpan: i dunno, i think most people are more likely to think in terms of width than position
07:05
<zcorpan>
i've commented on the bug
07:06
<zcorpan>
Hixie: also see comment 0 in 18501 ("It seems to me that what authors needs the ability to specify is:")
07:09
<Hixie>
yeah i looked at all the dupes pretty carefully
09:02
<hsivonen>
can someone remind me how an HTTP client determines if a response entity body exists?
09:05
<hsivonen>
in particular, how can a client differentiate between zero-length response entity body and the absence of a response entity body?
09:15
<hsivonen>
https://svn.tools.ietf.org/svn/wg/httpbis/draft-ietf-httpbis/latest/p1-messaging.html#message.body suggests there's no other way than looking at the request method and response code
09:16
<hsivonen>
yay for not speccing what the client should do if a message body is present when it MUST NOT be present
09:29
<annevk>
Hixie: I think I need some more concrete details on the event retargeting stuff
09:44
<annevk>
Hixie: I added a proposal to the bug
09:48
<annevk>
hsivonen: dude, cannot expect anything else from a spec written by jreschke really
09:48
<annevk>
hsivonen: in rare cases you can convince mnot and then they have to play ball, but other than that it's riddles all the way down
09:49
<jgraham>
A riddle inside a mystery inside an engigma inside a RFC?
09:49
<jgraham>
*enigma
09:50
<annevk>
http://wiki.whatwg.org/wiki/New_Features_Awaiting_Implementation_Interest is a great idea
09:50
<annevk>
jgraham: yeah, waiting for gsnedders to solve it
09:59
<annevk>
marcosc: feel free to update the wiki
10:00
<annevk>
marcosc: also, @WHATWG is not really an entity, it just shares info, I'm kinda hesitant to reply from it
10:08
<zcorpan>
the best part about @supports is that you get to repeat all your vendor-prefixed declarations once more.
10:09
<zcorpan>
now cue a vendor-prefixed impl of @supports and we're home
10:10
<Ms2ger>
fantasai++
10:12
<kennyluck>
What happned?
10:14
<Ms2ger>
XKCD happened
10:15
<darobin>
odinho: https://w3c-test.org:444/ is now open for you to have all sorts of fun with
10:15
<MikeSmith>
darobin: so what did the problem turn out to be?
10:15
<MikeSmith>
or rather, the cause of the problem?
10:15
<darobin>
MikeSmith: it was just that port 444 was not open
10:15
<darobin>
it's what I thought, I just wanted to check that it was okay to open it before doing so
10:16
<MikeSmith>
I mean the OPTIONS problem
10:16
<darobin>
oh, that we haven't solved :(
10:16
<MikeSmith>
ah
10:16
<darobin>
I can't for the life of me figure out what's going on
10:17
<odinho>
darobin: Okay, can add those two tests back then, and maybe plan some moar fun later ;-)
10:17
<darobin>
mmmm, actually I have an idea
10:18
<annevk>
FWIW, I'm likely to remove 'If the "user:password" format in the userinfo production is not supported for the relevant <scheme> and url contains this format, throw a "SyntaxError" and terminate these steps.' from XHR soonish
10:18
<annevk>
oh, and I guess I'll make XHR depend on URL
10:21
<MikeSmith>
darobin: so what's different about the port 444 server?
10:23
<zcorpan>
i hope the only difference is the different port number :-)
10:23
<zcorpan>
(useful for origin tests)
10:23
<MikeSmith>
ah OK
10:24
MikeSmith
reads the logs from yesterday
10:24
<darobin>
MikeSmith: the 444 is just for CORS testing for SSL, odinho wanted it
10:24
<MikeSmith>
了解
10:24
<darobin>
it's a separate problem from OPTION
10:24
<MikeSmith>
ok
10:25
darobin
cries a little
10:25
<darobin>
I hate Apache
10:32
<zcorpan>
given a worker like this: postMessage(1); close(); postMessage(2);
10:33
<zcorpan>
how many messages should be received on the other end?
10:33
<darobin>
sigh, our version of Apache is too old to support trace-level logging...
10:44
<zcorpan>
i think the answer is 2
10:46
<annevk>
zcorpan: depends on the queuing :)
10:47
<zcorpan>
annevk: can you elaborate?
10:47
<annevk>
zcorpan: I don't know how this part of workers work, but I assume these put all something on a queue of some kind
10:48
<annevk>
works*
10:48
<zcorpan>
if you don't know, you're not helping :-P
10:51
<annevk>
zcorpan: actually seems the answer is 0
10:52
<zcorpan>
annevk: why?
