09:13
<MikeSmith>
zcorpan: if you're around, I'm wanted to ask a question about the script ABNF at http://www.whatwg.org/specs/web-apps/current-work/multipage/scripting-1.html#restrictions-for-contents-of-script-elements
09:14
<MikeSmith>
or Hixie, if you're around
09:15
<zcorpan>
if websocket unmasking key is SHA-1(uuid, server nonce, client nonce, per-frame key)? how would you know what to use as per-frame key to get 0 as unmasking key?
09:15
<zcorpan>
MikeSmith: what about it?
09:17
<MikeSmith>
zcorpan: so the need to check for is "<!--" without a later "-->" is clear, but I'm trying to remember what I need to do for the "<script and "</script"
09:18
<MikeSmith>
is it not sufficient to have a check for "</script" followed by space, "/", or ">"?
09:19
<MikeSmith>
I don't understand why the ABNF has the stuff about "script-start = lt s c r i p t tag-end"
09:19
<MikeSmith>
instead of just for script-end
09:27
<zcorpan>
so e.g. "foobar(data1) <!--(escape) foobar(data2) <script>(script-start) foobar(data3) </script>(script-end) foobar(data2) <script>(script-start) foobar(data3) --> foobar(data1)" should be valid if i'm reading the abnf correctly
09:27
<zcorpan>
i don't understand why [ script-start data3 ] is in the first row
09:30
<zcorpan>
i also don't understand why it has [ escape ] at the end
09:31
<MikeSmith>
yeah
09:32
<MikeSmith>
even if I could implement that as spec'ed, I have no clue what kind of end-user-friendly error message I could report that would actually be useful
09:32
<zcorpan>
maybe i'm reading the abnf wrong, because it seems bogus to me :)
09:34
<MikeSmith>
I'm not sure those constraints are useful for the sake of document validation anyway
09:37
<MikeSmith>
since you can't feed the validator a DOM
09:37
<MikeSmith>
but instead a serialized document
09:39
<zcorpan>
a validator could be operating on the dom level if it's a browser extension
09:39
<MikeSmith>
well
09:39
<MikeSmith>
it's not going to be executing script hopefully
09:40
<zcorpan>
the browser executes script
09:40
<MikeSmith>
ah
09:43
<MikeSmith>
that's already a problematic use case
09:44
<MikeSmith>
for one thing, reporting the locations of any errors
09:46
<zcorpan>
you could point to the node and clicking it highlights the node in dragonfly
09:48
<MikeSmith>
yeah, could, potentially
09:49
<MikeSmith>
but the backend doesn't yet have any means for enabling that
09:56
<zcorpan>
sure, validator.nu isn't built for that environment in mind
09:57
<zcorpan>
s/in mind//
10:04
<MikeSmith>
anyway, I implemented the core checks for "/*" without a later "/*" and "//", for the script/@src case, and "<!--" without a later "-->" for both the src-less script case and script/@src case
10:05
<MikeSmith>
and got it up now at http://www.w3.org/html/check
10:07
<zcorpan>
cool
10:09
<MikeSmith>
I hope hsivonen will have time soon for some reviewing
10:10
<zcorpan>
i think it'd be more useful if the spec had parse errors for bogus <!--<script> cases in the parser
10:12
<zcorpan>
MikeSmith: i can't test xhtml with http://www.w3.org/html/check ?
10:12
<MikeSmith>
yeah you can
10:12
<MikeSmith>
under Options
10:12
<zcorpan>
oh
10:15
<zcorpan>
<script>&lt;/script></script> in xhtml is invalid according to the spec
10:15
<zcorpan>
but i don't think that's particularly useful to check for, personally
11:22
<Necra>
!loltip
11:22
<Necra>
oops
11:29
<MikeSmith>
zcorpan: I can't check for that in the datatype-checking code anyway
11:30
<MikeSmith>
because by the time it gets to that part of the code, "<" and "&lt;" are the same
11:38
<zcorpan>
MikeSmith: so? the abnf bans the textContent "</script>"
14:10
<bga_>
i can not send mail to es-discuss⊙mo :(
14:11
<bga_>
i have sended mail twice, but it do not displays in discuss tree
16:44
<_bga>
http://blog.mozilla.com/security/2010/12/27/addons-mozilla-org-disclosure/
17:01
<webben>
Would it not make sense to allow microdata to take value from the .value property of form fields? http://www.w3.org/TR/microdata/#values
17:02
<TabAtkins>
Hm, that's interesting. The immediate problem I can see is that, in the course of normal web crawling, you'd often see the value as "".
