00:20
<Hixie>
hsivonen: not sure what to do about http://lists.whatwg.org/htdig.cgi/whatwg-whatwg.org/2007-July/012131.html
00:59
<Lachy>
Hixie, yt?
01:00
<Lachy>
Typo in the definition for the sizes attribute "If the specified, the attribute..."
01:05
<Hixie>
Lachy: fixed
01:08
<Lachy>
Hixie, why have you included 8192x8192 and 32768x32768 in the example? Aren't they a bit excessive?
01:09
<Hixie>
man, nobody understands my sense of humour :-/
05:31
<MikeSmith>
heh
05:31
<MikeSmith>
comment on Mark Pilgrim's blog:
05:32
<MikeSmith>
commenter: I can’t help but think what an arrogant a$$ you’d have to be
05:32
<MikeSmith>
Mark replies: You must be new here.
05:32
<MikeSmith>
http://diveintomark.org/archives/2008/05/07/when-the-fall-is-all-thats-left#comment-12014
05:34
<Hixie>
heh
05:41
<othermaciej>
I got into a stupid flamewar on reddit over Mark's post
05:41
<othermaciej>
I am embarassed for myself
05:43
<othermaciej>
hey roc
05:43
<roc>
hey
06:06
<Hixie>
othermaciej: uri? :-)
06:06
<othermaciej>
Hixie: I'm too embarassed
06:08
<Hixie>
heh
06:09
<othermaciej>
ok if you really want to see it it's here: http://reddit.com/info/6igdz/comments/ but the thread is downvoted so it's a little buried
06:20
<roc>
ok, now I have to guess which one of these people is maciej. that should be fun
06:20
<othermaciej>
my reddit nick is obvious
06:20
<doublec>
"That's like telling me to go ahead and rationalize away being so smart and good looking.:
06:20
<doublec>
hehe
06:22
<othermaciej>
I'm not entirely embarassed about that part
06:22
<roc>
oh, right. I'm just a reddit noob
06:23
<othermaciej>
just resorting to "do you know who I am?!?!?" at the end
06:23
<doublec>
well, he did ask
06:24
<doublec>
he didn't appear to believe you though - unless he really is the ff team lead. Somehow I doubt it.
06:29
<jwalden>
I, for one, welcome our new ff team overlord
06:31
<Hixie>
wow, that guy really was more of a jerk
06:32
<Hixie>
i love how he didn't reply :-)
06:35
<roc>
our Gecko team isn't that much bigger, actually
07:16
<Hixie>
hsivonen: i don't really understand what i should do to address http://www.w3.org/mid/01575703-3A06-4F51-BE27-86A9EBB44C54⊙if
07:16
<Hixie>
hsivonen: in particular, i don't know what to say about xmlns that isn't either obvious, or out of scope
07:50
<MikeSmith>
Hixie: r1577 typo : "keywods"
07:50
<Hixie>
thanks, in the queue
07:51
<MikeSmith>
and comment: might it better be, "User agents must process the links on a per-*keyword* basis" (instead of "per-link" basis) ?
07:53
<Hixie>
nah, it's per-link, not per keyword
07:53
<Hixie>
e.g. link rel="alternate stylesheet" is one link
07:53
<Hixie>
and <link rel="up up icon"> is two links
07:54
<MikeSmith>
ah, OK
08:12
<Hixie>
MikeSmith: i agree entirely with your mail... so... how do we get a changelog? :-)
08:14
<annevk>
morning
08:15
<Hixie>
heya
08:15
<Hixie>
how's xtech?
08:15
<Hixie>
i haven't seen any controversial blog posts about any hot happenings yet
08:19
<MikeSmith>
Hixie: we get it by somebody taking time to sit down and write it up
08:20
<MikeSmith>
which I can help with (or do alone if I have to) starting this weekend (when I get back home to tokyo from dublin) and first part of next week
08:20
<Hixie>
the working group hasn't had much luck waiting for things to happen that way
08:20
<MikeSmith>
yeah
08:21
<MikeSmith>
which is why I think to make sure it gets done I will probably need to be the one to write it up
08:23
<MikeSmith>
unless some magic shoemaker's elves or something sneak in and do it first
08:23
<annevk>
Hixie, crockford attended my presentation which gave some discussion
08:23
<Hixie>
annevk: ooh, do tell
08:23
<annevk>
the lightning talks were pretty funny
08:24
<annevk>
but there was no real new information :(
08:24
<MikeSmith>
Hixie: about XTech, Doug Crockford was at annevk session on Access Control spec, and we had a 15-minute discussion at the end of that
08:24
<Hixie>
what was there to discuss?
