00:03
<nessy>
I'd call that web fundamentalism
00:04
<Hixie>
i don't even know what he means
00:05
<Hixie>
which i guess is his point
00:05
<Hixie>
hopefully he'll clarify
00:35
<jwalden>
Philip`: it's a character flaw; bz argues too much with people he should ignore, even when he knows better (and he'll be the first to admit it) :-)
00:39
<roc_>
everyone I know succumbs to that at times
00:42
<jwalden>
I tend to think bz succumbs more often than others
01:00
<Hixie>
I used to do that
01:00
<Hixie>
I designed mail filters to avoid it
01:01
<Hixie>
(the mail is filtered such that it takes enough effort for me to see it that I don't see it until after the thread is a few days old, at which point I don't feel the need to jump in anymore)
02:34
<Hixie>
should placeholder="" apply to type=password? what about type=date? type=number?
02:35
<Hixie>
yes no yes looks like the better answer
02:40
Hixie
decides yes no no
02:58
<jcranmer>
********* yesterday 7 ?
03:17
<Hixie>
hm?
03:18
<Hixie>
rb is such a troll it's hilarious
03:18
<Hixie>
his lst 6 or 7 e-mails have all said the exact same thing and avoided answering any of the actual questions he was asked
09:00
<yecril71>
The wording "static object" is ambiguous indeed, new or not new.
09:00
<yecril71>
I would rather say "snapshot" to avoid misunderstandings.
09:03
<Hixie>
"static" is far more correct english than "snapshot" for this case
09:07
<hsivonen>
is comatose politically incorrect? The kind of NodeLists the Selectors API returns are comatose, not live.
09:08
<hsivonen>
or a copy of the value
09:08
<Hixie>
"static" is the right word, people's misunderstandings notwithstanding
09:30
<takkaria>
7B7B7B7B7B7B7B7BThen the operations could still handle that common case of being the
09:30
<takkaria>
setter/getter methods.Then the operations could still handle that common case of being the
09:30
<takkaria>
oops, sorry
09:30
<takkaria>
Then the operations could still handle that common case of being the
09:30
<takkaria>
ack
09:31
<takkaria>
mouse is being a bit screwy. and you get no points for guessing what I'm reading
09:37
<jruderman>
i've been to that erotic-stories site
09:38
<yecril71>
Tell it to K&R.
09:39
<yecril71>
The effect of their technical vocabulary cannot be undone, I am afraid.
09:41
<Hixie>
?
09:41
<Hixie>
i rarely understand you, yecril71 :-)
09:42
<yecril71>
The meaning of "static" is engraved in stone by Kerrighan and Ritchie.
09:43
<yecril71>
It need not be perfect English but you cannot get rid of that.
09:43
<yecril71>
Especially in the context of a programming language.
09:49
<Hixie>
the meaning of static as defined by C-like languages is pretty much exactly what we're talking about here
09:49
<Hixie>
so that's fine
10:01
<zcorpan>
Hixie: perhaps you should define "static" alongside "live" and xref it
10:01
<Hixie>
well it now says "new"
10:01
<Hixie>
which seems clear to me
10:01
<Hixie>
since why would you return a new object each time if it was live
10:02
<zcorpan>
ok
10:05
<hsivonen>
http://blog.jclark.com/2008/11/what-allowed-in-uri.html
10:08
<Hixie>
not sure what to make of that
10:17
<yecril71>
See e.g. the documentation of "itoa", or "localtime".
10:17
<yecril71>
It is not reentrant *because* it returns a static object.
10:17
<yecril71>
(a pointer to one)
10:18
<yecril71>
Your argument about "new" is valid but:
10:18
<yecril71>
- it can be easily overlooked
10:19
<yecril71>
- the statement seems contradictory.
10:26
<Hish>
hi. regarding the <output for"id1 id2 id3> element: is there a way for the #id1 element to find out in which output element it is used for calculation?
10:27
<hsivonen>
yay for placeholder
10:30
<Hixie>
Hish: not currently
10:30
<Hish>
ok. thx.
10:41
<takkaria>
wow, WF2 integration work is pretty much done
10:46
<Hixie>
getting there
10:47
<Hixie>
i need to go through the wf2 spec at one point line by line
10:47
<Hixie>
making sure i haven't missed anything
10:47
<hsivonen>
Hixie: do you have plans on rewriting the meta sniffing algorithm as a state machine?
10:48
<Hixie>
which one?
10:48
<hsivonen>
Hixie: the prescan on bytes
10:48
<Hixie>
"8.2.2.1 Determining the character encoding"?
10:49
<hsivonen>
Hixie: yes
10:49
<Hixie>
no, no intention of rewriting that section.
