00:18
<annevk>
"We don't want the Web to be a bunch of cordoned-off apps." I wonder if Jeff Jaffe heard of SysApps.
00:51
<karlcow>
http://www.w3.org/QA/2013/05/perspectives_on_encrypted_medi.html
00:52
<zewt>
uninteresting
00:53
<karlcow>
Thus Spoke Zarathustra
00:55
<zewt>
i find it hard to believe anyone arguing for drm in an open web (an obviously fundamental incompatibility) doesn't have a personal stake in the matter, and arguing against it is, well, obvious; so arguments on both sides are uninteresting to me
00:57
<zewt>
(it's also an ancient argument, not significantly changed by the web context, so I'd be surprised if there's any new content in it this time around)
09:44
<jgraham>
Ms2ger: Yes, I think I agree with your reading of the spec
09:47
<Ms2ger>
Thanks :)
11:04
<hsivonen>
http://www.w3.org/QA/2013/05/perspectives_on_encrypted_medi.html#c4842306
12:19
<Ms2ger>
jgraham, do you know if Opera has tests for form.foo?
12:21
<jgraham>
Ms2ger: I don't know
12:21
<jgraham>
I might remember to look later
12:21
<Ms2ger>
I would be grateful if you could
12:22
<Ms2ger>
Our code for that is so full of suck... :/
12:41
<Ms2ger>
Consider http://software.hixie.ch/utilities/js/live-dom-viewer/saved/2245
13:05
<GPHemsley>
Is there precedent (that I can look at) for using ABNF in WHATWG specs?
13:06
<darobin_>
GPHemsley: yes, but no one has lived to tell the tale
13:07
<Ms2ger>
Does http://www.whatwg.org/specs/web-apps/current-work/multipage/the-button-element.html#publickeyandchallenge count as abnf?
13:09
<MikeSmith>
Ms2ger: seems like it ise
13:09
<MikeSmith>
eh or does ABNF use braces?
13:10
<GPHemsley>
I was referring specifically to RFC 5234: http://tools.ietf.org/html/rfc5234
13:10
<MikeSmith>
there was some other ABNF in the spec
13:10
<MikeSmith>
maybe still there
13:10
MikeSmith
tries to remember what
13:13
GPHemsley
also wonders if he should distinguish between "well-formed" MIME types and "valid" MIME types
13:13
<MikeSmith>
GPHemsley: that sounds odd
13:14
<MikeSmith>
I'd wonder who cares about the distinction
13:14
<MikeSmith>
not implementors
13:14
<GPHemsley>
k
13:14
<GPHemsley>
in other news, there is no .abnf class in the spec CSS file
13:15
<MikeSmith>
so I guess whatever other ABNF Hixie had in the there before, it's gone now
13:16
<MikeSmith>
ah I remember
13:16
<MikeSmith>
it was for the text content of script elements
13:16
<MikeSmith>
or maybe style elements
13:16
<MikeSmith>
can't remember which
13:17
<jgraham>
If you think you want to use ABNF you are probably doing it wrong
13:17
<darobin>
definitely
13:17
<MikeSmith>
http://www.whatwg.org/specs/web-apps/current-work/multipage/scripting-1.html#restrictions-for-contents-of-script-elements
13:17
<MikeSmith>
that is BNFish still
13:18
<GPHemsley>
that is precisely the ABNF I was referring to
13:18
<MikeSmith>
ah OK
13:18
<GPHemsley>
though I note it doesn't have special styling
13:18
<GPHemsley>
MikeSmith: Thanks for finding it
13:19
<GPHemsley>
jgraham, darobin: Care to elaborate/enlighten?
13:19
<darobin>
GPHemsley: ABNF is fine for describing things like valid document constraints
13:19
<MikeSmith>
yeah that's the case here
13:19
<darobin>
but it tends to suck if what you want is to describe implementation behaviour
13:20
<MikeSmith>
it's not defining parsing here
13:20
<darobin>
GPHemsley: as MikeSmith says, you *don't* want it to define parsing rules
13:20
<MikeSmith>
well it also suck if you have to write a parser based on it anyway
13:20
<darobin>
but it's good to describe constraints on content
13:20
<MikeSmith>
I did have to write a parser for the validator based on that ABNF
13:21
<MikeSmith>
and it was was PITA
13:21
<darobin>
yeah
13:21
<darobin>
supposedly you could generate one
13:21
<MikeSmith>
right, in theory
13:21
<GPHemsley>
in your defense, though, it has rules like "any string that doesn't contain a substring that matches not-data1"
13:21
<darobin>
but I've never seen a tool that really worked well for the languages I was interested in
13:21
<MikeSmith>
darobin: yep, I tried that in this case but gave up after running into nothing but fail
13:22
<darobin>
had the same problem with WebIDL
13:22
<darobin>
ended up hand rolling because none of the generators made any sense
13:22
<GPHemsley>
Well, the alternative is to have something like this, which seems like overkill: http://mimesniff.spec.whatwg.org/#parse-a-mime-type
13:23
<darobin>
GPHemsley: if you want to constrain how an implementation behaves, and you want error handling, you need stuff like that
13:23
<GPHemsley>
Also, I'm currently using the hodgepodge definition from HTTP to define what a MIME type is: http://mimesniff.spec.whatwg.org/#valid-mime-type
13:23
<MikeSmith>
we could just replace that ABNF with something based on the state-machine code I wrote for it in the validator. And then Hixie could run that through his pre-processor that turns the code into semi-prose.
