00:01
<Hixie>
"A
00:01
<Hixie>
er
00:01
<Hixie>
"A URL is a string."
00:01
<Hixie>
there.
00:02
<jcranmer>
"A URL is a non-zero sequence of octets arranged in an unspecified fashion"
00:02
<Hixie>
no, it could be zero-length
00:02
<Hixie>
:-)
00:02
<Hixie>
"" = a URL to the current page
00:02
<jcranmer>
but it can't be 0
00:03
<jcranmer>
I didn't say non-zero length :-)
00:03
<Hixie>
what's a zero sequence of octets then?
00:03
<jcranmer>
uh... a sequence that sums to zero?
00:03
jcranmer
frantically looks for someone who knows higher math
00:03
<Philip`>
A zero-length sequence sums to zero
00:04
<jcranmer>
well, we all know how many documents are non-conforming these days
00:04
<Philip`>
assuming you define sums in a sensible way
00:07
<Lachy>
Hixie, non-utf-8 escaped octets should be conforming. I don't believe there is a requirement that escaped octets in a path must represent UTF-8 bytes, only that non-ASCII characters must be represented by encoded UTF-8 octets.
00:08
<Hixie>
so a valid URL need not be a valid URI?
00:09
<roc>
I thought it was standardized that the first element in the document with a given ID is "the" element for that ID
00:09
<roc>
if not, it should be. The Web seems to depend on it
00:09
<Hixie>
not officially, but de facto, yeah. things get more exciting when you start changing the DOM though.
00:09
<roc>
hmm
00:09
<Lachy>
Hixie, isn't http://example.org/%FF a valid URI? I thought it was.
00:10
<Hixie>
yes, it is
00:10
<Hixie>
but should it be a valid URL?
00:10
<roc>
we always return the first element from getElementById, and if anyone tells us to do otherwise I will sneeze on them
00:10
<Lachy>
of course. Why shouldn't it be?
00:10
<Hixie>
what's the path?
00:10
<Hixie>
roc: even if someone inserts another element with the same ID before it?
00:11
<roc>
then we return the new element
00:11
<roc>
I mean "first in document order"
00:12
<Lachy>
Hixie, what do you mean? It's %FF (or the octect 0xFF)
00:12
<Hixie>
roc: i thought there was some issue where the order guarantee wasn't preserved in the face of dynamic changes
00:12
<Hixie>
Lachy: right, and what unicode character does %FF represent?
00:12
<Lachy>
why does it need to represent a unicode character?
00:13
<jcranmer>
because it means unescaping is always safe?
00:13
<roc>
maybe there is, but I've been hacking in this area recently and I'm pretty sure we preserve the order guarantee in the face of dynamic changes. We're certainly *trying*
00:13
<Hixie>
roc: ah, cool.
00:13
<Hixie>
roc: then we should fix the spec, indeed.
00:14
<Hixie>
Lachy: i guess it doesn't
00:14
<roc>
it's possible IE stuffs it up of course
00:14
<othermaciej>
I remember a while back Firefox would break the order guarantee in some cases
00:14
<roc>
if we do have bugs I'd rather fix them than preserve something insane
00:15
<Lachy>
I think it used to break it if a new element with a duplicate ID was inserted before an existing one with the same ID, or if one was moved from later in the document to before.
00:19
<roc>
ah OK, https://bugzilla.mozilla.org/show_bug.cgi?id=403868, fixed in November last year so in Firefox 3.
00:19
<roc>
Amazon apparently depends on the correct behaviour so it should be safe to write into the spec :-)
00:21
<Hixie>
we need someone to write Web DOM Core 5
00:22
<Hixie>
i think zcorpan volunteered once
00:22
<Hixie>
but i may be wrong :-)
00:23
<othermaciej>
Gecko has 3 getElementById implementations?
00:25
<roc>
yeah
00:25
<roc>
I've actually fixed that in my tree
00:25
<roc>
blame hyatt, he wrote one of them
00:26
<jcranmer>
and who wrote the others?
00:26
<roc>
dunno
00:26
<gavin>
the other two are html and xhtml
00:26
<roc>
digging up CVS blame from that long ago is something I only do when I'm in a *really* bad mood
00:27
<othermaciej>
there's one specific to xhtml, not xml?