10:49
<Hixie>
why?
10:49
<Hixie>
does it need it?
10:49
<hsivonen>
Hixie: it might be useful to implment it as a state machine at least
10:50
<hsivonen>
unless waiting for the first n bytes has a trivial perf hit even when the meta occurs slighly before n
10:50
<hsivonen>
Hixie: do you have plans on prescribing speculative script tag prescanning?
10:50
<Hixie>
i imagine waiting for bytes will take far longer than reading them
10:50
<Hixie>
no
10:50
<Hixie>
other than network traffic, it has no observable side-effects
10:51
<hsivonen>
ok
11:48
<mookid>
Hixie: someone said you would be the person to talk about with regards to an informal proposal
11:48
<Hixie>
sure
11:48
<mookid>
did you read any of it previously?
11:48
<mookid>
It was a discussion abuot an optional Accept attribute
11:49
<mookid>
for indicating content-type of a request
11:49
<mookid>
in the Accept header
11:50
<Hixie>
just saw an e-mail about it, filed it along with other forms-related feedback for future consideration (it can take years for me to get to replying to a particular e-mail, there's 2545 e-mails in my feedback pile right now)
11:50
<mookid>
it's not forms related :P
11:50
<mookid>
an optional tag for <a> or <img> - any hypermedia request
11:51
<mookid>
to indicate to the browser what Accept header to send
11:51
<mookid>
having this as part of HTML5 will stop encourageing developers to misuse URI's and include content-type in them
11:52
<Hixie>
yeah i guess it should have been put in the links folder
11:52
<Hixie>
no biggie
11:52
<Hixie>
to be honest though personally i think the whole http content negotiation feature is a big failure and we'd be better off removing the whole thing from http
11:53
<mookid>
it's a failure because HTML and Browser support for the rfc is pathetic? -_-
11:53
<Hixie>
i haven't seen many people using it, i've seen very few people needing it, and users seem to get confused by it whenever they are exposed to it
11:53
<mookid>
you can't use it because the markup is insufficient
11:54
<Hixie>
no i think it's a failure because it solves a problem that doesn't exist in a way that isn't compatible with the way most people (authors and users) think
11:54
<mookid>
they only think that way because the HTML they've been working with is crap.!
11:54
<mookid>
that's not a fair criticism of HTTP
11:54
<mookid>
that's a problem with HTML and browsers
11:55
<mookid>
content-type doesn't belong in a resource identifier, is what I'm getting at
11:58
<Hixie>
what problem does this solve?
11:58
<Hixie>
content-type is a waste of time too, we should just be using unique identification strings for content types like the PNG header
12:04
<mookid>
right - you're suggesting that content negotiation belongs in the URI
12:05
<mookid>
reading Roy Fielding's dissertation it becomes clear that this is not the intention of a URI at all
12:06
<mookid>
I'm suggesting a way that HTML can allow for both - I don't see how that's a bad thing
12:06
<mookid>
unless, of course, you understand HTTP better than Roy..! :P
12:08
<Hixie>
i'm suggesting that there not be any content negotiation because people don't need it or care for it
12:11
<Hixie>
anyway, bed time for me now
12:11
<Hixie>
nn
12:14
<hsivonen>
http://mxr.mozilla.org/mozilla-central/source/parser/htmlparser/public/nsIParser.h#89
12:14
<hsivonen>
that's a lot of different cases...
12:14
<hsivonen>
now I need to figure out which ones are tentative and which ones are confident
12:18
<hsivonen>
I guess < 9 are tentative and >= 9 are confident
12:42
<mookid>
Hixie: content negotiation is one of the main components of RESTful APIs - it allows your URI's to serve JSON/XML for application interaction and HTML for user interaction
12:43
<mookid>
if you take out content-negotiation that means one URI can only serve one content-type
12:43
<mookid>
one URI = one resource
12:44
<mookid>
one resource can have many representations
12:47
<hsivonen>
mookid: what problem is solved if to dereference a link you need a URI and an accept value and then you stick these into separate places in the HTTP request instead of baking the format identification into the URI?
12:47
<mookid>
because that's not the purpose of a URI
12:47
<hsivonen>
mookid: that's not a statement of solving a problem.
12:48
<mookid>
no, but the approach that - well everyone uses URIs to negotiate content so therefore it's the best way
12:48
<mookid>
is equally insufficient
12:48
<hsivonen>
why?