13:24
<jgraham>
GPHemsley: There are two valid approaches for speccing parsing algorithms that I have seen: explicit state machines and prose recursive descent parsers
13:24
<MikeSmith>
I much prefer http://mimesniff.spec.whatwg.org/#parse-a-mime-type even if/when it seems like overkill
13:24
<GPHemsley>
If I rewrite the definition of "valid MIME type" and related terms using my own ABNF, it would simplify the definitions (and the dependencies) and I think it would making defining parsing easier
13:25
<MikeSmith>
hmm
13:25
<MikeSmith>
it would not make it easier to implement man
13:25
<jgraham>
Most of the difficulty in parsing is handling edge cases
13:25
<GPHemsley>
I don't think it would make it any harder, either, though
13:26
<MikeSmith>
well only one person has to define the parsing behavior, so optimizing the spec to make things easier for one person doesn't seem like such a great plan
13:26
<GPHemsley>
I do wonder what makes it difficult, though
13:26
<MikeSmith>
browser implementors at least are very happy without ABNF and with state-machine-type definitions instead
13:27
<MikeSmith>
IETFers are a different story of course
13:27
<GPHemsley>
but ABNF can be used to define state machines, can't it?
13:27
<MikeSmith>
GPHemsley: look back at what darobin said
13:27
<MikeSmith>
especially about handling errors
13:28
<GPHemsley>
What I have in mind is something like "if it doesn't match ABNF-A, do X"
13:28
<MikeSmith>
hmm
13:29
<GPHemsley>
I don't plan to just say "here's the ABNF, bye"
13:29
<MikeSmith>
as somebody who may need to be implementing parsing for this, I would not like that much
13:29
<MikeSmith>
that's kind of the worst of both worlds
13:29
<MikeSmith>
I think you are likely to make everybody even less happy with that than now
13:30
<jgraham>
Yeah, unless there is exactly one error path, and even then, that sounds horrendous
13:30
<GPHemsley>
but isn't it essentially just a regexp?
13:31
<jgraham>
No actual browser when confronted with parsing problem thinks "hey, I know, I'll use regexps"
13:31
<Ms2ger>
Actually...
13:32
jgraham
suddenly becomes very afraid of the Gecko code
13:32
<GPHemsley>
^_^
13:32
<Ms2ger>
:D
13:33
<GPHemsley>
OK, so, then I have to ask: Do you all take issue with "valid MIME type" as it's currently defined?
13:33
<jgraham>
But seriously, bonsai or it didn't happen
13:33
<GPHemsley>
The link again: http://mimesniff.spec.whatwg.org/#valid-mime-type
13:34
<GPHemsley>
" A valid MIME type is a media-type, as defined in HTTP/1.1 or its successor, and is made up in part by a type and a subtype. The MIME type portion of a valid MIME type is the string represented by concatenating type, the string "/", and subtype. [HTTP]
13:34
<GPHemsley>
The MIME type portion of a valid MIME type excludes all parameters and is case-insensitive. [HTTP] "
13:34
<GPHemsley>
That's the part that I'm planning to replace with ABNF
13:34
<jgraham>
(I guess for the past half decade or so I mean mxr, but bonsai sounded better)
13:34
<GPHemsley>
heh
13:35
<GPHemsley>
I was like, "Do we really need to go back to CVS for this?"
13:36
<MikeSmith>
GPHemsley: making something essentially be a regexp is not a great goal. We have some code in the validator for parsing <time> that is an actual regexp where we do capturing groups and then check to see which capturing-group matches failed. i.e., pretty much what you've described
13:36
<MikeSmith>
and it's nasty
13:36
<GPHemsley>
Hmm
13:36
<GPHemsley>
OK
13:36
<Ms2ger>
jgraham, I guess I was thinking about pattern="" calling into the JS engine, which is kinda something else
13:36
<GPHemsley>
can you point me to a pretty example, then? :)
13:37
<jgraham>
Ms2ger: Right, in that case the input is a regexp, so there's not that much choice
13:38
<jgraham>
Obviously pattern should have taken BNF
13:40
<darobin>
GPHemsley: the thing is, in most cases you can't have isABNF || error()
13:40
<darobin>
what you want is ABNFConstruct1 || error1(), ABNFConstruct2 || error2(), etc.