00:27
<gavin>
no, I should have said XML
00:27
<roc>
actually gavin's not quite wright
00:27
<roc>
there's one for HTML+XHTML, one for XML, and one for XUL
00:27
<othermaciej>
I think hyatt's excuse for adding the xul one was that his request for making the general one do lots of caching was denied
00:27
<othermaciej>
(I vaguely remember him mentioning this once)
00:27
<gavin>
oh, I didn't know XHTML used the html one
00:28
<roc>
othermaciej: yeah, well I'm not defending any of the decisions made back then :-)
00:29
<roc>
but I will grumble that caching or not, nsXULDocument is the ugliest thing I've touched in a while
00:30
<othermaciej>
I will tell hyatt you send your love ;-)
00:31
<jcranmer>
I work in mailnews/... ugly is the norm
00:31
<Hixie>
Philip`: fixed the style xref link issue (apparently bert's script cross-references <em> elements)
00:33
<Philip`>
Hixie: Thanks
00:33
<Hixie>
anyone know how UAs determine which encoding to use when escaping URIs in scripts?
00:37
<Lachy>
I hope it's either UTF-8 or, failing that, the encoding of the script file
00:38
<Lachy>
but then most scripts don't declare their encodings and, IIRC, it inherits from the document anyway.
00:55
<Hixie>
which document?
00:56
<Hixie>
e.g. if i have na iframe showing doc1, and i get a reference to a function in that script, then navigate the iframe to doc2, different encoding, then call the function
00:56
<Hixie>
does it use the parent's encoding, doc1's encoding, or doc2's?
00:57
<Lachy>
I don't know, I was just speculating. Need to test it.
00:57
<Hixie>
yeah
00:58
<Hixie>
i feared that too
01:02
<MikeSmith>
Hixie: back now
01:02
<Hixie>
i sent mail
01:02
<Hixie>
(good morning)
01:05
<MikeSmith>
Hixie: g'morning.. reading now
01:06
<MikeSmith>
thanks
01:50
<Hixie>
what's the character that is represented in the most different ways in commonly supported encodings?
01:51
<jcranmer>
euro sign?
01:53
<Hixie>
actually that's got surprisingly little variation
02:00
<jcranmer>
:-/
02:03
<Hixie>
this can be reduced to a programming problem
02:03
<Hixie>
which is much more fun
02:03
Hixie
starts coding
02:05
<jcranmer>
writing documentation can be fun... if you're in the mood :-)
02:11
<Hixie>
can't do the docs without the tests
02:11
<Hixie>
and the tests aren't fun :-)
02:11
<jcranmer>
I do docs first and tests second
02:12
<jcranmer>
or, more accurately, I wait for somebody to complain about a regression, and then write the test
02:13
<Hixie>
i need to have the tests in this case to know what to document :-(
02:14
<Hixie>
so much easier to make stuff up
02:19
<Hixie>
U+2116 seems to be the winner
02:20
<jcranmer>
what's that?
02:22
<Hixie>
+2116, U+2235, U+02C7, U+2019, U+00D7, U+201D, U+00F7, U+00A7 actually
02:22
<Hixie>
all have 6 different ways of being encoded in the encodings i was looking at
02:24
<Hixie>
if i only look at single byte encodings, U+0160, U+017E, U+017D, U+0161 win
02:24
<Hixie>
that's better
02:25
<Hixie>
all letters with carrons
02:25
<Hixie>
carons even
02:27
<Hixie>
Codepoint U+017D has 5 different encodings: 0xDE in encodings: ISO-8859-13, Win-1257 0xAC in encodings: ISO-8859-10 0xAE in encodings: ISO-8859-2, ISO-8859-4 0xB4 in encodings: ISO-8859-15, ISO-8859-16 0x8E in encodings: Win-1250, Win-1252
02:27
<jcranmer>
Hixie: but isn't the point of a spec to make up some abstract spot to get implementors to chase towards so that by the time they get there, you write a new one, even harder to implement? That's more or less what CSS did...d
02:27
<jcranmer>
:-)
02:28
<Hixie>
html5 takes spec writing in new directions
02:31
<MikeSmith>
html5 takes a lot of things in new directions
02:32
<MikeSmith>
the Undiscovered Country
02:33
<MikeSmith>
which phrase, interestingly, despite the use of it in a different way for that Star Trek movie, is talking about death
02:33
<Hixie>
hey!
02:33
<MikeSmith>
in Hamlet, where the phrase came from
02:33
<MikeSmith>
Hixie: :)
02:34
<MikeSmith>
sorry, I mean it only in a good way
02:34
<MikeSmith>
I promise
02:34
<MikeSmith>
Rage, rage against the dying of the light!
02:35
<Hixie>
:-)
02:36
<jcranmer>
Wherefore art thou Romeo?
02:36
<jcranmer>
that's about the best I can manage...
