03:13
<takkaria>
bloody hell, why does CC feel the need to use RDF for metadata?
06:42
<hsivonen>
takkaria: http://intertwingly.net/blog/2008/02/09/Mashups-Smashups#c1202870522
08:27
<zcorpan>
Hixie: what's the point in making xml:lang non-conforming in text/html?
08:27
<zcorpan>
Hixie: (i.e. lang in the XML namespace)
08:28
<Dashiva>
zcorpan: How would you ever get that?
08:28
<zcorpan>
Dashiva: with scripting or by moving a node from an xml document to an html document
08:28
<Dashiva>
But then you're working on the DOM form, not text/html anymore, aren't you?
08:29
<zcorpan>
Dashiva: when i say text/html i mean the spec's definition of "HTML document"
08:37
<hsivonen>
zcorpan: presumably, scripts written for both HTML and XHTML wouldn't cause the DOM to throw if allowed
08:38
hsivonen
isn't sure if the DOM throws now
08:55
<Hixie>
zcorpan: because you couldn't serialise it, so it would just get you in more trouble than it would be worth
08:55
<Hixie>
oh he left
08:56
<Hixie>
zcorpan needs to be online more. :-P
08:56
<philipj>
he's likely on the way to the office
08:58
<MikeSmith>
I wish I were on the way to the office there.. I'm sure it's nicer today than muggy and smoggy Tokyo in summertime
08:59
<MikeSmith>
rather, I wish I were in Bergen
09:00
<hsivonen>
Schematron sees to be one of those things whose design works nicely on today's desktop in non-batch mode but that causes grief when offered as an online service with concurrent users.
09:01
<hsivonen>
that is, 12 MB means nothing if you need 12 MB * 1 at a time on a desktop with 4 GB of RAM
09:01
<hsivonen>
s/sees/seems/
09:04
<Dashiva>
Hixie: So it's possible for a DOM to "be" text/html, and therefore unconforming, but you'd need a DOM validator to "prove" it?
09:07
<MikeSmith>
hsivonen: true perhaps of a lot of XML technologies in particular
09:08
<hsivonen>
looks like news gets distorted by the time it hits slashdot: http://tech.slashdot.org/article.pl?sid=08/08/20/218240&from=rss
09:11
<Hixie>
Dashiva: DOMs are the only thing you ever conformance check, and they are always either HTML Documents or XHTML Documents
09:12
<Hixie>
Dashiva: search for "flagged as being HTML documents"
09:14
<Dashiva>
Hixie: But since you can't serialize it, you wouldn't be able to send it to the validator. So you'd need a UA-local validator (what I meant by DOM validator). Right?
09:18
<Philip`>
You could serialise to a special non-HTML format that is capable of representing any DOM and that is understood by your favourite validator
09:19
<anne_>
Hixie, all the "through the element" needs to be changed to "through the element or object"
09:19
<anne_>
Hixie, currently it makes no sense for Window etc.
09:19
<anne_>
Hixie, just "through the object" would work too probably
09:21
<hsivonen>
Dashiva: if someone contributes a streaming SAX parser for SDF... http://simon.html5.org/specs/sdf
09:21
<hsivonen>
or we get an XML5 that can serialize all DOMs
09:23
Philip`
wouldn't suggest using SDF, since it can have O(n^2) length compared to the XML-serialised length
09:24
<hsivonen>
Philip`: Namespaces strike again?
09:25
<Hixie>
Dashiva: you don't serialise it, you pass the serialised document to the validator and they turn it into a dom and check that
09:25
<Philip`>
hsivonen: No, just "<x><x><x><x><x><x>..." will result in loads of indentation
09:26
<Hixie>
anne_: yeah, i think i have a note about that somewhere. feel free to file a bug if you care enough. :-)
09:29
<hsivonen>
DOM Level 1 could be relatively reasonably by represented as JSON
09:29
<hsivonen>
too bad that Namespaces make even that inefficient
09:39
<Dashiva>
Hixie: The case we were talking about was invalid because of a script creating xml:lang, the validator might not run that script (if it runs scripts at all)
09:40
<Hixie>
Dashiva: ah yeah, checking such cases would need a full browser-like environment and a validator to be integrated
10:32
<Hixie>
well
10:32
<Hixie>
this is annoying
10:33
<Hixie>
TAB, CR, and LF aren't in Zs.