08:24
<MikeSmith>
er, what annevk just said :)
08:24
MikeSmith
types too slow
08:24
<annevk>
his claims that the spec should be abondoned
08:24
<annevk>
(art explicitly asked for his feedback at the end of the talk)
08:24
<MikeSmith>
yeah, "abandoned" was his exact word
08:25
<MikeSmith>
though he had great things to say about the cross-document messaging spec
08:25
<annevk>
also that we should solve everything with it and that lots of complexity for people like hsivonen is acceptable (hsivonen explained how easy Access Control was to implement for him)
08:27
<annevk>
other things: lightning talk introducing HTML5 concept from me and on HTML5 parsing from hsivonen
08:27
<annevk>
talk on RDFa that showed quite complex examples (the funny bit was that the namespace declarations were not shown which would've made it near impossible to grasp)
08:28
<Hixie>
sounds like the same thing as 2005
08:28
<Hixie>
glad i didn't go
08:35
<annevk>
it was mostly meeting the other people that made it good
08:36
<annevk>
though some of the stuff was interesting too, such as what SilverLight is doing, how OAuth works, that OpenID 2 is incompatible with OpenID 1, etc.
08:38
<Hixie>
what's silverlight doing?
08:39
<annevk>
it was quite confused, but the interface stuff they're doing looks quite fancy
08:39
<annevk>
they're far ahead in ready components and all though libraries will solve that problem for us I guess
08:40
<annevk>
he also advocated against cross-browser scripting and all quite a bit
08:41
<annevk>
quite aggresively, I suppose, though he also said it wasn't about replacing Flash (yeah right)
09:02
<Hixie>
annevk: it's about replacing html, not flash
09:04
<othermaciej>
a little of both I think
09:05
<Hixie>
flash has like 2% of the content market tops
09:05
<Hixie>
they're doing a platform play, to replace their windows monopoly -- that means 98%, not 2%.
09:06
<roc>
I think it's both
09:06
<roc>
flash isn't dominant now, but they're afraid it could be
09:07
<Hixie>
oh i'm sure they're not ignoring it
09:07
<Hixie>
i just mean that that isn't their primary goal
10:00
Lachy
wonders how many sites actually use silverlight at all?
10:01
<Lachy>
I've only ever seen it used on microsoft.com, and I refused to install the plugin
10:12
<gsnedders>
writing unambiguous ABNF is hard.
10:12
<gsnedders>
(and the damned signal is broken so I'm stuck nowhere on a train)
10:12
<gsnedders>
(and we literally just started moving!)
10:15
<Philip`>
It ought to be possible to get tools that check ABNF for ambiguities
10:15
<Philip`>
No idea if such things exist, though
10:15
<gsnedders>
I don't think so, sadly
10:16
<Philip`>
You could write it in a real parser-generator language, and have that parser-generator check for ambiguity, then manually translate it into ABNF
10:17
<gsnedders>
I guess, but I'm lazy :)
10:18
<gsnedders>
(train stuck again, still not made it to Inverkeithing)
10:18
<gsnedders>
(hopefully will make it to Cam. by 16:00 though)
10:18
<gsnedders>
(certainly before I'm meant to meet you and jgraham_ though)
10:23
<othermaciej>
yacc or bison makes it pretty straightforward to write a checked grammar
10:23
<othermaciej>
(just make all the actions empty)
10:24
<Philip`>
http://ejohn.org/blog/processingjs/
10:24
<gsnedders>
I'm not sure if I can bothered to learn another language, though
10:24
<gsnedders>
yacc is similar to BNF, no?