17:02
<TabAtkins>
I suppose the use-case is to let the user fill out a form and then extract the Microdata from it automatically?
17:07
<webben>
TabAtkins: That's a use.
17:08
<webben>
TabAtkins: I was actually looking at it in the context of http://tech.groups.yahoo.com/group/rest-discuss/message/17135?threaded=1&var=1&l=1
17:09
<webben>
TabAtkins: Which is discussing decorating HTML forms with RDFa semantics to decouple the domain vocabulary from the URL/form-encoding interface presented by the server and represented by @name
17:09
TabAtkins
is reading.
17:09
<webben>
TabAtkins: "<input id="case_title" name="Case.Title" property="dc:title"/>" does not work AFAICT
17:10
<webben>
TabAtkins: Since RDFa designates that dc:title is @content or the concatenation of text nodes.
17:10
<webben>
I thought maybe microdata could do this, but it doesn't look like it.
17:10
<TabAtkins>
It seems like the linked use-case is exactly the use-case I described.
17:12
<webben>
TabAtkins: Well if your use-case includes a programmer building a client to use the form, then yes.
17:12
<TabAtkins>
Which appears to be the described use-case, yes.
17:12
<webben>
ah ok
17:13
<TabAtkins>
Yeah, it seems to just be aiming for something that makes it easy to both extract the data client-side and server-side. Presumably there's already a framework server-side to convert the received form data into a more useful format.
17:14
<TabAtkins>
I'm not really sure what the value is of making @name and @property be different things.
17:14
<webben>
Well, it allows to use arbitrary @name's.
17:15
<TabAtkins>
Yes. So what's the value of that?
17:15
<webben>
Well, it allows you to present a consistent interface across different ways of processing name=value pairs.
17:16
<webben>
(For example, in PHP land, people tend to use foo[] for repeated values, but there's nothing baked into CGI about that.)
17:16
<TabAtkins>
How? There are only two ways I can see here - extracted as Microdata/RDFa on the client side, and received as name/value pairs on the server side.
17:16
<TabAtkins>
Those don't seem consistent unless you go to efforts to make them consistent.
17:17
<webben>
TabAtkins: I think the idea is the serverside processing and URLs could completely change, while clientside processing would persist.
17:18
<webben>
So for example you could change the @names and types of form attributes, but the client would persistently look for dc:title
17:18
<webben>
*of form fields
17:19
<TabAtkins>
Is there any good reason for this, though? Surely you could add an interpretation layer on the serverside that translates the kv pairs to whatever you needed?
17:20
<webben>
TabAtkins: Sure. Additional layers of processing doesn't sound more efficient than a persistent client interaction though.
17:21
<TabAtkins>
Sure. I'm just wondering about the utility of that. Extra indirection (a single piece of data referred to through two names) needs to be justified.
17:22
<webben>
TabAtkins: Well, at least it offers the potential to use a specific name (e.g. dc:title) rather than what's basically an opaque string (@name)
17:22
<webben>
s/specific/generically understood/
17:23
<TabAtkins>
@property is an opaque string as well. It has exactly as much meaning as @name, until you add meaning to it.
17:23
<webben>
TabAtkins: You can't reference a specific vocabulary via @name can you?
17:24
<TabAtkins>
Sure. I can put <input name="dc:title"> in my page.
17:24
<TabAtkins>
No current processor will recognize it as a curie, but I could do it.
17:24
<webben>
TabAtkins: That's still an opaque name in the absence of rules like microdata or RDFa for extracting it.
17:24
<TabAtkins>
Sure. Rules could be added if necessary, of course.
17:25
<webben>
Yes. It /might/ be that at least part of this usage could be addressed by providing such rules.
17:26
<webben>
I'm not sure there any standardizable way to map such vocabularies to a query string though.