10:33
<anne_>
you lose!
10:33
<anne_>
and will not collect 200$
10:34
<Hixie>
i guess i should use White_Space instead
10:34
Lachy
wonders why people keep putting the $ on the wrong end. it goes at the beginning.
10:34
<Hixie>
though that includes U+000B too
10:35
<anne_>
Lachy, cause that's how you say it
10:43
<zcorpan>
Hixie: there are other things in the dom that can't be serialized but are still allowed
10:44
<anne_>
who is w?
10:44
<anne_>
(re: recent commits)
10:44
<zcorpan>
hsivonen: setting the lang attribute in the xml namespace in html documents doesn't throw
10:45
<Hixie>
anne_: wakaba
10:45
<Hixie>
zcorpan: so?
10:45
<Hixie>
zcorpan: (like what?)
10:45
<zcorpan>
Hixie: like a textarea containing a comment node
10:45
<zcorpan>
Hixie: or a comment containing -->
10:46
<Hixie>
true
10:46
<Hixie>
that's mostly just because it'd be a pain for me to specify otherwise, to be honest
10:47
<zcorpan>
ok
11:15
<Philip`>
Hixie: How would an SVG file not labelled as SVG being sniffed as a valid image in <img> allow privilege escalation, given that scripts in SVG in <img> must not be run?
11:19
<Hixie>
Philip`: it wouldn't, but if scripts ever were to run, it would
11:19
<Hixie>
Philip`: it's more defence in depth than anything really
11:20
<anne_>
I think those rules match what we implement (though we also fail if the root element is not <svg> in the right namespace)
11:20
<Philip`>
Oh, okay
11:21
<Hixie>
anne_: that matches "User agents must not support non-image resources with the img element (e.g. XML files whose root element is an HTML element)."
11:21
<anne_>
ah
11:28
<Hixie>
ok bed time
11:28
<Hixie>
nn
11:36
<zcorpan>
it seems karl misunderstood the spec change wrt xml:lang
11:46
<anne_>
seems he did
11:46
<anne_>
admittedly it's quite hard
11:47
<anne_>
especially with the {xml:lang} versus xml{lang} nonsense
11:51
<hsivonen>
Namespaces make things hard.
11:52
<anne_>
I suppose this could be added to the anecdotical evidence thing
11:53
<hsivonen>
I made the Validator.nu error messages strictly correct but not necessarily obvious on this point as a juvenile act of rebellion against the badness of namespaces
11:53
<anne_>
maybe I should add the professor from my XML class who thought namespaces inherited into attributes
11:53
<anne_>
hsivonen, :p
11:53
<anne_>
lol, mibbit has a weird smiley for that one
11:54
<hsivonen>
anne_: your professor is in good company, since it seems that a W3C spec editor can have that misconception
12:00
<zcorpan>
hsivonen: <html xml:lang='en' lang='sv'> validates but <html lang='sv' xml:lang='en'> doesn't
12:00
<hsivonen>
hmm. that's weird
12:07
<zcorpan>
hsivonen: event-source was renamed to eventsource
13:32
<hsivonen>
zcorpan: <html xml:lang='en' lang='sv'> fixed
13:32
<hsivonen>
thanks
13:48
<hsivonen>
hmm. perhaps the spec check-in for renaming event-source wasn't properly annotated...