10:27
Philip`
sees that olliej (I think) already complained about how it uses ImageData non-conformingly
10:27
<Philip`>
(and also in a way that won't work if there's !=1 device pixel per canvas pixel)
10:29
gsnedders
wonders if anything actually uses the CRLF allowed in LWS in HTTP/1.1
10:30
gsnedders
has only ever seen it in test-cases
10:30
<gsnedders>
(i.e., do I actually have to allow LF there?)
10:50
<gsnedders>
http://pastebin.ca/1012564 — header-value is what was causing me issues at making sure it was unambiguous
10:51
<gsnedders>
anybody able to check that?
10:53
<gsnedders>
(no, I don't trust myself)
10:55
<Philip`>
What does "unambiguous" mean?
10:56
<gsnedders>
free of anything that is unclear
10:56
<gsnedders>
i.e., it is completely clear
10:56
Philip`
hopes it just means there's a single parse tree for any input, else he's confused
10:57
<gsnedders>
well, in the case of ABNF, there's no character which could match more than one production where it is in the string
10:58
<gsnedders>
Philip`: I can give you an English lesson tonight, if you want :P
10:58
<othermaciej>
in the case of grammars, "ambiguous" means something more specific than clear/unclear
10:59
<othermaciej>
it means there's only one way to parse
10:59
<othermaciej>
if you use an ABNF-like LALR(1) parser generator, it will report shift-reduce or reduce-reduce conflicts in such cases
10:59
<othermaciej>
though it is possible the grammar isn't ambiguous but you have just hit a limitation of LALR(1)
10:59
gsnedders
sucks at defining English
10:59
<othermaciej>
still, it is a pretty good approximation
11:00
<othermaciej>
I do think yacc/bison would make for a good way to test a grammar
11:00
<gsnedders>
othermaciej: I have enough to do without learning even more :)
11:00
<gsnedders>
(languages, that is)
11:00
<othermaciej>
this grammar might be simple enough for lex to do it
11:01
gsnedders
has an English exam next week, and a French one the week after
11:02
<gsnedders>
othermaciej: Where do you come from, with a name like Maciej, BTW?
11:02
<othermaciej>
gsnedders: straight outta hell
11:03
<gsnedders>
othermaciej: Ah. OK.
11:03
<gsnedders>
(No, seriously?)
11:04
<gsnedders>
Doing it in ABNF gives me a reason to write an ABNF checker, that checks for ambiguities, though
11:04
<gsnedders>
Not that I'll bother, probably :P
11:04
<othermaciej>
yacc is just a tool
11:05
<othermaciej>
learning its (trivial) syntax is almost certainly easier than writing a program to parse a different syntax and do the same checks
11:05
<gsnedders>
yacc is just another tool you have to learn how to use :)
11:06
<Philip`>
Learning how to use a tool is much easier than making your own, and often it's easier than doing your work without the tool
11:06
<Philip`>
Also it's a good opportunity for procrastination
11:07
Philip`
has spent a lot of time playing with LaTeX when he's meant to be writing reports
11:10
<gsnedders>
That's true. I just end up playing around with type-setting maths in LaTeX :P
11:13
gsnedders
wonders how to do this without being ambiguous
11:15
<gsnedders>
I need to properly parse headers, while still allowing garbage headers
11:16
Philip`
doesn't notice any obvious ambiguity in the 'header' thing
11:17
<gsnedders>
Philip`: In what I linked? Good.
11:17
<gsnedders>
Now to make it far more complex and more likely wrong.
11:18
<gsnedders>
I want something like `invalid-header = *( header-content / lws ) [CR] LF`, but that is totally ambiguous
11:18
<gsnedders>
(combined with the valid header)
11:22
<gsnedders>
("If a strict parser is in use, and anything is matched by the invalid-header rule, it is a fatal error.")
11:22
<Philip`>
ABNF is not a good way to do this :-p
11:22
<gsnedders>
It is mostly totally fine.
11:23
<gsnedders>
There are very few quirks like that
11:23
<Philip`>
Mostly, except for all the error handling
11:23
<gsnedders>
There's very little error handling at a parser level :)
11:23
<gsnedders>
It's mostly a case of either it is totally invalid, which is always a fatal error, or has a garbage as a header, which is a fatal error only in strict parsers
11:24
<gsnedders>
Once you start doing header specific stuff, then you get real quirks. But I'm not using ABNF for that.