17:26
<TabAtkins>
Hm? Of course there is. ?dc:title=foo
17:27
<webben>
TabAtkins: How about nested keys?
17:27
<TabAtkins>
You can't express that in forms.
17:30
<webben>
TabAtkins: well, you can ... e.g. name="book[4]title"
17:30
<TabAtkins>
Ah, yeah, sure, you can do server-side processing.
17:30
<webben>
exactly
17:31
<webben>
providing an independent mapping of the HTML representation to a shared vocabulary allows you to decouple the details of that processing from the client code.
17:32
<TabAtkins>
Hm, okay, I see your point. Trying to automatically express the relationship in the key sent over the wire is probably a harder problem than just letting people express it twice themselves (once explicitly in the @name, once implicitly in the microdata structure).
17:32
<webben>
I suspect so.
17:40
<webben>
TabAtkins: I think the point about value beginning in an invalid form (e.g. "" for a non-null data type) is a problem though.
17:41
<webben>
TabAtkins: I guess in so far as HTML5 can express data types/validity without script, extraction of data rather than extraction of input points could be premised on validity.
17:41
<TabAtkins>
Hm, that's true.
17:42
<TabAtkins>
But that would mean that, for example, non-required inputs would still be included.
17:43
<webben>
Excluding blank values might be possible; or including blank values might not be damaging - I'm not sure.
18:42
<matjas>
TabAtkins: [post] - [site] fo’ sho’!
18:43
<TabAtkins>
matjas: That's what I was thinking. [site] - [post] is only for corporate branding - I find putting the post first is a lot more attractive for blogs.
18:44
<matjas>
TabAtkins: replied on Twitter as well, but [p] - [s] is definitely the most useful variant if you consider small tab sizes (like in Chrome)
18:55
<webr3>
why not just [title]
18:57
<TabAtkins>
That makes it harder to, frex, remember where a bookmark came from.
18:57
<TabAtkins>
Anyway, done.
18:57
<matjas>
Yay!
18:58
<TabAtkins>
Also exposed a bug where I wasn't setting rel=canonical properly. I've rearchitected slightly to make future bugs of this sort less likely.
19:25
<matjas>
TabAtkins: heh, your site doesn’t work without the www.
19:26
<TabAtkins>
matjas: Indeed. That's the default for my provider - they seem to have a religious thing against it. I haven't cared enough to set it up differently.
19:43
<TabAtkins>
Hixie: Any reason why the section on CSS pseudoclasses only mentions a restricted list of elements that can match :active, when in practice all browsers allow all elements to be :active while they are clicked?
20:30
<Hixie>
TabAtkins: iirc it was based on a request of the csswg, but honestly i don't recall
20:30
<Hixie>
well
20:32
<annevk>
TabAtkins, wow, your provider is silly
20:32
<annevk>
TabAtkins, get DreamHost :)
20:33
<Hixie>
dreamhost are pretty silly too
20:33
<Hixie>
though maybe in a different way
21:00
<TabAtkins>
Hixie: I suspect it was a request from fantasai, since she dislikes the idea of arbitrary elements being activatable.
21:01
<TabAtkins>
Hixie: In any case, all browsers allow all elements to be :active.
21:01
<TabAtkins>
I can file a bug if it would help.
21:04
<Hixie>
TabAtkins: yeah, file a bug or send e-mail. IE didn't used to do this like the others.
21:05
<TabAtkins>
Yeah, but they didn't use to do :hover on non-links either.
21:05
<TabAtkins>
Will file.
21:05
<Hixie>
so?
21:05
<Hixie>
we're talking about :active, not :hover :-)
21:05
<Hixie>
if we do change this we'll have to define what :active means
21:05
<TabAtkins>
Right, but they're done through similar events, and conceivably are handled similarly in the browser.
21:05
<Hixie>
and how to do it from the keyboard
21:06
<TabAtkins>
You still can't activate arbitrary elements from the keyboard, unless they're tab-able or something.
21:07
<Hixie>
that seems suboptimal
21:10
<TabAtkins>
Huh. IE8 & 9 let any element be activated, but only the target of the click gets :active, not everything up the chain.