13:48
<hsivonen>
fixing anyway :-)
13:51
<anne_>
ah hsivonen has a flickr backup utility, good
14:02
<hsivonen>
hmm. I haven't updated Validator.nu's copy of the spec since March 20...
14:11
<hsivonen>
MikeSmith: do you need assertions.sch to work in an XPath 1.0 environment?
14:12
<hsivonen>
MikeSmith: that is, would your setup break if I start using XPath 2.0?
14:43
Philip`
notices that he failed to notice that the OpenGL 3.0 spec was released last week, after months of silence and delays, seemingly without any of the interesting features that had been promised earlier
14:43
<Philip`>
It seems the most radical change they've made is "In OpenGL 3.0 we have introduced a deprecation mechanism for the first time in OpenGL’s 16-year history."
14:47
<Philip`>
It doesn't seem entirely clear how the deprecation thing is meant to work - GL 3.0 implementations let you choose a "forward compatible" mode, which disables all the deprecated features, but also have to have a "full" mode which still provides them and is compatible with all old (GL <= 2.1) software
14:48
<Philip`>
so I guess the idea is that the features would be removed from the GL 3.1 "full" mode too, or maybe they're just going to hang around until GL 4.0, or until forever, since nobody's actually going to bother rewriting their code to not use the deprecated features
15:10
<MikeSmith>
hsivonen: using XPath 2.0 would not break anything for me
15:19
<Lachy>
http://ejohn.org/blog/queryselectorall-in-firefox-31/
15:21
<Lachy>
JohnResig, are the performance tests you used for that available? I'd like to test in them in Opera
15:45
<BenMillard>
each heading in my web page collection now provides an anchor point (via ID) and states how many entries are in that section: http://projectcerbera.com/web/study/2008/collection
15:46
<BenMillard>
Hixie, ^ is that a useful change?
16:29
<hsivonen>
MikeSmith: ok
16:30
<hsivonen>
Philip`: are the deprecated features conflicting with some insanely great improvements or are they just "bad"?
16:31
anne_
thought 1 and 2 were already incompatible
16:31
anne_
thus wonders what deprecation means in the context of OpenGL
16:36
<zcorpan>
http://tinyurl.com/559u23 (live dom viewer)
16:37
<zcorpan>
um, you need to insert a character before the script in firefox to work around the parsing bug...
16:38
<MikeSmith>
if there are cases where
16:38
<MikeSmith>
oops
16:38
<MikeSmith>
(ignore that)
16:44
<zcorpan>
http://tinyurl.com/5rj9l5 works in fx and saf
16:44
<zcorpan>
is there a spec that covers this?
16:45
<anne_>
ECMAScript?
16:45
<anne_>
though maybe I'm misunderstanding what you're testing
16:51
<Philip`>
hsivonen: They're not inherently bad ideas, they're just decade-old ideas and there are much better (more powerful, more efficient) alternative features that achieve the same effect on modern hardware
16:52
<Philip`>
like the whole glBegin();glVertex();glEnd() thing is deprecated, in favour of vertex arrays
16:53
<Philip`>
(The problem with all the legacy features is that they make drivers extremely complex, and make it hard to add new features since you have to consider interactions with all the old ones)
16:54
<Philip`>
anne_: OpenGL 1 and 2 aren't incompatible, in the sense that a program using the OpenGL 1.x API will still work in a GL 2.1 implementation (and will still work in GL 3.0 in 'full' mode)
16:54
<gDashiva>
zcorpan: What anne said, what is it that's working?
16:56
<anne_>
philip, interesting
16:56
<zcorpan>
anne_: ok
16:57
gavin_
just read http://lists.w3.org/Archives/Public/public-html/2008Aug/0510.html and couldn't agree more
16:57
<gavin_>
jgraham++
17:06
<gDashiva>
gavin_: Now watch it be completely ignored, or if we're lucky, marked as a troll.
