01:03
<Hixie>
well, i see my e-mail wasn't clear enough
01:03
<Hixie>
sigh
01:03
<TabAtkins>
On the other hand, I found the response pretty amusing.
01:05
<Hixie>
yes :-)
01:17
<Hixie>
hsivonen: yt?
01:48
<TabAtkins>
Oh wow, really? A legacy D-Link router? That's crazy.
01:52
<Hixie>
eseidel: i've begun changing the parser
01:52
<Hixie>
you can track changes at http://html5.org/tools/web-apps-tracker
02:20
<eseidel>
Hixie: thank you. I'll start implemnting your fixes tomorrow. I still need to talk to hsivonen, but he's been AWOL :p
02:21
<eseidel>
hsivonen: we should really sync up our test cases. We've written a bunch more for your fancy harness
02:25
<eseidel>
Hixie: I do not understand the mental model behind foreign content
02:25
<eseidel>
it was not clear to me from the spec
02:25
<eseidel>
sec
02:26
<Hixie>
mental model?
02:28
<eseidel>
Hixie: sorry, got a phoen call right as I was starting to explain
02:28
<eseidel>
it's unclear what to do with misnested tags in forieng content mode
02:28
<Hixie>
misnested how precisely?
02:29
<eseidel>
<div><svg></div></svg>
02:29
<Hixie>
(my usual answer is "just do what the spec says", but i'll wait until the spec bugs are fixed before saying that again :-) )
02:29
<eseidel>
is the idea that any non-foreign tag, breaks out of fo?
02:29
<eseidel>
fc, rather
02:29
<Hixie>
that should be treated as <div><svg></svg></div></bogus>, iirc
02:29
<eseidel>
interesting
02:30
<Hixie>
(is that what the spec now says?)
02:30
<Hixie>
basically the idea was that we wanted to add svg and mathml parsing with minimum impact on existing documents
02:30
<Hixie>
including documents that already contained bogus <svg> elements
02:30
<eseidel>
Hixie: I think I'll wait until you're done editing
02:30
<Hixie>
k
02:30
<eseidel>
since I don't feel i'm explaining myself well
02:31
<eseidel>
likely due to distractions
02:31
<Hixie>
k
02:34
<eseidel>
Hixie: one related confusion. If you see a mis-nested close tag in foreign content, right now the spec has you just pop all the way back until that tag makes sense. Which can ahve you popping through other insertion modes where the tag would be ignored. Another way to handle that would be any time you see a mis-nested close tag, to break out of FC immediately and re-process.
02:34
<Hixie>
do you have an example of popping through other inserting modes?
02:35
<eseidel>
<body><svg></html></svg> would be one example I think.
02:35
<eseidel>
I need to find an example where we ignore close tags, setc.
02:35
<Hixie>
you don't pop through other insertion modes where it would be ignored in that example do you?
02:36
<Hixie>
i mean, <body></html> won't ignore anything
02:36
<Hixie>
not sure what you mean
02:36
<eseidel>
<body><svg><col>Foo</svg>
02:36
<eseidel>
normally the "col" would be ignored, but since it's in the SVG, it causes it to break out of the SVG.
02:36
eseidel
looks for a better example
02:37
<Hixie>
no, that results in a <col> element in the SVG namespace
02:37
<eseidel>
Hixie: ah, <body><svg></foo><circle></svg>
02:38
<eseidel>
currently that would crash, if I remember the spec correctly (you'd walk off the bottom lookcing for <foo>
02:38
<Hixie>
i just fixed that one
02:38
<eseidel>
ok
02:38
<eseidel>
but </foo> woudl be ignored in body, so seems we shoudl ignore in in svg
02:38
<Hixie>
(not sure if i fixed it right)
02:38
<Hixie>
(let's see)
02:38
<eseidel>
Hixie: I'll implement whatever you wrote tomorrow, and then I can pick on you more then :)
02:38
<eseidel>
I'll leave you to work now
02:38
<Hixie>
k
02:38
<Hixie>
later
02:41
<Hixie>
(ok, looks like the </foo> in the example above just gets ignored now)
02:43
<eseidel>
(that's our current behavior too)
02:43
<Hixie>
(but in the case like <body><svg><g><foreignContent><b></g>, the </g> is not ignored, because the in body mode closes it)
02:44
<Hixie>
(not sure that this is completely coherent, but we'll have to try various examples and see if they all make sense now)
02:44
<eseidel>
k
02:44
<eseidel>
would like to find some way for hsivonen and I to better share test cases so we can implement these same things here
02:44
<eseidel>
our code is wildly different :)
02:44
<eseidel>
which is a good thing, I think
02:45
<Hixie>
yes
02:47
<eseidel>
<b id="1"><p><script>document.getElementById("1").id = "2"</script></p>TEXT</b> may turn into a spec bug... It's nice (mentally) to re-use the token, but "expensive" from an implementation perspective
02:48
<Hixie>
eseidel: as opposed to what?
02:48
<Hixie>
the spec used to clone the actual elements, but hsivonen asked me to use the token instead so he didn't need to keep a hold of the actual DOM since he's doing it off-thread
02:49
<eseidel>
ha.
02:49
<eseidel>
I think mozilla is crazy to do the parser off-thread. :)
02:49
<Hixie>
no disagreement here, but they get a minor advantage since they took the plunge and implemented it first :-)
02:49
<eseidel>
old WK clones the element, new one does too. I haven't checke IE
02:50
<Hixie>
IE doesn't clone
02:50
<Hixie>
it uses the same element
02:50
<Hixie>
you get non-tree DOMs
02:50
<Hixie>
it's quite special
02:50
<eseidel>
oh goodie :)
02:50
<Hixie>
at least old IEs did
02:50
<Hixie>
dunno what the recent ones do
02:50
<roc>
what's crazy about doing the parser off-thread?
02:50
<eseidel>
Hixie: I mean, we can store off the tokens (or the attributes from the tokens). its just more mallocs we'd probably like to avoid
02:51
<eseidel>
I'm not sure you want to pay all the sychronization cost.
02:51
<eseidel>
I mean, I liek the idea of using more threads in the engine, given the way hardware is going
02:51
<eseidel>
but doing all the dom access, which has to pretend to be single theaded, seems ouchy
02:51
<Hixie>
eseidel: why is it any more mallocs?
02:51
<roc>
the synchronization costs are low
02:51
<Hixie>
eseidel: the way they implemented it they have minimal synchronisation
02:52
<roc>
the DOM is not accessed directly from the parser thread
02:52
<eseidel>
Hixie: cause we have to copy the attributes array in case you change it in script
02:52
<Hixie>
(it's pretty cleverly done actually)
02:52
<Hixie>
eseidel: oh i see
02:52
<roc>
the parser generates a list of "tree operations" which are posted to the main thread
02:52
<eseidel>
i see.
02:52
<Hixie>
eseidel: can't you do a copy-on-write or something? i guess that'd be a performance hit too.
02:52
<eseidel>
Hixie: I'm not sure this will end up being a big deal. but for now we're not going to implement that quirk.
02:53
<eseidel>
Hixie: I have a test for it and a big FIXME so we won't forget :)
02:53
<Hixie>
cool
02:53
<eseidel>
hsivonen still owes me a response to my "<a>1<button>2</a>3</button>" test case
02:53
<eseidel>
I think minefield builds the wrong tree for that
02:54
<eseidel>
somehow </a> is breaking out of the <button> in minefield, which makes no sense to me
02:55
<Hixie>
there's some weird rules around </a> iirc
02:56
<Hixie>
oh that's for the <a><button><a> case, nevermind
02:56
<Hixie>
not <a><button</a>
02:57
<eseidel>
I filed https://bugzilla.mozilla.org/show_bug.cgi?id=578568
02:57
<Hixie>
ok i'm done for the day, will continue with these fixes tomorrow
03:07
<boblet>
hixie: for links.html#examples-1 you should combine George Washington’s itemprop="fn" and itemprop="n" to itemprop="fn n"
03:09
<boblet>
also having Jack Bauer’s given-/family-names in meta seems odd since they’re already in content
03:35
<MikeSmith>
http://www.boston.com/bostonglobe/ideas/articles/2010/07/11/how_facts_backfire/
03:36
<MikeSmith>
"when misinformed people, particularly political partisans, were exposed to corrected facts in news stories, they rarely changed their minds."
03:36
<MikeSmith>
"In fact, they often became even more strongly set in their beliefs."
03:37
<othermaciej>
I totally disagree with that story
03:37
<othermaciej>
cause I heard somewhere that people can change their minds
03:38
<boblet>
heh
03:39
<MikeSmith>
" the ones who were the most confident they were right were by and large the ones who knew the least about the topic"
03:39
<MikeSmith>
“It implies not only that most people will resist correcting their factual beliefs,” he wrote, “but also that the very people who most need to correct them will be least likely to do so.”
03:40
<boblet>
MikeSmith: so that explains … America then, huh
03:40
boblet
ducks
03:48
<MikeSmith>
boblet: explains quite a lot of things
03:49
<boblet>
MikeSmith: sadly, yes
04:27
<variable>
anyone here want a hope ticket for saterday/sunday for a cheaper price? I found out that I likely won't be able to stay the entire time ?
04:27
<variable>
woops
04:27
<variable>
wrong channel
04:47
<Hixie>
boblet: please file bugs for those (use the box at the bottom to do it automatically into the right component)
05:12
<variable>
Hixie, do you remember of any statments by you that about about direct client side includes?
07:42
<annevk>
we won't get to Last Call this year in the HTML WG?
07:42
<annevk>
geez
07:43
<annevk>
was'nt
07:43
<boblet>
Hixie: will do
07:43
<annevk>
wasn't Paul Cotton specifically assigned to help us get through this quicker or something?
07:43
<annevk>
oh well
07:43
<Hixie>
annevk: why wouldn't we get to LC this year?
07:44
<Hixie>
(if we don't get to LC this year, we surely can't make REC this quarter as per our charter!)
07:44
<Hixie>
i guess we can still make CR by 2012 like i predicted though
07:46
<zcorpan_>
Hixie: you mean WHATWG CR?
07:46
<Hixie>
i meant htmlwg CR
07:46
<Slaanesh>
It seems the longer the W3C process is dragged out, the more likely any given feature will become a victim of cross-browser implementations
07:47
<Hixie>
yup
07:49
<zcorpan_>
Slaanesh: i don't think the w3c process really affect browser implementations one way or the other
07:51
<annevk>
Hixie, that's what Sam said in his reply to Jonas
07:51
<Hixie>
ah
07:51
<annevk>
"unless we split the draft" ....
07:54
<mhausenblas>
MikeSmith around? re "Microdata/Microformat/RDF community should collaboratively write a document explaining about machine-readibility using non-RDF terms"
07:55
<mhausenblas>
it totally agree and would happily contribute to such an effort, MikeSmith
07:56
<MikeSmith>
mhausenblas: cool
07:56
<MikeSmith>
I'm happy to let anybody drive it
07:56
<mhausenblas>
maybe you want to take a quick look at http://linkeddata.deri.ie/node/58
07:56
<mhausenblas>
MikeSmith, I know who should not drive it ... /me :D
07:57
<mhausenblas>
link above was one of my earlier attempts (yes, restricted to RDF, I know, I know)
07:57
<mhausenblas>
but you get the idea?