11:24
<Philip`>
Seems easiest to say something like "If a prefix of the string matches 'header', then process it like a header. Otherwise, fatal error if strict, else skip to the next line"
11:24
<Philip`>
rather than trying to work out how to make a grammar rule that's a complement of another
11:25
<gsnedders>
It's actually really horrible to do using prose. I tried doing parts of it using prose, and it was far harder
11:25
<gsnedders>
I almost ended up using POSIX regex, as that not only has a format specification, but has negation, which ABNF does not
11:26
<Philip`>
Normal regexps only give you a boolean output, which isn't very useful when you want to get an HTTP header out of it
11:27
<gsnedders>
"Taking all matches and submatches of the header rule…"
11:27
<gsnedders>
or the "named header submatch"
11:27
<gsnedders>
It's certainly possible to do
11:27
<Philip`>
Just invent your own string-parsing grammar formalism
11:28
<gsnedders>
It's nicer to use ABNF as then I can just say, "This specification inherits all the rules from RFC3986, which as they are already given in ABNF they are not repeated here."
11:29
Philip`
goes away for a while
11:29
<gsnedders>
See you at 7, then
11:29
<Philip`>
Or 7:05 if I'm late
11:30
<Philip`>
It should only take me 24 minutes to get there, but I'm never very good at departing on time
11:33
<gsnedders>
Philip`: As long as you recognise me and James :P
11:34
<gsnedders>
Philip`: And I don't have a clue how long it takes me to get there, as I don't know where I'm coming from, and I don't know how long it takes to get around Cambridge
11:44
<gsnedders>
ABNF is too vague. It doesn't specify what to do if multiple alternations match a given input
13:52
<zcorpan>
Hixie: how about "HTML <code>Document</code>" and "XML <code>Document</code>"?
13:54
<zcorpan>
Hixie: that should make it clearer that it's about the dom and not a strem of bytes
13:54
<zcorpan>
s/strem/stream/
15:59
<Hixie>
zcorpan: yeah maybe
16:05
<Hixie>
good to see the accessibility experts dismiss the problems hit by people using ATs...
16:07
<Lachy>
Hixie, are you referring to this? http://www.paciellogroup.com/blog/misc/uc/
16:07
Lachy
is just reading it now
16:14
Philip`
assumes it's a reference to http://lists.w3.org/Archives/Public/www-archive/2008May/0016.html
16:21
<Lachy>
In the section Complex Data Images, in their proposal, it seems that they're using alt for text that would be more appropriate for either title="" or as an image caption.
16:25
<Lachy>
and it seems they didn't figure out what the example from the spec "Bubbles traveled everywhere with us" is actually a picture of. But I guess, it's a bit obscure from the caption.
16:26
<Lachy>
http://flickr.com/photos/carinda/1100670787/ Clearly, bubbles is a cat :-)
16:27
<Philip`>
http://lists.w3.org/Archives/Public/public-html/2008May/0079.html sounds like a reasonable argument that empty alt="" (as the spec requires) in those cases is bad, since it removes the distinction between purely decorative images (which you can happily ignore) and informative images (which you might want to see as text by default, but go to some extra effort to see when you decide the image is interesting)
16:32
<Lachy>
For those cases, I still like the kind="" attribute I suggested a few days ago, where for the example in the email, it would be: <img src="rendering-mode-pie-chart.png" kind="Pie Chart" alt="">
16:33
<Lachy>
I don't like the importantimage attribute, since it's likely to suffer from copy and paste errors, and be included where it shouldn't, and it's harder to detect whether or not it should be there.
16:33
<Philip`>
Why is that better than <img alt="Pie Chart" importantimage>? It's worse because it hides the important image entirely from current UAs, and I'm not sure (/can't remember) what makes it better overall
16:33
<Philip`>
Ah
16:34
<Philip`>
importantimage won't suffer from copy and paste errors because nobody will ever bother using it and so it won't be copied :-)
16:34
<Lachy>
with an attribute that says a) what the image is and b) why it's important, errors are more likely to be detected, and it would be easy to analyse with hsivonen's image report tool.