21:11
<bga_>
http://twitter.com/bga_/status/19863944524988416 :P
21:12
<bga_>
may be js not must know this. may be its security issue
21:14
<TabAtkins>
Bah, I can't be bothered to find the html bugzilla again. I'll just email. Same diff.
21:26
<TabAtkins>
Aren't there already some slightly hacky ways to force the event loop to spin in the middle of code? (Without breaking the code up and using setTimeout, that is.)
21:34
<TabAtkins>
(I'm thinking something involving databases has this effect.)
21:35
<Hixie>
(fwiw, i use the same mechanism for e-mail and bugs now, so it really makes no difference to me)
21:36
<TabAtkins>
Hixie: Yeah, I know, that's why I did it.
21:36
<Hixie>
this xml/html tf is baffling to me
21:36
<Hixie>
they were created to figure out what problem they were created to figure out?
21:36
<Hixie>
the w3c is getting dangerously close to poe's law
21:36
<TabAtkins>
No, they were created to make HTML act more like XML. Nobody wants to admit it, though.
21:36
<Hixie>
we already did that in the whatwg
21:36
<Hixie>
years ago
21:37
<Hixie>
html is closer to xml than it ever has been
21:37
<TabAtkins>
But the namespaces, Hixie! THE NAMESPACES!
21:37
<Hixie>
the namespaces are issue 41
21:37
<Hixie>
it seems premature to do anything while we're still waiting for teh chairs to decide on that
21:37
<TabAtkins>
Yup.
21:48
<karlcow>
I became insane, with long intervals of horrible sanity.
22:10
<TabAtkins>
annevk: Looks like my provider added a one-click option for allowing bare domains since I last checked. It automatically issues a 301 to the www. version, which is fine with me.
22:13
annevk
prefers without redundant characters
22:13
<TabAtkins>
I don't particularly care either way, I just want a canonical url for the google juice.
22:13
<annevk>
that is, long ago I bought into http://no-www.org/
22:16
<TabAtkins>
Doing what I just did (auto-redirect from bare to www) is compatible with no-www philosophy.
22:16
<TabAtkins>
I agree that not allowing the bare name isn't ideal. I only left it like that because I was lazy.
22:17
<TabAtkins>
Hehe, I love extra-www.org
22:20
<annevk>
I guess it is, yes, just seems like waste
22:20
<TabAtkins>
I don't understand what's being wasted.
22:20
<annevk>
characters :)
22:21
<annevk>
either characters or an additional request
22:21
<TabAtkins>
I don't understand how chars are wasted - people can still type/link the bare domain name.
22:21
<TabAtkins>
The additional request is required in one direction or the other, so I don't see much distinction about which direction it goes in.
22:23
<kennyluck>
huh. http://yes-www.org/ returns 403 while http://www.yes-www.org/ works
22:23
<TabAtkins>
Clearly they're doing the Hardcore option.
22:24
<karlcow>
one or the other it doesn't seem that important. How many type really type the letters of the domain name
22:24
<TabAtkins>
Yeah, I mean I just type "FACEBOOK LOG ME IN" into google.
22:24
<karlcow>
houla my usual dislexya
22:25
<karlcow>
how many time
22:25
<karlcow>
people
22:25
<karlcow>
geeez
22:25
<annevk>
actually TabAtkins, your site is not correct per no-www
22:26
<annevk>
not that it matters
22:26
<TabAtkins>
Hm?
22:27
<annevk>
check the validator
22:27
<TabAtkins>
Yeah, it won't work yet. I just made the DNS change, like, ten minutes ago.
22:28
<karlcow>
surprised nobody proposed worldwideweb.example.org
22:51
<karlcow>
another useless campaign http://www.antiifcampaign.com/
22:51
<TabAtkins>
Actually, that's not a bad campaign idea.
22:52
<karlcow>
it's poetic
23:01
<Philip`>
Seems a misleading name - it's not anti-if, it's just anti if-over-type-identifiers-as-a-cheap-substitute-for-polymorphism
23:07
karlcow
hesitates to launch the goto-back-campaign.com :p
23:12
<karlcow>
but for now I guess I will do the less-have-a-wonderful-dinner.com
23:32
<TabAtkins>
annevk: You're hopefully asleep by now, but could you respond to the "What to include in a propdef" thread on the csswg private list?