07:57
MikeSmith
takes a look at the link
07:57
<mhausenblas>
tx
07:57
<MikeSmith>
mhausenblas: well, we now know at least two people who should not drive it :)
07:57
<mhausenblas>
hehe
07:58
<mhausenblas>
seriously, I guess someone rather neutral would be ideal
07:58
<MikeSmith>
yeah
07:58
<mhausenblas>
just fire up an etherpad and spread the word (here?)
08:02
<Slaanesh>
zcorpan_: Not the current one, perhaps. But if w3c wants to be involved in UA conformance later, they risk being too late
08:03
<annevk>
fwiw, "the w3c" includes UAs
08:04
<Slaanesh>
"The subset of w3c left after removing the intersection with whatwg" maybe?
08:04
<annevk>
something like that, yeah
08:05
<Slaanesh>
I wonder, if the suggestion to drop all author conformance had passed, where would be we now? CR?
08:05
<annevk>
I was just saying it since some people seem to have the impression the W3C is some kind of standalone entity, while it's mostly just a collection of lots of companies
08:06
<Hixie>
Slaanesh: dropping all authoring conformance criteria would have made very little difference so far
08:06
<Hixie>
Slaanesh: most of the issues have been over editorial text
08:07
<Slaanesh>
But isn't much of the text about author conformance?
08:07
<Hixie>
no
08:07
<Hixie>
most of it is implementation rules
08:07
<Hixie>
(i.e. UA conformance criteria)
08:08
<Slaanesh>
Selective memory on my part, I guess
08:08
<Slaanesh>
... why do the church bells ring at 09:10 every day in Zurich?
08:09
<Hixie>
Slaanesh: calling the faithful to church
08:10
<Hixie>
(at a guess based on other churches in switzerland)
08:11
<MikeSmith>
mhausenblas: not sure what would be the best way to jump start it.. comes back to the usual problem of needing somebody who has the time and motivation to make it happen
08:12
<Slaanesh>
Hixie: I thought all the faithful were at work already ;)
08:12
<mhausenblas>
MikeSmith, I'm happy to *take* time for it (and I certainly have the motivation) ... BUT
08:13
<mhausenblas>
I also know that if I propose it the typical reaction here and in some other place would be a more or less friendly fck off
08:13
<MikeSmith>
Slaanesh: it seems like churches should be trying to call the faithless, anyway
08:14
<mhausenblas>
so, you see my problem, MikeSmith? ;)
08:14
<MikeSmith>
mhausenblas: random idea: maybe a Wikipedia article
08:14
<mhausenblas>
hm
08:14
<MikeSmith>
I realize that the Wikipedia editor community is as dysfunctional or worse than any
08:15
mhausenblas
has rather bad experiences with this ... and also seems a bit like misusing Wikipedia
08:15
<mhausenblas>
well, I think it's clearly against their policy (re using it as a discussion forum)
08:15
<Hixie>
wikipedia book? :-)
08:15
<mhausenblas>
hehe
08:15
<mhausenblas>
whatz your hunch Hixie?
08:15
<Hixie>
no idea what the topic is :-)
08:15
<mhausenblas>
I mean, assuming you'd be interested ;)
08:16
<mhausenblas>
ah
08:16
<mhausenblas>
re "Microdata/Microformat/RDF community should collaboratively write a document explaining about machine-readibility using non-RDF terms"
08:16
<jgraham>
eseidel: If you have testcases, please check them in to the html5lib repository
08:16
<mhausenblas>
MikeSmith and I are discussing how to kick this off
08:16
<jgraham>
eseidel: I can give you commit access if you need it
08:16
<Slaanesh>
A vcs repository?
08:17
<eseidel>
jgraham: I did not know there was a repository. b ut yes, we have lots of tests
08:17
<jgraham>
eseidel: Also, hsivonen already knows about the <button> weirdness I think. The conclusion is that the spec is probably broken
08:17
<eseidel>
jgraham: see http://trac.webkit.org/browser/trunk/LayoutTests/html5lib/resources/
08:18
<eseidel>
jgraham: yeah, I got CC'd on the right bugs
08:18
<Hixie>
mhausenblas: what does that mean?
08:18
<eseidel>
after the fact
08:18
<mhausenblas>
(MikeSmith I think we should also ask mamund on board - he's doing great stuff at http://amundsen.com/hypermedia/
08:18
<jgraham>
eseidel: http://code.google.com/p/html5lib/source/checkout
08:18
<mhausenblas>
Hixie, an attempt to overcome the silly low-level, syntax-driven fights
08:18
<mhausenblas>
for a greater, Webish data idea
08:19
<eseidel>
jgraham: crazy
08:19
<Hixie>
ah well i'm not your man
08:19
<Hixie>
i think webish data is a lost cause
08:19
<mhausenblas>
fair enough
08:19
<Hixie>
at least machine-readable data
08:19
<mhausenblas>
interesting. care to explain why?
08:19
<Hixie>
natural language interpretation is where the real solutions are going to lie
08:19
<mhausenblas>
ah, right, I see
08:19
<eseidel>
jgraham: I'll talk with abarth about it, and figure out how we'll go about syncing up
08:19
<Hixie>
we'll never get a critical mass of authors writing data in common vocabularies using any of the syntaxes discussed, microformats, microdata, rdf*, whatever
08:20
<Hixie>
i mean, we have enough trouble getting them to use even basic HTML semantics
08:20
<mhausenblas>
ok, yeah, these are the two fundamental directions I sense as well
08:20
<mhausenblas>
I agree that it's hard to achieve it, yes
08:20
<jgraham>
eseidel: OK. I can obviously sync up the tests that are already there if you like
08:20
<mhausenblas>
true (re "getting them to use even basic HTML semantics")
08:20
<Hixie>
i think it would require such a fundamental change in the way people are educated throughout the planet that it is for all intents and purposes impossible
08:20
<jgraham>
eseidel: (although I guess there might be licensing issues?)
08:21
<Hixie>
i mean, i seriously think that getting this done would require massive investment in education in every computer-literate country
08:21
<mhausenblas>
hehe Hixie, this is the only (main) difference where I seriously disagree with you ;)
08:21
<eseidel>
jgraham: abarth and I have written all the tests. as far as I'm concerned, they're public domain
08:21
<Hixie>
we're talking trillions of dollars, and serious political upheaval
08:21
<eseidel>
jgraham: I'm 100% certain abarth feels the same
08:21
<mhausenblas>
but it's happening (see data.gov.uk and data.gov, just to mention a few)
08:22
<eseidel>
jgraham: note, we don't pass quite all of them yet: http://trac.webkit.org/browser/trunk/LayoutTests/html5lib/runner-expected-html5.txt
08:22
<Hixie>
mhausenblas: oh individual places will provide data in structured form, sure
08:22
<mhausenblas>
and I'm not talking so much about the individuals producing structured data
08:22
<mhausenblas>
hehe
08:22
<jgraham>
eseidel: Sure :) It might be better for you to add them yourselves to the html5lib repository for the avoidance of doubt though
08:22
<mhausenblas>
my reasoning is very simple:
08:22
<eseidel>
jgraham: https://spreadsheets.google.com/ccc?key=0AlC4tS7Ao1fIdEo0SFdLaVpiclBHMVNQcHlTenV5TEE&hl=en is our status
08:22
<Hixie>
mhausenblas: but they will be rare enough that it's easier to write per-site hard-coded interpreters, and thus it's easier to use dedicated database formats for each one, than to try to embed the data in HTML each time
08:22
<jgraham>
eseidel: Also, I added some more tests to the html5lib repo recently
08:23
<jgraham>
So you might want to pull
08:23
<Hixie>
mhausenblas: trying to crawl the data out of the HTML only makes sense if the vast majority of the web is using this, which is what i don't think will happen
08:23
<eseidel>
jgraham: good! we would love to pass them!
08:23
<mhausenblas>
if you (== company, gov agency, whatever) already have HTML out there, why not putting a bit more structure into it (+links)
08:23
<mhausenblas>
right
08:23
<Hixie>
mhausenblas: if you (== company, gov agency, whatever) already have data, why not just put it out there in its native form? (e.g. a SQL database)
08:23
<mhausenblas>
true
08:23
<Hixie>
that'd be a bazillion times easier to deal with in practice
08:24
<Hixie>
it's not like these sites are all using the same useful vocabularies
08:24
<mhausenblas>
but then you're putting it on the Web, not in the Web
08:24
<Hixie>
sure but who cares
08:24
<mhausenblas>
well, it's not utilising the most important feature of the Web
08:24
<mhausenblas>
the links
08:24
<mhausenblas>
:D
08:24
<mhausenblas>
(and for the record: I do care ;)
08:24
<Hixie>
what matters is what the end user benefit is, not whether we're doing things theoretically correctly or in architecturally correct ways
08:24
<mhausenblas>
exactly
08:25
<mhausenblas>
Hixie, whatz your take on OData?
08:25
<Hixie>
better to have a bunch of dedicated sql queries than to try to reuse HTML because it's "in the web not on the web"
08:25
<Hixie>
OData?
08:25
<Hixie>
whatz?
08:25
<mhausenblas>
come on ...
08:25
<mhausenblas>
you do know MS OData? based on Atom?
08:26
<jgraham>
eseidel: Just send me the google account email addresses that you and abarth use and I will add you to the committers list
08:26
<mhausenblas>
s/whatz/what's
08:26
<Hixie>
one quick google search later, and i have a vague idea
08:26
<mhausenblas>
right
08:26
<Hixie>
no opinion, not sure what the point is though
08:26
<Hixie>
what's wrong with mysql dumps?
08:26
<mhausenblas>
nothing wrong
08:26
<mhausenblas>
just requires a lot of manual work to intregrate
08:26
<Hixie>
i mean, atom is hardly the best format for transferring a terabyte of climate data
08:26
<mhausenblas>
good point
08:26
<Hixie>
this is the argument i have a problem with
08:26
<mhausenblas>
neither is RDF ;)
08:27
<Hixie>
sql "just requires a lot of manual work to intregrate"
08:27
<Hixie>
so does everythinge lse
08:27
<mhausenblas>
no
08:27
<Hixie>
because nobody is using the same vocabulary as everyone else
08:27
<mhausenblas>
Webish formats don't
08:27
<Hixie>
you still have to hardcode everything
08:27
<Hixie>
well anyway
08:27
<Hixie>
i have no problem with people trying to solve this problem
08:27
<Hixie>
i just don't think it's worth my time :-)
08:27
<mhausenblas>
hehe, fair enough
08:27
<Hixie>
better to spend the time making computers be able to do this via natural language processing, imho
08:27
<mhausenblas>
just an aside: even Google is going into this business
08:28
<mhausenblas>
(see Fusion Tables)
08:28
<Hixie>
yes, i'm friends with the program manager of the fusion tables team
08:28
<mhausenblas>
interesting
08:28
<mhausenblas>
and?