10:53
<annevk>
I'm having a hard time finding the actual definition of postMessage, but if that queues tasks, and close() nukes all existing tasks and prevents further tasks to be queued, seems like 0 is the answer
10:54
<zcorpan>
the confusing part is that there are several event loops here
10:54
<zcorpan>
and neither close() nor postMessage() are clear on which they refer to
10:54
<annevk>
oh, postMessage() puts tasks on the other end
10:54
<zcorpan>
yes
10:54
<annevk>
then you're correct
10:55
<annevk>
but the spec could be a bit more explicit about task queues indeed
10:55
<annevk>
and also provide a forward reference to the actual postMessage method that is invoked
10:55
<zcorpan>
thanks. i'll file bugs
10:56
<annevk>
I was right though, depends on the queuing :p
12:09
<AryehGregor>
annevk, marcosc, https://www.w3.org/Bugs/Public/show_bug.cgi?id=12296 and https://www.w3.org/Bugs/Public/show_bug.cgi?id=12220
12:12
<annevk>
knew it
12:15
<hsivonen>
annevk: http://lists.w3.org/Archives/Public/www-archive/2011Dec/att-0021/encoding-labels.html contains at least one multibyte encoding
12:16
<hsivonen>
annevk: and it claims no one supports it
12:16
<hsivonen>
annevk: would the testing methodology have resulted in an indication of non-support when an encoding is, in fact, multibyte?
12:16
<annevk>
which one?
12:17
<hsivonen>
t.61-8bit
12:17
<hsivonen>
annevk: multibyte according to http://std.dkuug.dk/i18n/charmaps/T.61-8BIT
12:19
<annevk>
hsivonen: I suppose that could have happened, yes
12:20
<annevk>
hsivonen: though note that per http://wiki.whatwg.org/wiki/Web_Encodings Gecko is the only one to support that encoding
12:20
<annevk>
oh, IE might have support too under a different label
12:42
<annevk>
funny how this stuff works: http://lists.webkit.org/pipermail/webkit-dev/2012-January/019182.html
12:42
<annevk>
I'm pretty sure that group at Adobe is basically Vincent Hardy, who has been asking for this for a long time
12:44
<annevk>
he might even have done the original progress event spec in SVG, dunno
12:44
<annevk>
(which was intended for image downloads and such)
13:45
<annevk>
So I wanted to fix that URL bug in XHR, but I wonder if maybe the design in the URL spec should be tweaked a bit more
13:46
<annevk>
Instead of having a "fatal error flag" parsing would just return failure and URLUtils would not have an associated URL object
13:51
<annevk>
yup :)
13:55
<gsnedders>
jgraham: Waiting for me to solve what?
14:04
<darobin_>
jgraham: you probably want to chat with this guy https://twitter.com/graouts/status/271615076023287809
14:05
<darobin>
(he's a friend and generally a good guy)
14:30
<darobin>
good catch odinho
14:31
<odinho>
darobin: Only generally? :P
14:31
<darobin>
is anyone ever more than generally ;-)
15:14
<jgraham>
darobin: You are right, I do
15:15
<darobin>
jgraham: if he doesn't show up on IRC and you'd like a way to ping him outside Twitter just ask
15:15
<darobin>
(if you have a Twitter account, I don't know it)
15:16
<jgraham>
(I have an unused twitter account)
15:16
<jgraham>
Also, I commented on github
15:17
<annevk>
fwiw, new URL can now sometimes fail
15:17
<zewt>
are there any ctors on the platform that do that?
15:19
<zewt>
(failing instead of being in an error state does seem to make more sense as an interface)
15:27
<annevk>
zewt: you mean like XMLHttpRequest?
15:28
<annevk>
zewt: or maybe like new Image() if the argument cannot be coerced to an integer?
15:29
<zewt>
when can xhr's ctor fail? (webidl dispatch doesn't count, that's before it runs)
15:29
<zewt>
(note i'm not objecting, just wondering if this is a new thing or not)
15:29
<zewt>
gah i accidentally loaded the w3's copy of xhr2 and hooooooly shit they fucked it up
15:29
<zewt>
NO TEST COVERAGE
15:30
<zewt>
my eyes are bleeding
15:30
<annevk>
oh wait, XHR doesn't fail, doh
15:30
<annevk>
open() fails
15:31
<zewt>
i find it fascinating that they removed all links to the real spec (used to be the ED link went to yours), yet "participate" still points here
15:32
<annevk>
I found it rather odd to be "former editor" despite having written all the text
15:32
<annevk>
I commented on that here http://lists.w3.org/Archives/Public/public-webapps/2012OctDec/0538.html but thus far nobody seems to care
15:33
<zewt>
it's also fascinating that it apparently takes three people to copy one person's stuff
15:34
<annevk>
finally found an example, Worker throws, for exactly the same reason (URL cannot be parsed)
15:35
<annevk>
Worker can also throw for undisclosed security reasons
15:36
<annevk>
EventSource too
15:36
<annevk>
WebSocket too
15:40
<zewt>
hmm
15:41
<zewt>
if new URL(url.href).href always round-trips, then in principle the ctor base argument could just be URLString, but I guess it doesn't matter
15:42
<annevk>
what is a URLString?