13:40
<darobin>
actually if we had a language that did that nicely it could be sweet
13:40
<GPHemsley>
Well, yeah. I plan to use multiple ABNF rules
13:41
<darobin>
GPHemsley: at some point you're getting very close to prose then :)
13:41
<GPHemsley>
Perhaps I should just write it and let you all criticize it after the fact :P
13:44
<darobin>
GPHemsley: that's the right attitude :)
14:33
<GPHemsley>
Do WHATWG specs generally have abstracts?
14:36
<darobin>
GPHemsley: some do, some don't
14:36
<darobin>
(IIRC)
14:37
<Ms2ger>
jgraham, I was wondering if you remembered to look for those tests ;)
14:37
<darobin>
GPHemsley: using an abstract will simplify the life of whoever gets to steal your spec on behalf of a W3C WG
14:37
<GPHemsley>
Ah, so if I want to discourage that, I should remove it? ^_^
14:38
<jgraham>
Ms2ger: No, but now was a good time to remind me. Looking coming right up
14:38
<Ms2ger>
\o/
14:38
<jgraham>
I wouldn't get too excited
14:38
<GPHemsley>
darobin: Do you have any sense as to which portion is the majority?
14:39
GPHemsley
feels like the mimesniff abstract is mostly redundant to the introduction
14:39
<darobin>
GPHemsley: if you want to discourage that you should submit your spec to the HTML WG when it gets its new open license :)
14:39
<darobin>
GPHemsley: but overall, no, I don't know what the majority is
14:40
<GPHemsley>
Also, does anyone have any distinction between "key word" and "keyword"?
14:40
<darobin>
I think you want the latter
14:41
<GPHemsley>
darobin: I think I do, too, but that would go against the status quo
14:41
<GPHemsley>
(RFC 2119 uses "key word", and that is propagated to specs mentioning RFC 2119)
14:43
<darobin>
GPHemsley: ah, dunno then. I'd still use keyword at least outside of that context
14:43
<GPHemsley>
I'd like to use keyword inside of that context ;)
14:43
<darobin>
has anyone looked at making &apos: accepted in text/html as well? I thought it was but apparently not
14:43
<MikeSmith>
what rough beast, its hour come round at last, slouches towards Bethlehem to be born
14:43
<darobin>
GPHemsley: I wouldn't sweat it
14:44
<darobin>
MikeSmith: EME spec? :)
14:44
<MikeSmith>
heh
14:44
<MikeSmith>
well, now that you bring that up: http://www.w3.org/News/2013#entry-9813
14:44
<MikeSmith>
http://www.w3.org/TR/encrypted-media/
14:44
<MikeSmith>
the rest is future history
14:48
<jgraham>
Ms2ger: Well I found a file called Testcase%20showing%20the%20problem..html
14:48
<Ms2ger>
Heh
14:48
<jgraham>
Which isn'ta testcase and doesn't show the problem, but does seem to be related to named getters on forms
14:49
<jgraham>
Apart from that I'm not seeing much
14:49
<jgraham>
Asking zcorpan or hallvors on Monday might be a good strategy
14:49
<Ms2ger>
Sounds like I'm trying to coerce baku_ into writing them, then :)
19:20
<GPHemsley>
X-Content-Type-Options is now specced: http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2013-May/039561.html
19:49
<annevk>
GPHemsley: I don't think that's what actually happens for <img>
19:49
<annevk>
GPHemsley: I'm pretty sure that image/png will still decode image/gif
19:50
<annevk>
GPHemsley: because of the way image decoders work
20:29
<GPHemsley>
annevk: I'm not sure what you mean. How does what you said contradict what I said?
20:30
GPHemsley
is starting to wonder whether he remembers how to put thoughts into words.
20:41
<annevk>
GPHemsley: X-Content-Type-Options: nosniff still works for image/gif labeled as image/png I think
20:41
<annevk>
GPHemsley: I think the spec has that wrong, or did you test that case?
22:57
<GPHemsley>
annevk: I'm still not clear on what you're saying, but if you load an image in the main browsing context, the nosniff rule should apply. If you load an image through <img>, HTML bypasses the main sniffing algorithm using a context-specific sniffing algorithm.