08:28
<Hixie>
and nothing :-)
08:28
mhausenblas
was waiting for some dirty background stories
08:28
<Hixie>
google does many things
08:28
<mhausenblas>
indeed
08:29
<mhausenblas>
anyway, thanks for your time Hixie - hope to proof you wrong in some 5-10y :D
08:29
<Hixie>
good luck :-)
08:29
<mhausenblas>
hehe, thanks, will need it
08:30
<mhausenblas>
and, MikeSmith if you find someone to drive this, plz lemme know
08:30
<MikeSmith>
will do
08:30
<mhausenblas>
tx!
08:30
<MikeSmith>
maybe kennyluck will have some ideas
08:30
<kennyluck>
What was the discussion?
08:32
<MikeSmith>
kennyluck: see the logs
08:32
kennyluck
is reading.
08:33
kennyluck
has the feeling that he spends too much time in the IRC world. :)
08:57
<kennyluck>
I have to say I am super surprised by the fact that Hixie is not that into the "machine-readible" idea.
08:57
<kennyluck>
Then what is Microdata for, I wonder.
08:57
<kennyluck>
Anyway, I go for a fundamental change to computer-science education.
08:57
<Hixie>
lots of people disagree with me
08:57
<Hixie>
microdata is for them, in case i'm wrong
08:58
<kennyluck>
A class on HTML should be required for every collegue student. :)
08:58
<kennyluck>
Hmm... who are they? Hixie? In this channel?
08:58
<Hixie>
kennyluck: mhausenblas, for one
08:59
<Hixie>
tantek is another
08:59
<Hixie>
all the people who sent in use cases for microdata
08:59
<Hixie>
microdata is one of the areas of the spec for which i was the most explicit in terms of writing the use cases i was aiming for
08:59
<Hixie>
(mostly because it was so hard to get anyone to give me actual use cases)
09:00
<kennyluck>
Ah ha.
09:01
<Hixie>
see http://lists.w3.org/Archives/Public/public-html/2009May/0207.html
09:01
<kennyluck>
I am just very interested in the origin of Microdata, I guess.
09:02
<Hixie>
that e-mail is the index to the e-mails that are the origin of microdata
09:02
<kennyluck>
Thank you, Hixie.
09:02
<Hixie>
it's interesting to note that most of the use cases for which microdata is designed are not the use cases that people are always trying to say RDF and co will save the world for
09:03
<Hixie>
but that's another story
09:03
<kennyluck>
I guess #swig people should have joined the discussion at the beginning with #whatwg, and we might have had some consesus or something.
09:03
<kennyluck>
Sure
09:03
<Hixie>
i invited -- begged for, even -- feedback for like a year on the whatwg list
09:03
<Hixie>
before doing anything
09:04
<Hixie>
it was like pulling teeth
09:04
<Hixie>
i've never had so much trouble getting people to tell me what they wanted
09:04
<kennyluck>
OK. I feel sorry, really.
09:04
<Hixie>
it was like everyone thought it was obvious that RDF would solve the world's problems and bring world peace, but nobody could articulate a single concrete use case
09:04
<Hixie>
anyway
09:05
<kennyluck>
I noticed that #swig consists of almost academics, so.
09:05
<kennyluck>
Well, anyway.
09:07
<kennyluck>
FYI, the RDB2RDF effort aligned with your SQL dump idea, I think, Hixie.
09:07
<kennyluck>
We need a standardized format for dumping, anyway.
09:07
<Hixie>
why?
09:08
<kennyluck>
Although ofcourse RDF might be a poor choice.
09:08
<kennyluck>
standardized format for dumping?
09:08
<Hixie>
yeah
09:08
<kennyluck>
Well, this is what standardization is for, isn't it?
09:08
<Hixie>
what's the use case for which you need a standardised format for dumping... what? any relational database?
09:09
<kennyluck>
Yeah.
09:09
<Hixie>
why are you going to need the same format for 2GB of climate data and 1PB of mars imagery and 100MB of number station logs?
09:10
<kennyluck>
Hmm.. to save time for processing?
09:10
<kennyluck>
Becuase we really like "mashups" to happen.
09:10
<Hixie>
you think the format part of this is going to be where the complexity lies?
09:10
<kennyluck>
In a large scale, so saving processing time is a good thing to do.
09:10
<Hixie>
this is what gets me about the whole semantic web
09:11
<Hixie>
it's solving one problem -- the syntax the data should be in -- despite that being the single simplest part of the entire problem
09:11
<Hixie>
it's trivial to write a custom parser for each format you want to mash up
09:11
<Hixie>
the hard part is the UI and the data processing once you've parsed the data
09:11
<kennyluck>
But since you are not likely to give the parser to other people, so...
09:11
<Hixie>
who cares what the format is -- that's at most an hour or two to parse the incoming data, and in all likelihood less since most data sources will already have libraries to parse their data formats
09:12
<Hixie>
that is, to write the parser to parse
09:12
<kennyluck>
This is arguable, I am not very sure.
09:13
<Hixie>
it's like looking at the problem of how to solve world hunger, and doing it by standardising the seat belts in the trucks that will be used to carry the food
09:13
<Hixie>
and then saying that the problem is not essentially solved
09:13
<Hixie>
s/not/now/
09:13
<kennyluck>
The Google Official Blog has an article explaining "machine-readability"
09:13
<kennyluck>
It's the data provider who understands the data well.
09:13
<Hixie>
yeah, not everyone at google agrees with me either :-)
09:14
<kennyluck>
So the data provider ought to convert the data to something that everyone can parse easiliy.
09:14
<Hixie>
sure, but that doesn't have to be anything standard
09:14
<Hixie>
they can just provide the data in a MySQL IASM data table for all i care
09:14
<Hixie>
then all you have to do is install MySQL and use that locally
09:14
<Hixie>
30 minutes of work assuming MySQL is already installed
09:15
<Hixie>
anyway
09:15
<kennyluck>
Maybe.
09:15
<kennyluck>
Anyway, thanks for your links.
09:15
<Hixie>
my point is just that that seems like such a trivial part of the problem space that i don't understand why it's the main thing everyone is always talking about
09:17
<kennyluck>
This is quite fundamental. Even with NLP, you still need a regular data structure to store your processed data.
09:17
<kennyluck>
The Semantic Web effort basiclly asks you to provide your NLP result, I guess.
09:17
<Hixie>
yes, but what is useful for one NLP system isn't going to be useful for another
09:18
<Hixie>
to put it in an obvious form: RDF isn't the ideal format for bitmaps, for example
09:18
<kennyluck>
Google's NLP system is certainly useful for everyone. :)
09:18
<Hixie>
so if your NLP system outputs a bitmap, then RDF isn't going to help
09:18
<Hixie>
if your NLP system happens to output a quad graph, then sure, it might help
09:18
<Hixie>
if it outputs a tree, then JSON might be better
09:18
<kennyluck>
Hixie, this is not likely to be the truth. (ref. NLP system outputs bitmaps)
09:19
<Hixie>
it's just an extreme example
09:19
<Hixie>
but it's not that unlikely, consider e.g. google image search
09:19
<Hixie>
or google goggles, whose input is a bitmap
09:19
<kennyluck>
Hmm...
09:20
<Hixie>
or even moving video
09:20
<kennyluck>
We are talking about outputs.
09:20
<Hixie>
or youtube's autotranscription -- input is audio, output is timed track data
09:20
<Hixie>
neither side of which is optimally represented by RDF
09:21
<Hixie>
different problems have different ideal formats
09:21
<kennyluck>
Sure. But timed track data could be serialized into RDF.
09:21
<Hixie>
sure
09:21
<Hixie>
so could a bitmap
09:21
<kennyluck>
Ofcourse, audios can not.
09:21
<Hixie>
but it's not the optimal form
09:21
<kennyluck>
Come on.
09:21
<Hixie>
and why would you use a suboptimal form?
09:21
<kennyluck>
Why do you use HTML when docx is better?
09:22
<Hixie>
i disagree with the premise of the question
09:22
<kennyluck>
Sorry.
09:22
<Hixie>
(a) that one should always use HTML, and (b) that DOCX is always better
09:22
<Hixie>
when you're writing a document in a word processor, HTML is not the ideal format, and using it would imho be a bad idea
09:22
<kennyluck>
I mean, this is what standardization is, isn't it?
09:22
<kennyluck>
Sure.
09:22
<Hixie>
what is what standardization is?
09:23
<kennyluck>
But Semantic Web is about asking people to build mashups.
09:23
<kennyluck>
Like HTML is about asking everyone in the world to view it.
09:24
<kennyluck>
If you don't want people to view it, you don't output HTML. If you don't want people to do mashup based on your data, you don't output RDF.
09:24
<Hixie>
I don't think RDF is useful for creating mashups personally
09:24
<Hixie>
and i don't know what "HTML is about asking everyone in the world to view it" means
09:24
<kennyluck>
Anyway, I welcome JSON as well.
09:25
<Hixie>
there's nothing special or useful about JSON compared to RDF
09:25
<Hixie>
they're both just formats, which may or may not be useful in any given scenario
09:25
<Hixie>
the difference is that JSON advocates don't sell JSON as the solution to humanity's woes
09:25
<kennyluck>
JSON is better than HTML for building mashups.
09:25
<Hixie>
i have no idea what that means
09:26
<Hixie>
that's like saying pears are better than shoes for drawing maps
09:27
<Smylers2>
Boots are even better if it's a map of Italy you're drawing — you could lay it flat and trace round it.
09:28
<kennyluck>
I just want to mention we benefit a lot from the fact that HTML is so universal.
09:28
<Hixie>
Smylers2: and pears are good if you're drawing a map of hungary
09:28
<Hixie>
Smylers2: since you can eat pears
09:29
<Hixie>
kennyluck: sure, but HTML is a vocabulary, not a data model
09:29
<kennyluck>
It's of course not an optimal format for lots of things.
09:29
<Hixie>
kennyluck: HTML is to XML what Foaf is to RDF
09:29
<kennyluck>
But I still write my presentation in HTML.
09:29
<Hixie>
kennyluck: Foaf has clear use cases
09:29
<Hixie>
kennyluck: just like RDF has clear use cases
09:29
<Hixie>
er
09:29
<Hixie>
kennyluck: just like HTML has clear use cases
09:30
<Hixie>
RDF and XML, however, are metaformats
09:30
<Hixie>
and alone are just tools
09:30
<Hixie>
like JSON
09:30
<Hixie>
or MySQL IASM
09:31
<Hixie>
advocating that everyone should use Foaf for the specific use cases that Foaf addresses makes a lot more sense than advocating that everyone should use RDF
09:31
<kennyluck>
Alright.
09:32
<kennyluck>
I always prefer this approach, actullay.
09:32
<kennyluck>
FOAF and then RDF.