17:06
gsnedders
has a lucky feeling about this one
17:10
<BenMillard>
I set my e-mail client so that "alt" in a subject line gets the message automatically marked as "Read"
17:10
<gavin_>
http://lists.w3.org/Archives/Public/public-html/2008Aug/0517.html is mind boggling
17:11
<BenMillard>
there's actually very little which gets through as unread in my e-mail client
17:17
<anne_>
gavin_, yeah, there's an enormous gap between the two camps
17:23
<gDashiva>
BenMillard: I hope it's more like \balt\b :)
17:23
<Philip`>
Some subject lines just say "... alternative text ...", so you don't want the trailing \b
17:24
<BenMillard>
gDashiva, the less I see the more I know. ;)
17:25
<gDashiva>
We need a neural network to determine if an occurence of 'alt' refers to alternate text for HTML or not
17:26
<Philip`>
A much simpler heuristic is to look at the size of the thread - if the subject contains the substring 'alt' and there are at least 20 messages, it must be about <img alt>
17:26
<BenMillard>
the only things that interest me on Public-HTML are links to use cases and the results of research...maybe I should use a whitelist instead of a blacklist?
17:27
<gDashiva>
BenMillard: Nah, there's plenty of namedropping in the alt threads about research (or the need for it, or the non-need for it)
17:28
<BenMillard>
ok, I'll stick with what I've got...it ain't broke, so don't fix it
17:29
<BenMillard>
I've decided to do a bit of analysis of the collection before my midpoint comes up
17:30
<BenMillard>
this means figuring out things are useful to count
17:30
<BenMillard>
here's what I think for headings: http://projectcerbera.com/web/study/2008/collection#headings
17:31
<BenMillard>
Philip`, I think some of the heading analysis could be automated
17:31
<BenMillard>
like checking when heading levels get skipped
17:32
<BenMillard>
other things need judgement, like figuring out when the headings should have changed level but didn't
17:35
<jgraham>
BenMillard: Are you planning to look at table summaries? contrary to Al's opinion I think it would be exceedingly useful to have data here like 10% of @summaries are accurate descriptions of the table, 30% are SEO keywords 20% are descriptions of a table with a different structure etc.
17:36
<Philip`>
..., 90% are saying that it's a layout table, etc
17:36
<BenMillard>
jgraham, yes
17:36
<jgraham>
BenMillard: Cool.
17:36
<BenMillard>
although I thought summary had already been studied in quite large numbers by Mark Pilgrim or someone?
17:37
<anne_>
that was longdesc
17:37
<BenMillard>
anne_, ah, that must be it
17:37
<anne_>
(if we remember the same thing, yes)
17:38
<BenMillard>
anne_, yes, I just found it and that jogged my memory: http://blog.whatwg.org/the-longdesc-lottery
17:38
<BenMillard>
jgraham, the collection currently mentions the summary attribute 6 times at the moment
17:39
<jgraham>
BenMillard: OK. I guess it is also something that might be worth looking at from a different same e.g. a random sample from google to see how that influences the results
17:40
<hober>
hmm.
17:40
<hober>
awfully tempted to feed the trolls.
17:40
<hober>
WDYT? http://edward.oconnor.cx/tmp/alt-reply.txt
17:40
<BenMillard>
jgraham, yes that would be good. here's ~800 from Philip`: http://canvex.lazyilluminati.com/survey/2007-07-17/analyse.cgi/tagattr/table/summary
17:41
<Philip`>
http://canvex.lazyilluminati.com/misc/summary.html
17:41
<gDashiva>
hober: Seems fine to me
17:41
<BenMillard>
Philip`, ooh that's saved me some work :P
17:42
<Philip`>
hober: Judging by the URL, it's talking about alt, which is a bad sign :-p
17:43
<hober>
Philip`: well, that may have been a poor choice of names. it's more of a side issue from that interminable thread
17:44
<BenMillard>
Philip`, is that summary page using the 799 pages in your 2007-07-17 collection? If not, can you tell me how many pages it covers?