02:54
<Hixie>
http://www.hixie.ch/tests/adhoc/uri/encoding/001.html
03:02
<Hixie>
good lord
03:02
<Hixie>
IE8 actually sends ISO-885-13 over the wire
03:05
<othermaciej>
what is ISO-885-13?
03:05
<jcranmer>
8851, I presume?
03:07
<Hixie>
8851, sorry
03:07
<Hixie>
er, 59
03:07
<jcranmer>
er, right
03:07
<Hixie>
but actually if the page is ISO-8859-13, then the path component gets %-encoded as UTF-8 and the query component gets sent as raw Windows-1252
03:18
<Hixie>
oh jesus.
03:18
<Hixie>
so in IE
03:18
<Hixie>
if you set the .src of an iframe
03:18
<Hixie>
to a url that contains a unicode character
03:18
<Hixie>
IE will use the encoding of the document currently loaded in that iframe to encode the chararcter, and then won't escape it
03:19
<Hixie>
as far as i can tell
03:19
<Hixie>
http://www.hixie.ch/tests/adhoc/uri/encoding/001.html
03:19
<Hixie>
http://www.hixie.ch/tests/adhoc/uri/encoding/002.html
03:24
<othermaciej>
that's....
03:24
<othermaciej>
evil
03:24
<othermaciej>
I can see why people would like to pretend that when there are non-ASCII characters, it's an IRI
03:24
<othermaciej>
that would be a lot simpler than reality
03:25
jcranmer
wonders if Hixie will give up in several months and just start labelling all edge cases as places where the spec is undefined
03:27
<Hixie>
i haven't given up yet, and i've been doing this since 2003 :-)
03:27
<othermaciej>
jcranmer: if he was the type of man to do that, he would have done so long since
03:31
<Hixie>
bbl food
03:31
<jcranmer>
when/if it hits, it hits hard
03:42
<othermaciej>
I do fear that someday he'll find a browser behavior crazy enough to make him say "ok, fuck this, I quit"
03:42
<othermaciej>
but that hasn't happened yet
06:17
<Hixie>
actually it's self-limiting in a pretty neat way
06:17
<Hixie>
if the behaviour is too crazy, i just don't spec it
06:17
<Hixie>
i spec something better instead
06:53
<Hixie>
what's with the microsoft people putting their names in square brackets all the time
07:36
<Dashiva>
It's hip to be square.
08:15
<annevk>
Hixie, if you didn't get that e-mail about access-control then I'm not sure what you meant
08:16
<Hixie>
what was the link again?
08:16
<annevk>
Hixie, you wanted personalized widgets on the server right without having to use HTTP headers?
08:16
<Hixie>
i've sent a lot of mail, i'm not sure of the context here
08:16
<annevk>
Hixie, http://lists.w3.org/Archives/Public/public-webapps/2008AprJun/0255.html
08:18
<Hixie>
right the idea is that the xbl might be returned differently based on cookies
08:18
<Hixie>
but sicking's idea would not send cookies without a flag
08:18
<Hixie>
so i'd have to add a flag to xbl2 for when we want cookies to be sent
08:19
<Hixie>
personally i'm not sure sicking's idea makes a whole lot of sense, as i later mentioned
08:19
<Hixie>
but assuming you do adopt it, then those xbl changes will be made later
08:20
<annevk>
but then on the server side you still need to use HTTP headers
08:20
<annevk>
because otherwise the flag will be ignored by the user agent
08:21
<annevk>
I thought that wasn't acceptable
08:21
<annevk>
if it is, we might as well drop <?access-control?> ...
08:28
<Hixie>
well that's fine
08:29
<Hixie>
doesn't matter what the server side has to do
08:29
<Hixie>
i was just talking about <xbl>
08:29
<Hixie>
er, <?xbl?>
08:29
<annevk>
oh?
08:29
<Hixie>
(and you still need <?access-control?> for the more common no-cookie case anyway)
08:29
<annevk>
I see...
08:30
<annevk>
In that case I did misunderstood you
08:37
<annevk>
wtf
08:38
<annevk>
I pointed out in an e-mail why WebKit requested not throwing an exception for responseXML and responseText based on site compatibility and IE being inconsistent and they're saying no arguments were made?!