09:32
<Hixie>
not FOAF and then RDF
09:32
<Hixie>
just FOAF
09:32
<Hixie>
that FOAF uses RDF is an implementation detail
09:32
<Hixie>
you can't generalise from "use XHTML" to "use XML"
09:33
<Hixie>
you can't generalise from "use FOAF" to "use RDF"
09:33
<kennyluck>
Probably.
09:33
<kennyluck>
Microdata also allows mixing namespaces. So it indeed meets these use cases.
09:34
<Hixie>
microdata is just another RDF serialisation
09:34
<Hixie>
(amongst other things)
09:35
<kennyluck>
Anyway, use cases are important, that I agree with you.
10:46
<phrearch>
morning
10:47
<phrearch>
i wondered if there is some sort of an agreed standard how to do routing in jsonrpc websockets?
10:48
<phrearch>
i could add uri data like /path/to/handler, but i thought there may be better ways to do this
10:50
<annevk>
jsonrpc websockets?
10:50
<annevk>
websockets is not even an agreed upon standard :)
10:51
<phrearch>
annevk: just thinking how to handle this. i like the django approach of handling urls. maybe i can use something simular for jsonrpc dispatching
10:52
<phrearch>
like calling functions like /path/to/function, and let the jsonrpc handler dispatch to the right function
10:53
<Rik`>
phrearch: http://substack.net/posts/85e1bd/DNode-Asynchronous-Remote-Method-Invocation-for-Node-js-and-the-Browser ?
10:55
<phrearch>
Rik`: thanks, ill take a look how they do things
10:58
<phrearch>
hm interesting stuff about RMI
10:59
<phrearch>
i still need some way to call a function from the server on the client
11:05
<Rik`>
isn't that what DNode provides ?
11:06
<phrearch>
sure, but i'm writing a python/twisted app
11:07
<phrearch>
i like the way they call remote functions in js
11:17
<MikeSmith>
annevk: thanks for the tweaks to the Media Queries spec
11:26
<MikeSmith>
I've been reading up on node.js
11:26
<MikeSmith>
seems like it's really turning into something very useful
11:26
<MikeSmith>
with a big and growing community around it
11:27
<Rik`>
MikeSmith: I find it very useful for my websockets demo :)
11:27
<MikeSmith>
:)
11:27
<Rik`>
(thanks to micheil btw)
11:28
<MikeSmith>
yeah
11:36
<annevk>
is ms2ger on IRC?
11:36
annevk
would like some rationale behind http://www.w3.org/Bugs/Public/show_bug.cgi?id=10164
11:41
<MikeSmith>
I've never seen ms2ger on IRC
11:41
<MikeSmith>
not under that name at least
11:41
<MikeSmith>
the identity of ms2ger seems to be a mystery
11:42
<MikeSmith>
maybe he/she is somebody who doesn't like realtime communication so much
11:56
<annevk>
hmm, does document.domain make sense in a document created through createDocument() ?
11:57
<annevk>
there's a concept of cookie-free Document objects but maybe that should be generalized
11:57
<annevk>
now document.cookie has that concept and document.domain specifically mentions XMLHttpRequest
11:57
<annevk>
seems somewhat wrong
12:28
<oal>
How would I create a modern, simple, wysiwyg editor for a website? Is an iframe with designMode and execCommand out dated?
13:15
<kennyluck>
Oh, Ms2ger just logged off.
13:44
<annevk>
getComputedStyle defined: http://dev.w3.org/csswg/cssom/#extensions-to-the-window-interface
13:45
<annevk>
finally, took like 3 years or so
13:45
<annevk>
where "3 years" is more like a week spread out over a long time with lots of learning in between, but still
13:46
<Philip`>
"the style rules associated with d." - s/d/doc/ ?
13:47
<annevk>
hmm, I should rename in all places at once
13:47
<annevk>
thanks
13:48
<annevk>
you're not in the acknowledgment list
13:48
<annevk>
weird
13:48
<annevk>
fixed
13:48
<Philip`>
I don't remember commenting on it before
13:49
<annevk>
yeah, it's not really too weird
13:49
<annevk>
there has not been much review at all
14:19
Ms2ger
waves at annevk
14:20
<annevk>
hey man
14:21
<annevk>
I see you added a comment :)
14:22
<Ms2ger>
Yeah, I'm just too lazy to file bugs through bugzilla
14:22
<annevk>
fair enough
14:23
<annevk>
you're patching Gecko these days?
14:23
<Ms2ger>
Yes
14:24
<Ms2ger>
And a bit of testing
14:25
<annevk>
nice
14:25
<annevk>
converging is taking quite some time, but it's actually happening these days, which is a nice change
14:26
<pleb1985>
I have tried googling but I gave up after 10 minutes and thought I could ask here instead: is video+audio live upstream possible with proposed html5-related technologies?
14:27
<annevk>
with proposed technologies yes
14:27
<annevk>
but nothing is actually implemented or done with respect to that
14:27
<pleb1985>
which should I be looking at?
14:28
<annevk>
I think you want to start by reading this email: http://lists.whatwg.org/htdig.cgi/whatwg-whatwg.org/2010-July/027129.html
14:28
<pleb1985>
presumably there's lots of butthurt going on around codecs etc?
14:29
<annevk>
no discussion yet actually
14:29
<pleb1985>
i'd use more mature terminology but I'm not sure that's really necessary/appropriate
14:29
<pleb1985>
;)
14:29
<annevk>
heh
14:29
<pleb1985>
is it likely there will be wg for that?
14:29
<annevk>
I hope we can just use VP8 but dunno
14:30
<annevk>
the other problem is network protocols
14:30
<pleb1985>
Ian is the web equivalent of neo from the matrix so wrt VP8:
14:30
<pleb1985>
I believe.
14:30
<annevk>
well, for your scenario I guess WebSocket+VP8+<device> is enough
14:30
<annevk>
so not much network protocol trouble
14:31
<annevk>
but P2P...
14:31
<pleb1985>
client server is fine :)}
14:32
<zcorpan_>
annevk: websockets does messages, not streams
14:32
<annevk>
zcorpan_, streams are just message blocks, no?
14:33
<zcorpan_>
annevk: on the network, yeah i guess, but the current api for websockets is not appropriate for sending streams
14:34
<annevk>
hmm
14:34
<annevk>
socket.send(Stream)
14:34
<annevk>
i think it'll work fine
14:34
<zcorpan_>
not with how it's currently defined :)
14:35
<annevk>
well it's not defined yet for Stream objects
14:35
<annevk>
but I'm pretty sure that is how it's going to work
14:36
<zcorpan_>
would you abort sending a stream if you invoke send() again?
14:36
<annevk>
I'm not sure it's worth to discuss the specifics now :)
14:38
<annevk>
And I suppose we could define a new API as well if that turns out to make more sense
14:39
<pleb1985>
socket's a bad name then eh? :)
14:53
<pleb1985>
annevk: i assume there's a lot of enthusiasm for <device> .. ?
14:53
<pleb1985>
well, everyone other than the flash camp ;)
14:55
<annevk>
pleb1985, people certainly want a feature like that, yes
14:55
<pleb1985>
.. and the UA guys?
14:56
<annevk>
I think we do too, but there's plenty of other stuff that requires attention too
14:56
<pleb1985>
for sure, I think you'll get more mileage from that than web sockets
14:57
<annevk>
I guess there are some uses for just having camera input and no dedicated network integration, but you really want network integration for most use cases
14:58
<pleb1985>
mmm, what sort of timescales would you expect on something like that?
14:58
<pleb1985>
year?
14:59
<annevk>
dunno really
14:59
<pleb1985>
fair enough :)
15:00
<annevk>
if we had a few implementors and a set of people hashing out the network protocol that would be likely, but so far there's not much of that
15:05
<pleb1985>
implementors on the server side?
15:06
<pleb1985>
or you mean UA side and by 'we' you mean whatwg
15:06
<pleb1985>
it's confusing. ;)
15:11
<annevk>
on the user agent side
15:12
<pleb1985>
fwiw, I'd like to see something happen there :)
15:14
<foolip>
I'm not sure I understand why <device> needs to be an element, is it expected to have any rendering or do anything useful without scripts involved?
15:15
<pleb1985>
I guess to ease detection ?
15:16
<foolip>
detection of what, by whom?
15:16
<pleb1985>
of the requirement for device connectivity
15:16
<pleb1985>
to the user
15:17
<annevk>
foolip, see the email
15:17
<foolip>
annevk, I read the email, after which I asked myself above question
15:17
<annevk>
foolip, Ian thinks this way you can design the best "interactive UI"
15:17
<pleb1985>
foolip: you're saying you would sink that type of GUI into the browser instead, similar to http auth?
15:18
<annevk>
foolip, i.e. that the user actually knows what he just did
15:18
<jgraham>
foolip: I thought the idea was to have some UI for accessing the device associated with the element
15:18
<foolip>
ok, so browser-specific (probably ugly) rendering of <device> ?
15:18
<annevk>
foolip, yes, like <input type=file>
15:18
<jgraham>
foolip: Something like input type="file"
15:19
<jgraham>
Well at least our story is consistent :)
15:19
<annevk>
where is this mercurial w3c thing?
15:19
<annevk>
plh wants me to create a XHR placeholder file
15:20
<annevk>
I wonder if I have an account there
15:20
<annevk>
if not this sounds like way too much work
15:20
<foolip>
If one is going to do video conferencing then surely one doesn't want an unstylable chunk of <device> ruining the design?
15:20
<foolip>
of course <input type=file> has the same problem
15:20
<annevk>
an infobar too
15:20
<annevk>
doesn't seem like such a big deal
15:22
<MikeSmith>
list of w3c mercurial repos is here:
15:22
<MikeSmith>
http://dvcs.w3.org/hg
15:22
<MikeSmith>
push access is by normal w3c username+password
15:23
<MikeSmith>
auth on that is per-WG, done by checking the w3c user groups db
15:23
<annevk>
oh
15:23
annevk
tried dcvs :)
15:24
<annevk>
now I have to learn hg
15:24
<annevk>
crap
15:24
<annevk>
quite a bit of make work imo
15:24
<jgraham>
http://hginit.com/
15:25
<jgraham>
It is _so_much_ better than CVS
15:25
annevk
installs mercurial
15:25
<Philip`>
jgraham: You could say that about pretty much anything
15:25
<jgraham>
Well worth it as soon as you have even the barest glimmer of understanding
15:25
<annevk>
I only use CVS as changelog
15:25
<jgraham>
Philip`: I do :)
15:25
<annevk>
and offsite storage facility
15:26
<MikeSmith>
here's the tutorial: hg clone, .. make changes.., hg commit, hg push ... then hg pull, .. make changes.. repeat
15:26
<MikeSmith>
oh, with some hg merge thrown in
15:26
<Philip`>
Unless someone else changes the repository, then you need to pull and up and merge if you already committed?