17:45
<Philip`>
hober: I think you typoed "editor" as "editors" :-)
17:47
<Philip`>
BenMillard: http://krijnhoetmer.nl/irc-logs/whatwg/20070513#l-348 - it looks like that came from a list of pages from Yahoo (passing it common words ("a", "the", etc) as search terms)
17:48
<BenMillard>
Philip`, so it's covering 105 pages?
17:49
<hober>
Philip`: :)
17:49
<Philip`>
(It'd be fairly trivial to get a list of all the summary values used my newer collection of ~130K pages, though it'd take significant effort to categories them)
17:49
<Philip`>
BenMillard: According to myself (who I wouldn't necessarily trust), it is
17:50
<BenMillard>
yes, I wouldn't like to categories ~1,000 uses of summary
17:50
<Philip`>
It'd be easy to count how many contain the string "layout", though
17:50
<BenMillard>
indeed
17:51
<BenMillard>
if you randomly take summary strings from ~105 pages from that ~130,000, I could add those to study 210
17:51
<BenMillard>
as this seems to be a hot topic which reccurs
17:55
<BenMillard>
Philip`, could you count the total number of tables in the big and then see what proportion of them used summary?
17:55
<BenMillard>
*big sample
17:59
<JohnResig>
is dev.whatwg down?
18:05
<heycam>
JohnResig, it's very slow today
18:07
<BenMillard>
jgraham & Philip`, just noticed my collection contains "Summary="a summary of ..." where ... is the <caption> text." for this: http://dev.opera.com/articles/view/19-html-tables/#morestructure
18:08
<BenMillard>
should I studying each summary value in the context of the page it came from?
18:08
<BenMillard>
then I could check if it repeats the previous sentence, or is the previous heading rephrased slightly, etc
18:13
<Philip`>
BenMillard: Hmm, I can't look at anything now, since the university's internet connection has died, so I can't access all my files
18:13
<BenMillard>
Philip`, that's ok, I'm not your boss :)
18:14
<BenMillard>
you can do it whenever is convenient for you
18:14
<Philip`>
You might want to remind me later because otherwise I'll just forget :-)
18:14
<BenMillard>
Philip`, I have the same affliction :P
18:14
<BenMillard>
I'll put a note in the collection, then I'll stumble upon it from time to time anyway
18:15
<anne_>
jgraham, Philip`, WG dinner tonight, maybe tomorrow
18:15
anne_
has to go
18:15
<Philip`>
Hopefully it won't be long until somebody notices the whole university is offline and plugs the cable back in
18:15
<Philip`>
anne_: Tomorrow sounds reasonable to me
18:18
<jgraham>
Philip`: Someone just plugged me back in :)
18:19
<Philip`>
BenMillard: I see 3301 pages using @summary (on any element, but I assume pretty much all on table), out of 92066 pages using table
18:19
<Philip`>
jgraham: Indeed, I just noticed it started working too :-)
18:20
Philip`
likes how the CL has a totally independent open wireless network that just uses standard ADSL, so he can still connect when the proper network is down
18:22
<Philip`>
Oh, I've already got that data online - http://philip.html5.org/data/tag-count-pages.txt and http://philip.html5.org/data/attr-count-pages.txt
18:22
<Philip`>
(Excludes attributes with fewer than five occurrences, because there were zillions of them)
18:23
<BenMillard>
Philip`, awesome. :)
18:25
<Philip`>
(Oops - when I say "totally independent", I don't mean that at all, since it uses the same wireless access points and the internal CL network; but it's independent in its connection to the outside world)
18:26
<Philip`>
Oh, they broke the university again
18:26
Philip`
goes home
18:28
hsivonen
wonders if key compiler devs are on board with C++0x
18:29
<Philip`>
GCC and MSVC have already shipped parts of it, so I assume they are
18:29
<hsivonen>
ok
18:31
<BenMillard>
Philip`, can you count the number of summary attributes and the number of table elements rather than the number of pages?