15:43
<zewt>
a will it blend
15:43
<zewt>
DOMString
15:44
<annevk>
I'm not sure I follow. I think the main remaining annoyance is changing protocol
15:45
<annevk>
I'm sort of wondering if setting protocol should only work as long as you don't change relativeness.
15:45
<annevk>
So if the current scheme is a relative scheme it can only be another relative scheme, and if the current scheme isn't, it can only be a non-relative scheme
15:46
<annevk>
that way you keep the invariants and don't need something weird like the "relative flag"
16:05
<annevk>
In other news, I'm hitting more cases that at some point need to be defined in "Fetch"
16:20
<odinho>
zewt: I think linking up tests to the spec page is a very nice idea. I hope it's the styling you're mostly worried about :-)
16:20
<annevk>
oh hey hallvors!
16:20
<annevk>
hallvors: so I'm planning on making another change to XHR
16:20
<annevk>
hallvors: currently it's either same-origin or cross-origin
16:21
<annevk>
hallvors: I think we should put some checks on cross-origin too, for either http/https scheme in the request URL
16:21
<annevk>
hallvors: and then have "otherwise" be a case where it's just a network error
16:21
<annevk>
hallvors: thinking about it more, I guess that's another thing that "Fetch" should handle
16:42
<annevk>
Hixie: should http://xhr.spec.whatwg.org/#dom-xmlhttprequest instead read "set document to the active document of the global object on which xhr's interface object is located"?
16:42
<annevk>
hallvors: ^^
16:54
<annevk>
odinho: Velmont: you around?
16:55
<annevk>
odinho: Velmont: is the "userinfo" bit from http://fetch.spec.whatwg.org/#redirect-steps tested? I'm thinking it's bogus
17:08
<odinho>
annevk: There was some, but xhr tests are kinda broken.
17:09
<odinho>
Ah, fetch - not xhr :P Hm. They should be.
17:10
<annevk>
odinho: so you have Location: http://test:test⊙tc/ somewhere in there which is valid except for the test:test part?
17:11
<annevk>
hmm
17:12
<odinho>
No. They are not I have some started things lying on T server at Opera, but I never actually moved them anywhere because I hit some snags I think. Long time ago now. Never thought more about it :P
17:13
<odinho>
Just that small thing might be easy to add though.
17:23
<annevk>
well, I might want to kill it
17:23
<annevk>
so I'd prefer it if you don't :)
17:25
<odinho>
annevk: http://test.s0.no/w3c-tests/webappsec/tests/cors/submitted/opera/staging/redirect-userinfo.htm
17:25
<odinho>
^_^ Opera implements it.
17:25
<odinho>
WebKit also passes. (if the test is correct though)
17:26
<odinho>
Firefox times out.
17:26
<odinho>
IE10 passes.
17:27
<annevk>
k
17:27
<odinho>
So pretty god coverage. 3/4 engines do it.
17:27
<odinho>
(with the above caveat)
17:28
<annevk>
ah yeah, redirects only work for simple requests, so yeah, that seems good then
17:28
<annevk>
Fetch is gonna be one complicated algorithm :/
17:28
<annevk>
(though I guess that's also kinda fun)
17:31
<odinho>
(it doesn't really have to though ... but ohwell :P)
17:32
<annevk>
odinho: what doesn't have to?
17:33
<odinho>
Added some more tests to it. They're more to do with actual "what is userinfo" so I guess some of them are bad. Opera kills requests with //user:pass@ and //user:@ but not for //:pass@ //:@ or //@. Guess you can tell me how that stuff works from head? :]
17:34
<odinho>
annevk: preflighted requests _could_ also do redirects, I haven't seen any problems with it, as long as the preflight got a 200 OK back on the preflight request itself. -- So, possible, as in, it'd be possible to make the spec more complicated if you wanted :D I think.
17:37
<odinho>
"Userinfo is a username, optionally followed by a ":" and a password." -- so sounds sane.