15:27
<jgraham>
It is pretty good about telling you what you need to do
15:27
<jgraham>
It might just take a while to get the mental model
15:27
<MikeSmith>
Philip`: no, not if nobody's changed anything
15:27
<annevk>
oh god
15:27
<jgraham>
annevk: What's the problem
15:28
<annevk>
complexity
15:28
<annevk>
so I created w3c-hg, I run hg-init in there?
15:28
<jgraham>
It's not that bad
15:28
<annevk>
euh, without the dash?
15:28
<jgraham>
You are cloning an existing repository
15:28
<jgraham>
You just run hg clone http://url/of/repo
15:29
<jgraham>
hg init is only needed to create one from scratch
15:29
<annevk>
so within w3c-hg I run hg clone http://dvcs.w3.org/hg/webapps/ ?
15:29
<annevk>
or should I create a folder webapps first?
15:29
<annevk>
I guess I should
15:29
<jgraham>
No need to create the folder first
15:29
<annevk>
ok
15:30
<annevk>
how do I add a folder via hg?
15:30
<jgraham>
Just do hg clone
15:30
<MikeSmith>
annevk: you can't create top-level folders
15:30
<jgraham>
You mean add a folder to the repository?
15:30
<annevk>
yes
15:30
<MikeSmith>
we have to ask systems team to add one
15:30
<annevk>
MikeSmith, inside webapps
15:30
<jgraham>
MikeSmith: Really?
15:30
<MikeSmith>
annevk: oh
15:31
<jgraham>
annevk: Just create a folder and do hg add /path/to/folder
15:31
<MikeSmith>
didn't knwo PLH had added that
15:31
<MikeSmith>
jgraham: can't create them there at least
15:31
<MikeSmith>
but any webapps WG member can create within the webapps folder
15:31
<Philip`>
Or just hg add /path/to/folder/file and it'll automatically add the folder if needed
15:31
<jgraham>
MikeSmith: That's creating whole repositories isn't it?
15:32
<MikeSmith>
jgraham: yeah
15:32
<jgraham>
MikeSmith: Right, I would imagine that is harder :)
15:33
<variable>
http://hginit.com/ --> a decent hg tutorial
15:34
<annevk>
I did hg commit but hg push doesn't work
15:34
<annevk>
method not allowed
15:35
<Philip`>
Has it asked for a username/password yet?
15:35
<annevk>
at one point during commit and I provided my W3C username
15:35
<annevk>
didn't ask for anything since
15:35
<annevk>
password it might pull out of .ssh?
15:35
<jgraham>
Commit shouldn't have asked for username
15:35
<annevk>
just like it does for CVS?
15:36
<jgraham>
Since it only changes things in your local repository
15:36
<annevk>
well, it did the first time
15:36
<Philip`>
I've got it set up with ~/.hgrc containing:
15:36
<Philip`>
[auth]
15:36
<Philip`>
w3c.prefix = dvcs.w3.org/hg/
15:36
<Philip`>
w3c.username = blahblah
15:36
<Philip`>
w3c.password = blahblah
15:36
<jgraham>
That is very surprising, no?
15:36
<MikeSmith>
annevk: it doesn't use ssh at all
15:36
<annevk>
i have no idea
15:36
<annevk>
i don't know how this works
15:37
<annevk>
but i'm learning slowly
15:37
<jgraham>
annevk: You have a copy of the whole repository and all the history locally
15:37
<jgraham>
Making a checkin changes your local copy of the repository
15:38
<annevk>
sure sure
15:38
<annevk>
i'm just saying what happened
15:38
<jgraham>
Later you push the changes to the remote
15:39
<annevk>
Philip`, added that file, still failing in the same way
15:40
<Philip`>
It might be that you need the https:// URL
15:40
<Philip`>
rather than http://
15:41
<annevk>
hmm
15:41
<Philip`>
You can just edit .hg/hgrc in the checkout directory
15:41
<Philip`>
(I think)
15:41
<Philip`>
to change to https
15:41
<jgraham>
You can
15:42
<Philip`>
(Unlike SVN, which needs some crazy --relocate thing that rewrites your whole working copy)
15:42
<annevk>
it asks for username again...
15:43
<annevk>
(I started over)
15:43
<annevk>
but now it worked
15:43
<annevk>
is there a version where all this is checked out?
15:44
<annevk>
hg serve?
15:44
annevk
finds http://test.w3.org/webapps/
15:44
<MikeSmith>
http://dvcs.w3.org/hg/webapps/file/eda29ce012e3/XMLHttpRequest/
15:44
<MikeSmith>
http://dvcs.w3.org/hg/webapps/raw-file/eda29ce012e3/XMLHttpRequest/info.htm
15:45
<annevk>
but that's a static link no?
15:45
<annevk>
where's the latest version?
15:45
<Philip`>
http://dvcs.w3.org/hg/webapps/file/tip/XMLHttpRequest/ if you want the latest
15:45
<annevk>
so what's test.w3.org/webapps/ ?
15:46
<annevk>
so http://dvcs.w3.org/hg/webapps/raw-file/tip/XMLHttpRequest/info.htm
15:46
<annevk>
hmm
15:48
<annevk>
how do you move files?
15:48
<Philip`>
hg mv
15:51
<annevk>
god it requests my username for each commit
15:51
<annevk>
but not for hg push
15:51
<annevk>
silly as hell
15:55
<Philip`>
You probably want to edit ~/.hgrc again, to say
15:55
<Philip`>
[ui]
15:55
<Philip`>
username = Blah Blah <blah⊙bb>
15:55
<Philip`>
so it knows what name to give you in the commit logs
15:56
<annevk>
can be a different username?
15:56
annevk
tries
15:56
<Philip`>
It can be anything
15:56
<Philip`>
I don't know if it has to be an email address but that seems conventional
15:57
<annevk>
you win the interwebs
15:57
<annevk>
i did annevk <annevk⊙oc>
15:58
<Philip`>
You can put other people's names in commits to your local repository, and when you push to the W3C repository it'll just trust the commits that you're pushing
15:59
Philip`
doesn't know whether there's anywhere that records who actually pushed a change
17:18
<oal>
I asked this earlier, but I didn't get any response: If I want to make a simple wysiwyg/rich editor with html5-ish technology, where should I start? Are iframes and execCommand out dated?
17:36
<boblet>
oal: things should pick up in ~4 hours, maybe try asking again in a bit
17:37
<oal>
boblet, ok thank you :)
17:37
<oal>
At least it was a reply ;)
17:37
<boblet>
sorry I can’t help
17:37
<boblet>
heh
17:41
<AryehGregor>
oal, I'm not an expert, but contenteditable, designMode, execCommand, this is what you should be looking at, yes.
17:42
<AryehGregor>
(you can do much more complicated stuff using canvas, but you'd have to implement it all from scratch; this is what Bespin does)
17:42
<AryehGregor>
Hah, Google thinks the contents of Bespin's website is "Loading...": http://www.google.com/search?q=bespin
17:42
<AryehGregor>
That's what you get for not having fallback.
17:42
<oal>
AryehGregor, I've been experimenting with a div+contenteditable, but I'm unable to get selecting/applying effects correctly
17:43
<AryehGregor>
I don't know how interoperably implemented all this stuff is in practice.
17:43
<AryehGregor>
I do know that it's what people generally use for this.
17:43
<AryehGregor>
If you had a specific question, maybe you could ask that, someone might know the answer.
17:44
<Philip`>
oal: Why do you not want to reuse one of the existing WYSIWYG HTML editors?
17:45
<oal>
Philip`, because I need something that can "float" on an existing page, not with static, locked toolbars etc
17:46
<oal>
And I don't need all the features of CKeditor and the "big" ones
18:03
<oal>
Should I use <strong> or <b> in html5?
18:04
<Philip`>
Depends on the semantics of your content
18:05
<oal>
Philip`, in which ways?
18:05
<Philip`>
They mean different things, as defined by the spec
18:05
<oal>
Ok, I'll look it up
18:15
<boblet>
oal: http://html5doctor.com/i-b-em-strong-element
18:16
<boblet>
Philip`: “Why do you not want to reuse one of the existing WYSIWYG HTML editors?” because they’re all unmitigated shite?
18:16
<blue_oak>
Does anyone know why the string to number and number to string conversion algorithms are written as they are when the input element is in the Date and time state?
18:16
<boblet>
well partially mitigated by trying to support older browsers
18:17
<boblet>
but anything using Word as a UI metaphor to emulate should be put down (mercy killing)
18:17
<oal>
boblet, thanks. :)
18:17
<boblet>
oal: heh, np. always easy to recommend your own stuff ;-)
18:18
<oal>
bobchao, is that your site?
18:18
<bobchao>
oal: boblet I think? :)
18:18
<boblet>
oal: I’m one of the authors yep
18:18
<boblet>
bobchao: namespace collision!
18:18
<oal>
bobchao, sorry mate. Tab doesn't always react the way I want :P
18:19
<TabAtkins>
xmlns=http://www.example.com/names#bob
18:19
<bobchao>
boblet: true, haha actually this is the second time :P
18:19
<oal>
boblet, looks like a good source for html5 stuff. thank you
18:19
<boblet>
heh. I have the same problem with mike][inq and MikeSmith
18:20
<boblet>
oal: it’s a superlative resource! and I don’t only say that because some of the articles are mine ;-)
18:20
<boblet>
honest!
18:20
<oal>
Yes, I'll definitely use it again :)
18:24
<boblet>
oal:if you can make something that’s not
18:24
<boblet>
woops
18:24
<oal>
Wrong channel?
18:25
<boblet>
oal: if you can make something that outputs decent code and doesn’t look as ugly as a MS product, I’ll use it too
18:25
<boblet>
fat-fingered
18:26
<oal>
If I only knew how I should solve all the issues I stumble upon, I might get a decent result, but now things are going sloow...
18:27
<boblet>
oal: hopefully you’ll choose to only support modern browsers, as I’m sure the millstone of ahem older ones doesn’t help current projects any
18:28
<boblet>
oal: one I liked was wmd http://wmd-editor.com/
18:28
<oal>
boblet, definitely will focus most on newer browsers
18:29
<boblet>
oal: regardless I think you’ll need lots of luck (it’d be a touch project) so good luck!
18:29
<boblet>
:)
18:29
<oal>
I know. If I get things going, the goal is to make a simple editor, not anything fancy
18:34
<dandaman>
im trying to set a thing border that will surround my content on my site automatically(so it will depend on the viewer's resolution) is there a way to do that?
18:34
<dandaman>
<body style="border:thin solid #003366;width: 150px">
18:34
<dandaman>
kinda like that
18:34
<dandaman>
except the width not being set
18:35
<TabAtkins>
body { border: thin solid #036; margin: 0 20px; } ?
18:35
<dandaman>
in css
18:35
<dandaman>
?
18:35
<TabAtkins>
Yes. It's all CSS.
18:36
<dandaman>
it doesnt hug it perfectly
18:36
<dandaman>
but oh well, it works
18:36
<TabAtkins>
I dunno quite what you're exactly asking for.
18:37
<dandaman>
well im doing this for a mobile app
18:37
<dandaman>
so the screen size makes it work
18:37
<dandaman>
thanks
18:37
<TabAtkins>
O...k?