08:38
<annevk>
hah
08:41
<annevk>
"Oops. Adding Anne." :/
08:41
annevk
is on the list
08:52
<Hixie>
annevk: i recommend just repeating yourself
08:54
<annevk>
yeah, maybe during the weekend
08:54
<annevk>
I should really make this presentation I have to give later today
09:34
<Hixie>
nggh. http://www.hixie.ch/tests/adhoc/uri/encoding/003.html
09:37
<annevk>
"%-escaped ISO-8859-13"
09:37
<annevk>
for query
09:37
<Hixie>
you must be using opera or firefox
09:38
<annevk>
yeah, Opera
09:39
<Hixie>
IE uses -15 (unless you go back/forward, in which case it starts doing even weirder things)
09:39
<Hixie>
safari uses -2
09:39
<Hixie>
iri says to use utf-8
09:41
<Hixie>
oh except IE uses raw octets and everyone else escapes
09:41
<Hixie>
everyone ignores the encoding of the script when doing this
09:42
<Hixie>
that's interesting
09:42
<Hixie>
of course it means now i have to find a way to associate a Document with every script
09:42
<Hixie>
that's exciting
09:43
<annevk>
doesn't the origin of a script arrive at some Document somehow
09:44
<Hixie>
not always
09:47
<Hixie>
bed time
09:47
<Hixie>
nn
14:58
Philip`
continues to wish he could tell Gmail that anything sent to whatwg⊙wo is not to be considered spam
15:02
<Lachy>
what is it about whatwg emails that get caught?
15:02
<Lachy>
does it do the same for public-html? (I would classify a lot more of those as spam than I would whatwg)
15:03
<Philip`>
Lachy: Don't know what it is, but it caught your latest email
15:03
<Lachy>
great.
15:03
<Philip`>
It only seems to be around one every couple of weeks, and I've not noticed any pattern
15:04
<Lachy>
can you set up a filter or a whitelist or something?
15:04
<Philip`>
I don't remember public-html emails getting marked as spam, but that's mostly because my memory isn't great
15:05
<Philip`>
Hmm, now I can't connect to Gmail at all :-/
15:08
<Philip`>
Ah, back now
15:09
<Philip`>
Lachy: I can set a filter to label/archive/etc WHATWG mails, but that just makes the messages in the spam folder have a WHATWG label - I can't see any way for a filter to affect the spam filtering
15:20
<annevk>
then you search every 30 days for label:whatwg in:spam
15:20
<annevk>
or something like that
15:21
<jcranmer>
looks like gmail needs better "adaptive junk mial controls"
15:21
<jcranmer>
s/ia/ai/
15:26
<Lachy>
Philip`, have you tried contacting google and sending them a bug report about it?
15:31
<annevk>
interesting, WHATWG has now attraced crowd un-aware of what the W3C is doing
15:31
<annevk>
attracted*
15:34
<Lachy>
annevk, who is unaware?
15:41
<annevk>
the people talking about cross-site requests
15:41
<annevk>
and you, apparently :D
15:44
<Lachy>
ah, I hadn't read those mails yet
16:35
<billyjack>
annevk: Frode Børli seems to have quite prolific of late as far as postings to whatwg go
20:26
<mcarter>
Hixie, whats the URI scheme for secure web socket... "wss://" ?
20:51
<Lachy>
mcarter, I thought the web socket proposal just used HTTP URIs
20:52
<Lachy>
in which case, couldn't HTTPS be used for establishing secure connections?
20:52
<mcarter>
Yeah, thats what I had proposed. There had been discussion in here about the potential of using an alternative, such as "ws://"
20:52
<mcarter>
i'm just curious what a secure WebSocket would look like in the context of "ws://" as the URI
20:54
<Lachy>
ah, ok. I had missed that discussion. I'll read the logs about it later
22:27
<Hixie>
wss:// makes sense to me
23:02
<mcarter>
the default port for wss:// would be 443 then, I assume
23:06
<MikeSmith>
annevk: URL for your Reboot session/proposal?
23:08
<Hixie>
yes
23:08
<Hixie>
i suppose so
23:29
<Lachy>
OMG! Rob's resonse on bug 5772 about IDs is ridiculous. I give up.
23:33
Hixie
sends google's feedback on the geolocation charter
23:38
<hober>
Lachy: yeah, I couldn't make heads or tails of it
23:40
<Hixie>
well he didn't remove the NEEDSINFO
23:41
<Hixie>
so i don't see it on my list!
23:50
<Lachy>
Hixie, his last comment that I was talking about on that bug was the one before yours
23:51
<Hixie>
oh ok
23:51
<Hixie>
my comment was "wtf" right?
23:51
<Lachy>
I'm sure he'll reopen it again later and you'll have the pleasure of reading it then
23:51
<Lachy>
yeah
23:53
<Lachy>
I don't get how allowing duplicate IDs doesn't defeat the purpose of an attribute designed as a unique identifier, or how he can claim that "backwards compatible part is a red herring here" and then point out that it doesn't work at all in existing UAs, in the same sentence.
23:58
<Hixie>
those things don't even seem related enough to be in the same sentence