18:31
<jcranmer>
well, boost provides most of the C++0x libraries
18:31
<hsivonen>
clearly, C++ embodies drastically non-Exupérian design principles
18:31
<Philip`>
BenMillard: http://philip.html5.org/data/tag-count-total.txt and http://philip.html5.org/data/attr-count-total.txt
18:31
<BenMillard>
Philip`, aha!
18:31
<Philip`>
BenMillard: (940942 table, 12834 @summary)
18:32
Philip`
fails to find helpful definitions of Exupérian on Google
18:33
<Philip`>
Aha, "Perfection is achieved, not when there is nothing more to add, but when there is nothing left to take away."
18:34
<hsivonen>
Philip`: that's the one
18:34
<BenMillard>
I prefer: "Perfection is having just enough."
18:34
<Philip`>
C++ has never had perfection as a design principle - it's just designed to be useful :-)
18:57
<BenMillard>
Philip`, 940,942 <table> elements with 12834 summary attributes means summary attribute appears on 1.3% of all <table> elements in that sample of 130,000 pages, right?
18:57
<BenMillard>
(maths isn't my strong point, even basic maths)
18:58
<Philip`>
BenMillard: Yes
18:58
<Philip`>
BenMillard: (Well, 1.4% if you round-to-nearest)
18:58
<Philip`>
BenMillard: (And not precisely 130,000, but close enough)
18:58
<BenMillard>
LOL, I've forgotten how to round numbers :D
18:59
<BenMillard>
940,942 <table> elements with 12,834 summary attributes means summary attribute appears on 1.4% of all <table> elements in that sample of ~130,000 pages?
19:00
<Philip`>
Yes
19:00
<jcranmer>
Philip`: round half-up, half-down, half-even, half-odd?
19:00
<Philip`>
jcranmer: Doesn't matter, since it's 1.36...% so it's not a half :-p
19:17
<BenMillard>
Philip`, the summary proportion looks about right compared with this (bottom of the <table> attributes list): http://code.google.com/webstats/2005-12/tables.html
19:40
<BenMillard>
Philip`, thanks for help with that. I'm off for dinner now, cya.
20:06
<hsivonen>
jgraham++
20:07
<hsivonen>
jgraham: there seems to be no documented reasoning why action 72 is a better solution to issue 20 than doing the kind of tweaks you mentioned
20:13
virtuelv
thinks that include proposal would mostly be solved with an equivalent to -o-content-size
20:14
<virtuelv>
http://lists.w3.org/Archives/Public/www-style/2008Jul/0380.html
20:16
<virtuelv>
the security implications you get in combination with legacy cmses and clients supporting something like <include /> just strikes me as "potentially ugly"
20:33
<hsivonen>
what does Dreamweaver generate into summary='' by default?
21:07
<jgraham>
hsivonen: I have a feeling that Dreamweaver generates the dimensions of the table, which could clearly just be determined on the client side (I might be wrong about what it does though; this is just a memory of the last time this discussion came up)
21:09
<hsivonen>
jgraham: sad if that's the case
23:47
<rubys>
anybody know the current implementation status of web sockets? A pointer to a webpage with this answer would be fine...
23:53
<Hixie>
i believe no browser implements it, but mcarter has a js-based shim implementation iirc
23:54
<rubys>
http://www.orbited.org/
23:57
<rubys>
That looks to be the one. Thanks! Two more (hopefully quick) questions. Is the video element compatbile with rtsp? Is there anything in HTML5 or related work that addresses video capture?
23:58
<Hixie>
<video> is intended to be compatible with any streaming protocol, though if there are any issues please do raise them and i'll try to address them
23:59
<rubys>
fair enough
23:59
<Hixie>
regarding video capture, do you mean streaming video from a camrea, or taking stills from a video stream?