18:38
<AryehGregor>
Why margin: 0 2px; instead of margin: 0;?
18:39
<TabAtkins>
I had no idea what he was asking for, so I went with some margins.
18:39
<TabAtkins>
If he just wanted a literal border on the edges, then yeah, margin:0;.
18:40
<dandaman>
on the edges of the text
18:40
<dandaman>
not on the edges of the browser
18:40
<TabAtkins>
Oh, you want it to wrap whatever you've got?
18:40
<TabAtkins>
Then margin:0 auto;
18:43
<dandaman>
thanks
18:44
<dandaman>
auto did not work :\
18:44
<TabAtkins>
No, it works. You may be doing something preventing it from working.
18:44
<TabAtkins>
Can you provide a live site and a description of what precisely you're looking for?
18:46
<dandaman>
let me try....
18:46
<dandaman>
http://www.w3.org/TR/REC-CSS1/
18:46
<dandaman>
i guess see how the note is being bordered
18:46
<dandaman>
by that red
18:46
<TabAtkins>
(Also, for future reference, #css is a better channel for these sorts of things.)
18:46
<dandaman>
i want basically that
18:46
<dandaman>
except a border all over everything
18:46
<dandaman>
that fits snugly
18:46
<dandaman>
currently i have all my text centered
18:46
<dandaman>
and the border reaches accross the entire screen
18:47
<TabAtkins>
Yes.
18:47
<dandaman>
which i dont want
18:47
<dandaman>
i want it only close to the text
18:47
<TabAtkins>
Set display:table-cell on the body as well.
18:48
<TabAtkins>
Or rather, display:table, since you still want the margin:0 auto; to apply.
18:48
<dandaman>
AWESOME
18:48
<dandaman>
THANKS
18:48
<dandaman>
sorry caps
18:48
<dandaman>
but you win
18:48
<TabAtkins>
(Eventually you'll be able to use width:fit-content;, but for now you can only do it in FF with width:-moz-fit-content;.)
18:49
<TabAtkins>
Note: won't work on IE7 or earlier.
18:50
<dandaman>
hmm, is there a way to have a different background color within the border and a different bg color outside of the border?
18:50
<TabAtkins>
Set one background on <body>, and one on <html>.
18:51
<dandaman>
good call
18:56
<dandaman>
when i go <html background="#FFFFFF"> it doesnt work, the body style= background takes over the whole page and <html style= "background: #FFFFFF"> makes the page go blank (and when i hit view source there is nothing there)
18:56
<dandaman>
I am using grails to run the server btw
18:56
<dandaman>
dunno if that makes a difference
18:57
<TabAtkins>
Oh, right, sorry. The <body>'s background is hoisted to the entire viewport, so that won't work.
18:57
<TabAtkins>
In that case, you'll have to create a wrapper, move your current body styling to it, then set the "back" background on <body>.
18:58
<TabAtkins>
Well, hum. Actually, setting backgrounds on <html> and <body> work fine for me in Chrome.
18:58
<dandaman>
just throw a <div> around everything?
18:58
<TabAtkins>
Worst case, yeah.
18:58
<dandaman>
yeah im doing this on chrome
18:59
<TabAtkins>
Hmm, then. If I explicitly set a background through <html @style> and <body @style>, they both take effect exactly as expected.
18:59
<dandaman>
div worked
19:00
<dandaman>
might just be grails
19:00
<TabAtkins>
That would be... bizarre. A server-side framework shouldn't have any effect on how client-side CSS works, unless it's inserting additional CSS without you knowing.
19:01
<TabAtkins>
And when I say "shouldn't" I mean "actually can't, because it's physically impossible".
19:01
<AryehGregor>
My website sets separate backgrounds on <html> and <body>.
19:02
<TabAtkins>
Yeah, I'd messed up. <body>'s background gets hoisted to the viewport *if* <html> doesn't have a background.
19:16
TabAtkins
just typed </meh> instead of </em>.
19:16
<TabAtkins>
I think that's the exact opposite semantic.
19:18
<Workshiva>
</meh> could complement </sarcasm>
19:43
boblet
is <meh>
19:43
<boblet>
nn all
19:48
<hsivonen>
gsnedders: I saw Avenue Q. It was fun, but it didn't quite live up to the expectations (the best songs were before the intermission and it kinda flattened out from there)
20:48
<hober>
anyone going to oscon this year?
21:01
<zcorpan_>
Hixie: r5162, i think it also affected <option>
21:32
<AryehGregor>
Blast Opera's nonsensical treatment of # in data URLs.
21:40
<zcorpan_>
not supporting fragments in data urls is what's nonsensical
21:41
<cardona507>
saying Opera Mini loads quicker than the stock android browser is whats nonsensical. :P
21:42
<AryehGregor>
zcorpan_, I have so far had documents in data URLs randomly break several times in Opera and Opera only because of its fragment treatment. I have never wanted to actually use a fragment in a data URI that I can recall.
21:44
<AryehGregor>
Okay, I just tried drawing shadows with canvas, and got three significantly different results in three different browsers.
21:44
<AryehGregor>
I have no idea who's actually right, or if the spec is just vague.
21:45
<AryehGregor>
It doesn't look vague to me, unless Gaussian blurs aren't well-defined.
21:45
<cardona507>
different results in different browser?! I am shocked :)
21:45
<TabAtkins>
The spec for <canvas>'s shadows is basically just "whatever Safari did", so it should be fine.
21:45
<cardona507>
*browsers
21:45
<AryehGregor>
TabAtkins, http://www.whatwg.org/specs/web-apps/current-work/multipage/the-canvas-element.html#shadows
21:46
<AryehGregor>
<!doctype html> <canvas height="300" width="300"></canvas> <script> window.addEventListener('load', function () { context = document.getElementsByTagName("canvas")[0].getContext("2d"); context.shadowBlur = 50; context.shadowColor = 'black'; context.fillRect(100, 100, 100, 100); }, false); </script>
21:46
<TabAtkins>
AryehGregor: No, that's well-defined. Any problem with it is a browser bug.
21:46
<AryehGregor>
In which browser, and how do I tell?
21:46
<AryehGregor>
I'm guessing they implemented this feature before it was specced.
21:46
<arosenberg>
I'll retry posting my thought from the other day.
21:47
<TabAtkins>
Calculate the gaussian matrix yourself from the appropriate stdev, run it manually, then diff against the browser?
21:47
<AryehGregor>
IIRC, the algorithm follows WebKit, so maybe I should assume everyone but them is wrong.
21:47
<TabAtkins>
That's probably correct.
21:47
<arosenberg>
There seems to be no way to get the response code or headers from an injected script tag.
21:48
<AryehGregor>
The treatment of the blur parameter is at least slightly insane. Maybe it should be cleaned up in the spec, since clearly there's no interop here anyway.
21:48
<arosenberg>
And yet the twitter API returns results in the headers with a 400 response for rate limiting.
21:48
<TabAtkins>
It is immensely insane, in fact.
21:48
<arosenberg>
XHR has provisions for this info, and I think it should be addressed for script tags as well.
21:49
<arosenberg>
(BTW, I'm thrilled that onerror now gets called for a script tag.)
21:49
<TabAtkins>
AryehGregor: Since we're discussing what precisely a "blur value" means in CSS right now (since every single browser treats it differently), it should sync up with whatever we decide for that.
21:49
<AryehGregor>
TabAtkins, this is in the context of a post I was making there.
21:49
<AryehGregor>
(Violently disagreeing with you, as it happens.)
21:49
<TabAtkins>
Ah, haven't read the thread yet today.
21:49
<TabAtkins>
I'm saving it for last.
21:50
<AryehGregor>
arosenberg, try posting to the whatwg list, it's a better way to get a response.
21:50
<zcorpan_>
arosenberg: xhr has same-origin policy, script doesn't
21:51
<TabAtkins>
AryehGregor: Ah, you haven't actually sent the post yet.
21:51
<AryehGregor>
No.
21:51
<zcorpan_>
arosenberg: so being able to read headers from arbitrary sources cross-origin would be a new security problem
21:52
<zcorpan_>
arosenberg: i guess it could be exposed for same-origin only, but what's the use case?
21:52
<Philip`>
AryehGregor: It doesn't follow WebKit, it follows CoreGraphics
21:53
<Philip`>
i.e. what Safari does
21:53
<Philip`>
not (necessarily) any other WebKit ports
21:53
<AryehGregor>
Oh, I see.
21:53
<AryehGregor>
So Chrome won't match WebKit here.
21:53
<AryehGregor>
Er.
21:53
<TabAtkins>
Mac/Safari, to be precise.
21:53
<AryehGregor>
Won't match Safari
21:53
<AryehGregor>
.
21:53
<Philip`>
since WebKit just defers to the graphics layer for almost all of this
21:53
<Philip`>
TabAtkins: Win/Safari too, since that uses CG
21:53
<TabAtkins>
Oh, CG is one Win too?
21:53
TabAtkins
knows nothing.
21:54
<TabAtkins>
AryehGregor: What's the gist of your violent disagreement?
21:54
<AryehGregor>
TabAtkins, the spec should say something pixel-perfect even if that means it's author-incomprehensible. Authors can figure out what values they like quickly by trial and error.
21:55
<AryehGregor>
They have to do that anyway, reading the spec won't tell them by itself.
21:55
<AryehGregor>
(even if it gives a general idea of how much blur to expect)
21:57
<TabAtkins>
AryehGregor: I am *potentially* okay with pixel-perfect spec descriptions. What I'm *not* okay with is the actual component value provided by the author being an opaque number, when we can just as easily make it sensical.
21:57
<AryehGregor>
If anyone has the latest IE9 handy, I'd be interested to hear what that canvas looks like.
21:58
<AryehGregor>
Does it match the SVG here? http://www.w3.org/Bugs/Public/show_bug.cgi?id=10169#c1
21:58
<Philip`>
http://test.w3.org/html/tests/submission/PhilipTaylor/canvas/index.2d.shadow.blur.html
21:58
<Philip`>
Opera (on Linux) looks correct, I assume Safari is still correct, Firefox (3.6 on Linux) looks all wrong
21:59
<AryehGregor>
On Linux, Opera looks correct on both, Chrome looks wrong on both, Firefox 4 looks wrong on small blurs but okay on big blurs.
21:59
<AryehGregor>
Actually, it looks wrong on big blurs too, but not so drastically.
21:59
<AryehGregor>
What does IE9 look like? I'm betting it's correct.
22:00
<arosenberg>
zcorpan_: I haven't carefully considered the security implications, but I don't see the differences between a script tag's arbitrary source ability and the headers from same.
22:00
<Philip`>
(Search for "sigma" in http://test.w3.org/html/tests/submission/PhilipTaylor/tools/canvas/tests2d.yaml which is how the bottom images in my tests are generated)
22:00
<Philip`>
(which hopefully matches the spec's requirement)
22:02
<AryehGregor>
Philip`, do you think my markup in these two cases is supposed to display the same? http://www.w3.org/Bugs/Public/show_bug.cgi?id=10169#c1
22:02
<AryehGregor>
As far as I can tell, it's supposed to be a 100x100px black box offset 100x100px with five sigmas of blur in both cases.