12:48
<mookid>
because it's purely the inadequacy of html and browsers that has led to that
12:48
<hsivonen>
sounds like dogma to me
12:49
<hsivonen>
not like an actual problem
12:49
<mookid>
have you read Roy Fielding's thesis? there are benefits from having one URI for a resource
12:50
<hsivonen>
mookid: I have read parts of it, but I haven't read it from start to finish.
12:50
<hsivonen>
mookid: can you point me to a benefit?
12:50
<mookid>
yes
12:50
<mookid>
So I can send a colleague a message; 'you can get the report at http://example.com/report';, and they can use that URL in any user agent that is appropriate. A browser is a special case in which many different content-types are dealt with. The same benefit is not achieved if the content is negotiated via the URL, since the user would have to know the type their user agent required and modify the URL accordingly
12:51
<mookid>
it means that you can use one URI to serve several user agents
12:51
<hsivonen>
what other user agent would the colleague use?
12:51
<mookid>
they could maybe use excel
12:51
<mookid>
or.. word.. or wahtever
12:51
<mookid>
this is the direction web applications are going
12:52
<mookid>
I'm not making this up for nothing!
12:52
<hsivonen>
so isn't the accept header bit then something that the agent should generate instead of something that the the hypertext reference should contain?
12:52
<mookid>
browsers are different case really
12:53
<mookid>
they, generally, pass the content to the OS if they can't understand it
12:54
<mookid>
most UA's just handle one filetype
12:54
<mookid>
browsers do lots - whcih is why it would be nice to have that ability reflected in html
12:54
<mookid>
does that make sense?
12:55
<hsivonen>
so if you feed them general Web URLs, you are likely to end up frustrated, which is why the user probably wants to know a priori that the URL works with a particular kind of non-browser HTTP agent
12:55
<hsivonen>
mookid: no, I don't think your argumentation makes sense
12:55
<mookid>
they will, generally, have some idea - the worst that can happen is there UA will say "nope, no good here"
12:55
<mookid>
well it makes sense because I'm suggesting the best solution is to provide a way to use HTTP as it was intended
12:56
<hsivonen>
If you send me an URL to a report, why would I try it in Excel if you didn't tell me that it has been specifically crafted to work usefully in Excel?
12:56
<mookid>
if anyone should be justifying themselves with a use case it should be the guys saying "content negotiation is done in URIs"
12:57
<mookid>
well the idea is that as the application frameworks for these kinds of APIs grow.. all user agents will be able to interact with the data
12:57
<mookid>
e.g. opening it in iTunes streams a voice-translated version for the blind
12:57
<mookid>
that type of thing
12:57
<hsivonen>
no, only user agents that support the particular flavor(s) of data availabale will work
12:58
<hsivonen>
it's useless to try a URL that deferences into KML in Excel, for example.
12:58
<mookid>
right.. so the worst that can happen is a UA says 'nope sorry I dont know what to do with that'
12:58
<hsivonen>
if I have a zillion agents in addition to a browser, do I try each one when you send me and architecturally correct URL?
12:59
<mookid>
well you pick the one you know/guess is appropriate
12:59
<mookid>
you're assuming complete ignorance on behalf of the user
13:00
<hsivonen>
no, I'm assuming that a user who gets an HTTP URL by email without additional instruction dereferences the URL using a browser
13:00
<mookid>
I would send an email out saying "here's the report: example.com/report, you can open it in excel, powerpoint, and word" - how would you do it with URIs? :)
13:01
<hsivonen>
Here's the report in various formats http://example.com/report.xls, http://example.com/report.ppt and http://example.com/report.doc
13:01
<mookid>
but if they're all the same resourcee
13:02
<mookid>
just different representations
13:02
<mookid>
then that's a misuse of the term URI
13:02
<mookid>
I've spent a fair amoutn of time studying this
13:02
<mookid>
I'm not suggesting this for fun..!
13:02
<hsivonen>
I posit that the case where the alternative representations of a resource are truly equivalent is so rare that it is pointless to design for it
13:03
<mookid>
like RSS feed and index.html ?
13:03
<mookid>
:)
13:04
<hsivonen>
an RSS feed and index.html are rarely equivalent in the sense that one didn't contain some information that the other contains
13:04
<mookid>
I think my report example makes sense.. the only reason these kinds of interactions arent the case is because web application development is young.. and because html discourages peolpe from using URIs as they were intended
13:04
<mookid>
they are completely equivalent in the majority of cases
13:04
<mookid>
look at blogs..
13:04
<mookid>
look at technorati..
13:05
<hsivonen>
mookid: well, depends on at what point in time you cast "intended" in stone
13:05
<mookid>
well..