17:41
<odinho>
Checked the referenced document in the fetch spec, rfc2616 -- but that was super unhelpful. You should just link to the real URL spec :-)
17:44
<annevk>
odinho: just @ doesn't mean anything
17:44
<annevk>
the others should fail
17:44
<annevk>
odinho: I will update it shortly
17:46
<annevk>
well, shortly, next time I touch fetch, dunno when that'll be
17:51
<odinho>
IE10 is green on all, opera green on all but :pass@ :@ @, chromium green on all but :@ @, firefox only green on :@ :pass@
17:53
<odinho>
So noone does what it "should" be, then: green on all but @
19:22
<zcorpan>
annevk: firefox fails to resolve data:test
19:46
<SimonSapin>
zcorpan: data:,test works, though
19:46
<SimonSapin>
or data:test,
19:46
<zcorpan>
SimonSapin: yes, they are legal
19:53
<annevk5>
Yeah I know, but they might change that. We could put data in the main URL parser, but that is not going to scale...
19:56
<SimonSapin>
it’s a completely different syntax, isn’t it?
19:57
<annevk5>
Well all URLs have their own syntax bits...
19:57
<annevk5>
mailto is not handled in any special way either
19:58
<SimonSapin>
in any case I would like to have the syntax well defined. Using email libs to parse the charset in data: is, well, not pretty https://github.com/Kozea/WeasyPrint/blob/master/weasyprint/urls.py#L188
19:59
<annevk5>
data's syntax is defined, no? There might be some issues with the MIME type parsjng
20:00
<annevk5>
My plan is to define URL processing at some point, with per-scheme rules for schemes that need it and do not have it
20:02
<SimonSapin>
annevk5: it’s defined in http://tools.ietf.org/html/rfc2397#section-3 , but I’m a bit fuzzy on quoted strings etc.
20:03
<annevk5>
Everyone is
20:05
<annevk5>
I might look into a clearer definition at some point, but maybe I can convince you to write it? :)
20:05
<SimonSapin>
maybe :)
20:08
<SimonSapin>
annevk5: do you think it is possible to define to find a web-compatible subset of the syntax for data: headers that does not depend on all of MIME?
20:09
<annevk5>
You need MIME, but what exactly that is...
20:11
<SimonSapin>
I could very well be convince to help with this, but I’d need some guidance :)
20:14
<annevk5>
Tomorrow :(
20:14
<annevk5>
Euh, :) silly phone
20:15
<Ms2ger>
That's pretty phoney
22:53
<GPHemsley>
AFAICT, no browser detects OTF or TTC fonts
22:55
<GPHemsley>
Also, is there a way for me to turn off automatically downloading files in Safari?
23:05
<GPHemsley>
They don't seem to detect WOFF, either
23:05
<GPHemsley>
And Chrome has strange behavior when a file is sent with the WOFF mime type
23:06
<GPHemsley>
if the file has no file extension, it doesn't add one
23:06
<GPHemsley>
but if the file has a '.woff' file extension, it adds another one
23:06
<GPHemsley>
so it tries to get me to download and save the file as signature.woff.woff
23:06
<GPHemsley>
(versus just signature without it)
23:08
<Hixie>
i would expect browsers to only try to sniff for fonts in @font-face
23:10
<GPHemsley>
hmm
23:10
<GPHemsley>
TTF seems to be sniffed outside of @font-face, though
23:10
<Hixie>
funky
23:10
<Hixie>
how does that manifest?
23:12
<GPHemsley>
just as a binary download
23:12
<GPHemsley>
the others are treated as plaintext
23:12
<Hixie>
isn't that just detecting non-text?
23:13
<GPHemsley>
oh, good point
23:13
<Hixie>
what mime type are you sending it as?
23:13
<GPHemsley>
I forgot the TTF signature was binary
23:13
<Hixie>
(the other types aren't? funky!)
23:13
<GPHemsley>
nope, they're 'ttcf', 'OTTO', and 'wOFF'
23:13
<Hixie>
with no binary bytes in the first 512?
23:13
<GPHemsley>
TTF, on the other hand, is the binary representation of '1.0'
23:14
<GPHemsley>
the test files consist only of the signatures and nothing else
23:14
<GPHemsley>
(which is surprisingly successful in other areas)
23:14
<GPHemsley>
the signature for bitmaps, for example, is only 'BM"
23:14
<Hixie>
aah
23:14
<SimonSapin>
GPHemsley: what binary representation?
23:14
<GPHemsley>
SimonSapin: 00 01 00 00
23:15
<GPHemsley>
(2 bytes for major, 2 bytes for minor)
23:15
<SimonSapin>
oh, ok
23:15
<GPHemsley>
anyway, back to the Thanksgiving festivities...
23:17
<zewt>
that doesn't sound like a very useful magic, heh
23:18
<zewt>
not unless you know in advance that you're looking for a font type, at least