22:02
<AryehGregor>
Hmm, no, I miscalculated for canvas?
22:02
<AryehGregor>
sqrt(100) = 10, not 5.
22:02
AryehGregor
tries again
22:02
<Hixie>
zcorpan_: how?
22:04
<zcorpan_>
Hixie: <body><option></option>
22:04
<Philip`>
AryehGregor: The canvas one is a solid box on top of a blurred box, the SVG one looks like simply a blurred box
22:04
<Hixie>
oh right because we made </option> not so special after all
22:04
<Hixie>
hmm
22:04
<Hixie>
damn
22:04
<Hixie>
i shouldn't have put it in the list of special tags then
22:04
<AryehGregor>
Hmm.
22:04
<Philip`>
(where by "looks" I mean from the source code, not the rendering)
22:05
<Philip`>
You could draw the canvas's box at x=-200 with shadowOffsetX=200
22:05
<Philip`>
Uh, 300
22:05
<Philip`>
or whatever
22:05
<Philip`>
to get just the shadow rendered
22:06
<AryehGregor>
Hmm, good idea.
22:06
<Hixie>
i would really appreciate it if someone could check http://html5.org/tools/web-apps-tracker?from=5163&to=5164 with a fine tooth comb to make sure i didn't screw up anything
22:07
<AryehGregor>
Philip`, nice, that works perfectly. Thanks!
22:08
<AryehGregor>
Pixel-perfect.
22:13
<AryehGregor>
Should I file a bug against Firefox?
22:13
AryehGregor
doesn't see an existing one, unless it's: https://bugzilla.mozilla.org/show_bug.cgi?id=478459
22:15
<eseidel>
Hixie: you are a machine
22:15
<eseidel>
zomg it's a hsivonen!
22:16
<eseidel>
hsivonen: and here I thought you were avoiding me :p
22:16
<annevk>
oh god
22:16
<annevk>
tyler close wtf
22:16
<annevk>
guess it's time for bed
22:17
<annevk>
eseidel, lol
22:17
<eseidel>
nah, hsivonen is still avoiding me :p
22:19
<Hixie>
eseidel: once i get started on bugs i can crank through them easily yeah :-)
22:20
<Hixie>
eseidel: check out my downhill gradients on http://www.whatwg.org/issues/data.html :-)
22:20
<Hixie>
(blue line is bugs)
22:22
<Philip`>
AryehGregor: Filing bugs is good :-)
22:22
<Philip`>
AryehGregor: unless you want browsers to converge on a behaviour that's different to what's currently specced
22:23
<AryehGregor>
Philip`, I already filed an HTML5 bug for that. :)
22:24
<eseidel>
Hixie: I was just commenting to abarth that it would be nice if there was a version of the spec which showed diffs since a revision/date. I'm curious if such exists already? Basically I want a version of the spec with red/green in it to show me what has changed since I implemented it (i.e. at a specific date or revision).
22:24
<eseidel>
a diff sorta gives me that, but I want full context of the spec
22:24
<eseidel>
I'm sure I could generate such a thing myself from svn, but perhaps it already exists?
22:28
<Hixie>
eseidel: the spec is >5 MB, I'm not aware of any HTML diffing tool that can handle it
22:28
<eseidel>
ha
22:28
<Hixie>
eseidel: the tracker page lets you see diffs in the source between two arbitrary revisions, though
22:29
<zcorpan_>
you could generate an html diff for the a multipage page
22:29
<zcorpan_>
s/the//
22:31
<Hixie>
i think it would be really cool if someone could make a version of the spec that let you see diffs between versions and let you see inline blame annotations for everything so you could navigate the spec's history and changes
22:31
<Hixie>
but i'm not gonna do it myself :-)
22:32
<variable>
Hixie, the spec is stored via svn?
22:32
<annevk>
yeah, svn.whatwg.org/webapps/source
22:32
<Hixie>
what anne said
22:32
<variable>
then I know there is already a program that does that
22:32
<variable>
svn has a native command line version and I know I've seen a web version
22:33
<variable>
right - WebSVN
22:37
<Slaanesh>
Man, chrome is such a pain... clearing the location bar after each javascript command
22:38
<zcorpan_>
Slaanesh: yeah, what's up with that
22:38
<Slaanesh>
Let's blame Hixie
22:39
<TabAtkins>
Slaanesh: Hit F12, then Esc if the console isn't up yet.
22:39
<Hixie>
variable: i meant some sort of dynamic "web 2.0" "ajax" app version of the spec, not the spec source, but websvn would certainly be a step in the right direction
22:40
<zcorpan_>
s/"web 2.0" "ajax"/html5/
22:40
<Slaanesh>
TabAtkins: I assume you're trying to enable the jsconsole, but that doesn't work here, since it has superpermissions
22:41
TabAtkins
didn't realize that permissions mattered in any way here.
22:41
<variable>
I always "web 2.0" to mean "user content"
22:41
<variable>
Hixie, what particular features did you want fancy animations when the user clicks "diff"?
22:42
<Slaanesh>
Well, F12 also doesn't open it :P
22:42
<Hixie>
i meant something where you would be reading the spec and as you hovered over the spec it would highlight each fragment that had been updated, and in the margin would just the most recent revision that had that change, or something, and you could click on it to switch to that revision, or to see the diff, or something
22:42
<Slaanesh>
shift-ctrl-j, apparently
22:42
<Hixie>
i really have no concrete idea here :-)
22:42
<eseidel>
Hixie: it seems minefield and webkit both blindly coalesce text into the previous node, ignoring the "if it was the last node inserted by the parser" check for character token insertion
22:42
<eseidel>
example: <table>A<td>B</td>C</table>
22:42
<eseidel>
the spec says that should be two separate notes "A" and "B"
22:42
<variable>
Hixie, ah, like built into the current spec with the fancy commenting and such
22:43
<eseidel>
both minefield and WK make "AC"
22:43
<eseidel>
sorry "A" and "C" not "A" and "B"
22:44
<Hixie>
variable: yeah something
22:45
<eseidel>
Hixie: I'm attempting to test IE now.
22:45
<Hixie>
eseidel: you mean with hsivonen's parser or the old one?
22:45
<eseidel>
minefield, so new parser
22:46
<Hixie>
eseidel: i seem to recall there were pretty important performance reasons for not doing that, but ok
22:46
eseidel
wonders if there is a version of live dom viewer or plexode which works in IE
22:47
<eseidel>
hmm, maybe
22:50
<Hixie>
e.g. it lets you reuse your buffers in your parser without having to worry about being compatible with the buffers your CreateTextNode() implementation uses
22:50
<Hixie>
and it means you only have to have one "active" buffer at a time
22:51
<Hixie>
so you don't run the risk of having to increase the length of your buffer and finding that you can't just allocate more RAM, you have to move the entire buffer around first
22:52
<eseidel>
Hixie: I'm not sure I understand
22:52
<Hixie>
which bit?
22:52
<eseidel>
once you've create a text node, it can't be using parser buffers
22:52
<eseidel>
you might have to make the text node's buffers bigger if you append to it, yes.
22:53
<eseidel>
Hixie: I mena, there are some nice things about the way the spec is worded. prevnets strange things when scripts start inserting text nodes
22:53
<eseidel>
it won't be hard to implement
22:53
<eseidel>
Hixie: just noting that both WK and minefield fail that example atm
22:53
<eseidel>
Hixie: but I'm not against implementing it that way
22:53
<Hixie>
ah ok
22:53
<Hixie>
i misunderstood, sorry :-)
22:54
<Hixie>
usually when people tell me the browsers don't match the spec it's a passive aggressive way of telling me the spec is wrong :-)
22:54
<abarth>
svn.whatwg.org is slow...
22:55
<abarth>
soon i will have my own clone of the HTML5 repo and there will be no stopping me :)
22:58
<Hixie>
who is eric⊙wo? is that eseidel?
23:00
<jgraham>
I thought hsivonen was one of the people asking for the multiple-test-nodes thing before
23:00
<jgraham>
But I could be wrong. I recall having a conversation with Philip` about it
23:00
<abarth>
Hixie: yes
23:00
<eseidel>
Hixie: there can be only one!
23:01
<abarth>
we're looking at the details of text node coalessing
23:01
<abarth>
looks like the new Firefox doesn't quite follow the spec
23:01
<abarth>
in crazy foster parenting cases
23:01
<Hixie>
eseidel: i couldn't understand one of your bugs, so i marked it NEEDSINFO
23:02
<zcorpan_>
eseidel: try the zombie dom viewer for ie
23:02
<Hixie>
eseidel, abarth, hsivonen, zcorpan_, jgraham: I've resolved every parser bug I could find and every bug marked P1/crit; if I missed any please mark them P1/crit and let me know
23:03
<abarth>
Hixie: thanks. that's super helpful
23:03
<Hixie>
np
23:03
<Hixie>
sorry it took so long
23:03
<jgraham>
Hixie: Awesome
23:03
<abarth>
Hixie: did you look at the script@onload bug after you closed it?
23:03
<abarth>
Hixie: i'm not sure whether i'm misunderstanding the spec or whether your rationale is backwards
23:04
<abarth>
http://www.w3.org/Bugs/Public/show_bug.cgi?id=9984
23:04
<jgraham>
I verified that the end tag in foreign content fixes worked the way I expected in the html5lib test cases
23:04
<jgraham>
I mean I verified that html5lib with the new spec passed the tests
23:05
<Hixie>
jgraham: cool, i wasn't sure if what i did there made sense, thanks
23:05
<Hixie>
abarth: looking
23:05
<jgraham>
Hixie: Did hsivonen file a bug about <button>? I forget
23:05
<Hixie>
didn't see one
23:05
<jgraham>
Hixie: I think it was more-or-less equivalent to my ad-hoc fixes, but expressed in a slightly simpler way
23:06
<jgraham>
Hixie: The list of bugs that I know about is http://wiki.whatwg.org/index.php?title=ParserIssues
23:06
<Hixie>
abarth: hm, no, you're right, i forgot how the spec was written
23:06
<Hixie>
abarth: i don't understand the bug then
23:06
<Hixie>
abarth: what do you want changed?
23:06
<Hixie>
jgraham: looking...