13:05
<hsivonen>
mookid: the original HTTP didn't even have content types
13:05
<mookid>
I'd call an rfc.. an intention
13:05
<mookid>
HTTP 1.1 did
13:05
<mookid>
and does
13:05
<mookid>
so that's a mute point
13:05
<hsivonen>
so one might write Accept off as later day astronauting ;-)
13:05
<mookid>
?
13:06
<mookid>
you aren't doing a very good job of convincing me I have to say
13:06
<hsivonen>
mookid: HTTP conneg is younger than URLs
13:06
<mookid>
I've spent a fair while looking at this
13:06
<hsivonen>
I could say the same thing.
13:06
<mookid>
well I'm looking at how it is defined now.. and how people are using it moving forwards (which is presumably what HTML5 should be looking at)
13:06
<mookid>
well you admitted that you havent read that much into it
13:07
<mookid>
why would you not want to give developers the option at least?
13:07
<hsivonen>
I haven't read the canonical theory in its entirety. I have considered the practical side quite a bit.
13:07
<mookid>
how can you do that if you dont have a full grasp of the concepts?
13:07
<hsivonen>
mookid: giving the option is not free
13:07
<hsivonen>
mookid: also, conneg sucks for seachability
13:07
<mookid>
free?
13:08
<mookid>
does it?
13:08
<hsivonen>
mookid: implementing the feature has an opportunity cost: the developers wouldn't be doing something else
13:09
<mookid>
these kinds of functionality is what helps the framework developers
13:09
<mookid>
if you can standardise this stuff it will assist tooling for proper HTTP transactions
13:09
<hsivonen>
mookid: yes, without representation-specific URIs and with a countably infinite set of potential Accept values, how do you index all representations of a resource?
13:10
<mookid>
that's an issue of producing a format for spiders
13:10
<mookid>
that's a non issue if all it requires is another content type built specifically for crawlers
13:10
<hsivonen>
well, doing that isn't costless, either
13:10
<hsivonen>
afk, I'll be back
13:10
<mookid>
depends how your aplpication is built
13:10
<mookid>
if you're just providing an extra view to a model
13:10
<mookid>
it's a 2 second job
13:11
<mookid>
across the board
13:12
<mookid>
content negotiation is bad right now for searchability because it *is* done in the URI
13:12
<mookid>
that's the point I'm trying to get across here
13:13
<mookid>
google are championing REST for a reason
13:13
<mookid>
:)
13:14
<takkaria>
relying on authors to provide an index of their representations to spiders seems very prone to spamming attempts
13:15
<Philip`>
Why is it any more prone to that than providing a separate HTML document which links to lots of other URIs that spiders will follow?
13:17
<Philip`>
(Unintentional errors seem a more important problem, because people would forget about their spider-specific index and it would be buggy or out-of-date)
13:17
<takkaria>
fair point
13:20
<hsivonen>
mookid: is google really championing REST according to Fielding or POX over HTTP Web services with a lower case s?
13:21
<hsivonen>
I'd argue that AtomPub-based services are their own class of services with Atom-aware clients. They aren't totally generic conneg
13:44
<hsivonen>
zcorpan: I believe I've now addressed the problems that blocked building v.nu on Windows
13:59
MikeSmith
wonders why/whether "using HTTP as it was intended" should be a goal
14:00
<MikeSmith>
I would doubt that most people would say we are now using HTML as it was originally intended
14:03
<hsivonen>
looks like I need to expand my fragment parsing API to deal with fragments rooted at foreign content...
14:19
<hsivonen>
when a document has been created with document.open() and everything is document.written, does the first level of document.write count as being equivalent to a network stream or does it count as the first level of document write?
14:19
<hsivonen>
I assume the former considering the behavior of Hixie's live dom viewer
14:24
<zcorpan>
hsivonen: that's nice
14:27
hsivonen
finds
14:27
<hsivonen>
// Hack to pass on to the dtd the caller's desire to
14:27
<hsivonen>
// parse a fragment without worrying about containment rules
14:27
<hsivonen>
if (aMode == eDTDMode_fragment)
14:27
<hsivonen>
mCommand = eViewFragment;
14:28
<hsivonen>
Hixie: have you studied the purpose/impact of Gecko having some kind of magic fragment mode for containment rules?
14:54
<pergj>
hsivonen: I don't remember seeing anything in the HTML5 spec that would give different behaviour for document.write immediately after document.open
14:56
<hsivonen>
pergj: but if document.write after document.open counted as the first level of document.write, how could Hixie's live dom viewer have the right document.write semantics?
14:58
<pergj>
hm... Not sure...