16:35
<Philip`>
With the importantimage thing, the alt attribute says (in combination with its human-understable context) what the image is and why it's important
16:36
<Lachy>
but it also pretends that the text given is its alternative, when it isn't.
16:36
<Lachy>
it's entirely possible to have an important image with a good alt text supplied
16:36
<Philip`>
It doesn't pretend the text is an alternative, if the spec doesn't say that the alt text is necessarily an alternative
16:37
<Lachy>
yes it does
16:37
<Lachy>
that's the whole point of it.
16:37
<Lachy>
also, importantimage is a really bad name
16:37
<Philip`>
The point is to help users, and users might be better helped if alt text isn't always a strict alternative
16:38
<Philip`>
I agree that particular syntax isn't much good
16:39
<Lachy>
importantimage suffers from many of the same problems as noalt
16:39
<Lachy>
it's basically the same thing with a different name
16:40
<Philip`>
importantimage images will still have alt text, so non-hyper-intelligent UAs won't be totally stuck when trying to present the image textually
16:41
<Philip`>
and people won't stick in importantimage to make the validator shut up, because their pages will be equally machine-checkably-valid regardless of whether they have that attribute
16:41
<Lachy>
kind="" also gives additional information about the picture, which may not necessarily come from the alt text. http://krijnhoetmer.nl/irc-logs/whatwg/20080506#l-162
16:42
<Philip`>
alt="A painting of a white house with ..."
16:42
<Philip`>
What are UAs meant to do with 'kind'?
16:42
<Lachy>
same thing they're supposed to do with importantimage, but also say what kind of image it is
16:46
<Philip`>
If all they're doing is serialising the alt and kind values and some connecting words into a text stream (like "Image type: Painting. The house is ..."), why not have authors put that whole serialised text into the alt attribute (alt="Image type: Painting. The ...") instead of adding the complexity of an extra attribute?
16:48
<Lachy>
a better example would be the flowchart one from the spec:
16:48
<Lachy>
<img src="images/parsing-model-overview.png" alt="The network passes data to the Tokeniser stage..." kind="Flowchart">. It would seem wrong to say alt="A flowchart illustrating how the network...".
16:49
<Lachy>
the UA doesn't need to read out the kind by default if there is alt text supplied. Only if the user asks for it.
16:50
<Philip`>
Why would the user want to know that the image a flowchart?
16:50
<Philip`>
Why would the user explicitly ask for it when it's going to say "no 'kind' was specified" for pretty much 100% of images?
16:51
<Philip`>
(I think your argument makes sense, but I'm not convinced yet :-) )
16:52
<Lachy>
it depends how the UA indicates its presence, or what kind of UA it is
16:53
<Philip`>
It seems it would be useful to work from concrete examples of what some UAs could/would do, rather than just presenting some metadata and hoping someone else will figure out what to do with it
16:54
<Lachy>
for a screen reader, it could use any aural indicator to to indicate the presence of more information (like, e.g. a sound effect in the background as it's reading the alt text (probably a bell or tone of some kind)
16:55
<Lachy>
for a text only UA like Lynx, it could include it by default in the textual output somehow. e.g. [Image (Flowchart): The network....]
16:58
<Lachy>
I think screen readers should have done something similar things with other hidden metadata like longdesc. IIRC, users have to ask for it without knowing it's there in current implementations.
17:00
<Philip`>
Why would users of screen readers or Lynx want to know it's a flowchart?
17:02
<Lachy>
to help them decide if it's worth downloading and opening in external viewer, espcially if they want to zoom in on it to see it better
17:07
<Philip`>
I'm not sure why that would help them decide in a non-negligible number of non-contrived cases - I expect they would care about the information presented by the image (which they can find from the alt attribute) when deciding whether to look at it more closely, not about whether it's presented as a flowchart or a photo or a cartoon
17:09
<Philip`>
and it might be a photo of a flowchart, and the author will spend lots of time deciding what an appropriate 'kind' is, when actually it's a waste of time because it won't help people enough to be worthwhile
17:11
<Lachy>
it doesn't have to be used all the time
23:38
Philip`
's hands are cold :-(
23:39
<jgraham>
Heh
23:55
<jwalden>
the latest opera beta didn't have new-style postMessage support, did it?