23:07
<jgraham>
Hixie: It seems that you did change <button> somewhat
23:07
<Hixie>
abarth: (btw feel free to reopen bugs -- i don't see changes unless the bug is reopened)
23:07
<Hixie>
jgraham: all those bugs are resolved
23:07
<abarth>
Hixie: it's a place where FF and the spec are different. i don't particularly care about the behavior, but we either need to change the spec or get henri to agree to do what the spec says
23:08
<Hixie>
jgraham: and the e-mail had a bug for it too so that's dealt with now also
23:08
<Hixie>
abarth: ah ok
23:08
<abarth>
there's some subtly involving the requirements of an off-the-main-thread parser, which I don't quite grasp
23:09
<Hixie>
i'm not sure how we would change things to have it happen the other way around to be honest, looking at the spec more closely now
23:10
<Hixie>
short of just making the insertion pointer or whatever it's called undefined just for the event
23:11
<abarth>
ok
23:12
<abarth>
in the code, it's just a matter of switching the order of two lines of code
23:12
<abarth>
so we can do whichever
23:12
<daedb_>
heh, IE9 is really funny... <!doctype html><canvas></canvas> puts the canvas inside the body, but <!doctype html><style></style><canvas></canvas> puts canvas inside the head :)
23:12
<abarth>
Hixie: one more thing: https://bugs.webkit.org/show_bug.cgi?id=42112
23:12
<jgraham>
abarth: Thanks for the reply about the cross-domain stuff. That looks like exactly what I was hoping for :)
23:12
<AryehGregor>
daedb_, yeah . . . HTML5 parser plz :(
23:12
<abarth>
Hixie: ap is worried about the null character's replacements looking ugly
23:13
<abarth>
Hixie: that bug has a link to some chinese site that looks pretty bad
23:13
<cardona507>
daedb_: yeah - IE9 is wacky
23:13
<daedb_>
AryehGregor: I found that with your test case :p
23:13
<abarth>
jgraham: glad it was helpful
23:13
<abarth>
Hixie: the old webkit behavior was to strip nulls in text nodes, but not in tag names, etc
23:13
<jgraham>
abarth: (I haven't actually read it in detail yet)
23:16
<sicking>
if I change something in browser/components/feeds , where do I need to rebuild to pick that up?
23:16
<abarth>
sicking: wrong channel?
23:16
<sicking>
abarth: hah, yes indeed :)
23:17
<sicking>
chatzilla doesn't have optimal focus management :)
23:19
<Hixie>
abarth: yeah, it's basically intentional
23:19
<Hixie>
abarth: there shouldn't be NULs there at all and they're likely to be triggering all kinds of security problems
23:19
<Hixie>
abarth: making it ugly is one way to draw attention to it
23:21
<abarth>
as brendan would say, i don't have a dog in this hunt. I've forwarded your comments to ap.
23:22
<ap>
Hixie: I doubt there is actually a security aspect to null handling in DOM
23:22
<Hixie>
there have been _many_
23:22
<ap>
Hixie: maybe in 199x
23:23
<ap>
Hixie: there were more recent "security" issues about stripping nulls from tags, but those were very remotely related to security, in my opinion
23:23
<jgraham>
Hmm, Opera typically doesn't render past embedded nulls, so we are presumably badly broken with any site that has lots of them
23:24
<jgraham>
So a link would be appreciated :)
23:24
<ap>
jgraham: https://bugs.webkit.org/show_bug.cgi?id=42112
23:25
<ap>
Hixie: and by the way, the more recent "security" issues were caused exactly by the attempts to give null some magical meaning, so some software might have potentially gotten it differently than other
23:26
<jgraham>
abarth: So are the security problems with the "subspace" design considered acceptable because it is too hard to inject code into an unwilling site this way, only one that specifically uses the design?
23:26
<zcorpan_>
jgraham: opera renders past nulls but not in view source
23:27
<abarth>
jgraham: the security problems arise from using these document.domain tricks. the recommended course of action is to ignore document.domain and use postMessage instead
23:27
<abarth>
jgraham: document.domain is very complex and best ignored
23:28
<jgraham>
abarth: From the point of view of a browser implementor though
23:29
<jgraham>
Obviously we have to support document.domain
23:29
<jgraham>
(I agree that *sites* shouldn't use it)
23:30
<jgraham>
(the rough context is that I want to be sure that implementing the HTML5 spec in Opera won't open up any significant security vunerabilities or privacy leaks or whatever compared to what we currently do)
23:31
<jgraham>
zcorpan_: Oh.
23:31
<abarth>
jgraham: yep
23:31
<abarth>
there's a deeper issue is opera
23:31
<abarth>
last time i checked
23:31
<abarth>
opera used dynamic instead of lexical authorization
23:31
<abarth>
which makes the document.domain attacks easier to pull off
23:32
<abarth>
HTML5 requires lexical authorization
23:32
<abarth>
and opera might have switched since i tested it last
23:32
<abarth>
http://www.adambarth.com/papers/2009/barth-jackson-li.pdf
23:32
<jgraham>
I am not sure I understand the difference
23:32
<abarth>
Section 2.1
23:32
<jgraham>
But I will read the paper :)
23:33
<abarth>
in the spec
23:33
<abarth>
it's the difference between the active script
23:33
<abarth>
and the "first script"
23:33
<jgraham>
Ah
23:33
<abarth>
if you're able to run the webkit security layout tests
23:33
<jgraham>
OK. I remember that being "fun"
23:33
<abarth>
there's a bunch of stuff in there about that
23:34
<jgraham>
We can probably find a way to run those
23:34
<jgraham>
and really should be if we are not already
23:34
<Hixie>
ap: that has not been my experience, but i guess our experiences can differ :-)
23:34
<ap>
Hixie: can you point me to some bugs caused by null terminated strings used in DOM?
23:35
<Hixie>
not off-hand
23:35
<jgraham>
abarth: Thanks
23:35
jgraham
decides it is time for slep
23:35
<jgraham>
*sleep
23:35
<ap>
Hixie: and that wouldn't happen if nulls were inserted via DOM manipulation (nulls are still allowed there, correct?)
23:36
<Hixie>
abarth: "first script" was renamed at some point btw (i forget to what)
23:37
<Hixie>
ap: most of the kinds of bugs i've seen reently have been with server-side tools treating NULLs differently than browsers, and failing to properly filter content
23:37
<Hixie>
ap: but NULLs in general are frequently the source of random bugs, and I really would rather never have to worry about it, hence why I always try to write specs to get rid of them as soon as possible
23:37
<ap>
Hixie: yes, that's the kind I referred to as "security"
23:38
<Hixie>
ap: i see no harm in doing so, especially considering how rare NULLs are in "real" content
23:38
<ap>
Hixie: that's not the first time I say that I disagree with this general principle :)
23:38
<Hixie>
ap: i know
23:39
<ap>
Hixie: you may be giving some a false sense of security - they'll write code expecting that the parser gets rid of nulls for them, and then a Dom manipulation will insert a null
23:39
<ap>
Hixie: also, these new requirements are breaking existing content, and slow down decoding/tokenizing slightly
23:40
<Hixie>
ap: there are pros and cons on both sides
23:40
<Hixie>
ap: such is life
23:40
<ap>
Hixie: I suggest that we don't make changes that are not for the better then
23:41
<Hixie>
ap: i disagree that it's not for the better.
23:41
<ap>
Hixie: at this point, I struggle to find a use case that's improved by converting nulls to u+fffd
23:41
<Hixie>
ap: i think it improves security for most users of the HTML parser (who don't have a DOM to worry about), and it doesn't break enough content to be an issue, and the performance impact is minimal.
23:42
<Hixie>
ap: but as you say, we've had this discussion before
23:42
<Hixie>
ap: we're not covering new ground here
23:42
<ap>
Hixie: ah, the eternal "HTML is not for browsers" argument
23:42
<ap>
Hixie: well, previously we were talking about new specs, not about changing HTML in incompatible ways
23:42
<Hixie>
if you're just going to misrepresent what i'm arguing then there's not much point me arguing
23:43
<ap>
Hixie: I didn't intend to misinterpret. did I misunderstand you?
23:43
<Hixie>
i didn't say HTML was not for browsers
23:44
<ap>
Hixie: well, you said that most users of HTML parser don't have the DOM to worry about
23:44
<Hixie>
indeed
23:44
<Hixie>
only five users of the HTML parser have a DOM to worry about
23:44
<Hixie>
out of all the people who parse HTML, that's not the majority
23:44
<Hixie>
it also happens to the be most competent at writing and using HTML arasers
23:44
<Hixie>
parsers, even
23:45
<Hixie>
and thus not the most important when it comes to making sure that it's easy to not screw up
23:45
<Hixie>
s/to the be/to be the/
23:45
<ap>
Hixie: it seems to be the primary target of security attacks though
23:45
<ap>
Hixie: btw, my understanding is that some Web spiders are now interpreting JS already - is that not true?
23:46
<abarth>
ap: that's correct
23:47
<Hixie>
are you saying that abarth and eseidel are so incompetent that what I do with NULLs in the HTML5 spec will affect the number of security bugs in their code? give me a break
23:47
<abarth>
ap: they tend to take a browser and hack it up to get it to crawl nicely
23:49
<ap>
Hixie: it sounds like the only case that's helped by this change is code that incorporates a full HTML5 tokenizer, and then does something truly naive with it
23:49
<ap>
Hixie: since otherwise, they will happily convert null to null, not knowing that there is a spec saying it should be turned into FFFD
23:49
<Hixie>
you've just described a large portion of future HTML parsing software
23:51
<ap>
Hixie: are you saying that it will be easier to incorporate an HTML5 library than to use whatever scripting language mechanism there is to convert a result of a curl/wget download to a native string?
23:52
<Hixie>
do you think that's what i'm saying?
23:52
<ap>
Hixie: yes
23:52
<Hixie>
could you please at least assume i'm not an idiot?
23:52
<Hixie>
seriously
23:53
<ap>
Hixie: you're probably moving too fast for me to follow
23:53
<Hixie>
it will obviously not be easier to wget a file, turn it into a string, and then pass it to a parser than it will be to wget a file, and turn it into a string.
23:54
<ap>
Hixie: you said that a "large portion of future HTML parsing software" will do something, and I'm trying to imagine how I would write a piece of software that would fit the picture
23:54
<Hixie>
to parse HTML, you can either hack it using regexps, or you can use an HTML parser
23:54
<Hixie>
few people will write their own HTML parsers, since it's a non-trivial task
23:55
<ap>
Hixie: yes, if the software doesn't need a DOM, they'll probably just regex parts of wget output
23:55
<Hixie>
there's nothing we can ever do to make regexp parsers do anything, since they ignore the spec by definition
23:55
<Hixie>
so all we can affect are the people using the libraries, which presumably follow the spec
23:55
<ap>
Hixie: and if there is a DOM, they can shoot themselves in the foot even if they don't execute JS from the documents
23:56
<Hixie>
dude, they can shoot themselves in the foot without any help from us
23:56
<Hixie>
the idea is to reduce the likelihood of that
23:56
<Hixie>
we can never make it impossible
23:56
<ap>
Hixie: what I'm saying is that it's not reducing the likelihood, but giving a false expectation
23:56
<Hixie>
and i disagree
23:56
<Hixie>
i don't think it's giving anyone any expectation
23:57
<Hixie>
since these people aren't going to know about this at all
23:57
<Hixie>
anyway i really have to do work
23:57
<Hixie>
if you want this changed, please file a bug or send mail
23:59
<ap>
there is a WebKit bug, that's my form of feedback