16:58
<BenMillard>
takkaria, you just sent 5 empty messages. So now who's "fail"! :P
16:59
<Dashiva>
They're not empty, they're images encoded with lenpeg v2
17:00
<BenMillard>
oh, they show up empty in Opera and the logs
17:03
<Dashiva>
http://www.dangermouse.net/esoteric/lenpeg.html
17:03
<BenMillard>
yeah, I've just reached the bottom of that :(
17:05
<takkaria>
ah, oops
17:05
<takkaria>
my mouse is doing weird things today
17:06
<takkaria>
mixed with X11's middle-click-is-paste thing, bad things happen
17:16
<MikeSmith>
takkaria: your mouse should meet Hixie's cats
17:16
<takkaria>
:)
17:30
<BenMillard>
MikeSmith, nice job on the TPAC 2008 minutes and overview page.
17:33
<MikeSmith>
BenMillard: thanks for your prompting. I still need to add some of the links you suggested
17:41
<jmb>
so, a really unscientific measure of how complex the html5 parsing algorithm is in comparison with a pre-existing html4 parser:
17:41
<jmb>
sloccount on libxml's HTML{parser,tree}.c => 5188 sloc
17:42
<jmb>
sloccount on hubbub's tokeniser and treebuilder source trees => 9127 sloc
17:44
<takkaria>
I imagine hubbub has more boilerplate than lixml though
17:45
<jmb>
could well. hence "unscientific" :)
17:50
<jmb>
oh, in case anyone cares, that's libxml 2.7.2, and hubbub svn head
17:56
<gsnedders>
First two thoughts when writing CV: 1) What do I put in it? 2) How do I mark it up?
17:57
<takkaria>
look at Tim Bray's
17:57
gsnedders
doesn't like the markup of Hixie's :)
17:58
<takkaria>
http://www.tbray.org/ongoing/When/200x/2005/11/12/Template.html
17:59
<Philip`>
Nobody cares about the markup, since they'll just print it out :-p
17:59
<gsnedders>
Is it really a table though?
18:04
<takkaria>
it's not really a table, I don't think, but I don't think it matters
18:04
<takkaria>
you evidently do :)
18:04
<gsnedders>
http://alexking.org/site/projects/html-resume-template/demo/resume.php semms a lot closer
18:10
<BenMillard>
gsnedders, mine is here and I was told "your CV rocks" by 1 potential employer: http://sitesurgeon.co.uk/
18:10
<gsnedders>
How ugly.
18:10
gsnedders
ducks
18:10
<BenMillard>
yeah, the visual aspects aren't my strong point :)
18:10
<BenMillard>
the markup and content is what you seemed interested by
18:11
<BenMillard>
anyway, dinnertime now, cya
18:38
<gsnedders>
Prince doesn't support meta@charset seemingly
18:42
<takkaria>
no, it doesn't
18:51
<jcranmer>
I presume if I had <input type="xxxinvalid">, would getting .type return "text" ?
18:51
<jcranmer>
s/would/then/
19:05
<gsnedders>
jcranmer: yes
19:05
<jcranmer>
ah, good :-)
19:27
<Dashiva>
public-hmtl is back in high gear again, I see
19:29
<takkaria>
lots of revs but the break is on
20:15
<Lachy>
Hixie, Thanks for finally adding placeholder :-)
23:34
<Hixie>
man i need to eat before i can reply to the crazy stuff on the w3c html lists
23:35
<Hixie>
roy fielding in particular appears to live on a different planet
23:37
<roc>
the one email I read of his suggested that testing your HTML in a browser is bad because you might be tempted to make it less pure
23:40
<jcranmer>
bwahuh?
23:40
<gavin>
But now that Firefox is getting just as buggy and complex as
23:40
<gavin>
the other major browsers, they have no reason to switch at all.
23:40
<gavin>
Firefox usage hasn't increased since it decided to be no better
23:40
<gavin>
than the others.
23:41
<gavin>
that's an interesting assertion!
23:41
<jcranmer>
numerous factual errors in ther
23:41
<jcranmer>
there
23:41
<jcranmer>
shall I pull out the mandatory xkcd reference?
23:42
<Lachy>
jcranmer, xkcd references are always funny
23:45
<Philip`>
jcranmer: XKCD references are getting a bit tired now :-p
23:46
<jcranmer>
it's the one where the character responds to trolls with long, multi-paragraph replies detailing why the poster should just die
23:49
<Lachy>
http://xkcd.com/493/
23:49
<Lachy>
is that the one you meant?
23:50
<jcranmer>
no, not that one
23:51
<Lachy>
ok, then I can't find it
23:52
<jcranmer>
I'm not finding it either
23:52
<svl>
This one: http://xkcd.com/406/ ?
23:54
<jcranmer>
yes!