00:00
<annevk>
(which is what's used for cross-references)
00:00
<jennb>
Hixie: yup, reloading changed to fail
00:00
<Hixie>
jennb: all my browsers show "pass"
00:00
<Hixie>
jennb: are you testing with your patch?
00:00
<Hixie>
jennb: (your patch would make it say "fail")
00:02
<annevk>
zewt, fixed
00:02
<annevk>
zewt, I hope you are Glenn btw
00:02
<zewt>
yep
00:03
<annevk>
anyway, should've been asleep long ago once again; maybe I can figure out some stuff on the plan tomorrow
00:04
<annevk>
such as how to structure the initial sections and how assimilate the borrowed terminology
00:04
<annevk>
how to*
00:04
<jennb>
Hixie: On released Chrome, "fail". On Chrome dev trunk, "pass". That's what I expect.
00:04
<annevk>
and plan should read plane
00:04
<jennb>
Hixie: The bug wants 142.html to select Windows-1252 like it used to.
00:05
<Hixie>
jennb: matching IE8 seems like a better goal than matching an old Chrome
00:06
<Hixie>
jennb: especially when Firefox and Opera agree with IE8
00:06
<jennb>
Hixie: released FF gives me "fail" on 142.html test
00:07
<Hixie>
jennb: yeah but that's a separate bug
00:07
<Hixie>
jennb: they don't treat 8859-9 as -1254
00:07
<Hixie>
jennb: the point is they get 1254/8859-9 rather than 1252/8859-1
00:08
<jennb>
Hixie: oh, i see. My IE also gives me "pass". 
00:08
<jennb>
Hixie: i'll add a note to the bug that IE8 behaviour.
00:08
<jennb>
s/that/about
00:08
<zewt>
http://www.hixie.ch/tests/adhoc/html/parsing/encoding/142.html in FF3.6 is "fail", FWIW
00:09
<Hixie>
zewt: see above, it's a pass for the purposes of this discussion
00:09
<Hixie>
the failure is the failure you see on 003.html
00:09
<Hixie>
not the failure we're testing
00:10
<zewt>
i mean re: '<Hixie> jennb: all my browsers show "pass"'
00:10
<jennb>
zewt: that was before I reloaded his test.
00:11
<Hixie>
zewt: ff3.6 shows "pass" if you ignore the fact that it actually says "fail" and just look at the actual results :-P
00:11
<ap>
Hixie: ff3.6 says "Encoding used by browser is: Windows-1252"
00:11
<zewt>
ignoring the result, "Encoding used by browser is: Windows-1252", eg. it picks up "cotent"
00:11
<Hixie>
it does?
00:11
<Hixie>
huh
00:11
<Hixie>
well, ff4 uses 1254
00:12
<Hixie>
i wonder if they have the same regression on that page then
00:12
<ap>
Hixie: could also be complicated by content-based encoding sniffer in Firefox
00:13
<Hixie>
that test should be safe from that
00:13
<zewt>
yeah, i also see ISO-8859-9 used in FF4b11
00:13
<jennb>
if FF has improved their charset detection, we should too. ;-)
00:13
<ap>
Hixie: they do have the same regression on actual page, yes
00:13
<Hixie>
interesting
00:13
<Hixie>
does the actual page work in IE8?
00:13
<Hixie>
(i have IE8 up here if you can pass me the URL)
00:14
<ap>
Hixie: https://bugs.webkit.org/show_bug.cgi?id=54582 (you need to log in, not just open the url)
00:14
<ap>
Hixie: ugh, not sure if I can share the credentials :(
00:16
<Hixie>
well i can promise that i won't abuse them, since i don't understand a word of this site :-)
00:18
<Hixie>
ok i got to log in
00:18
<Hixie>
looks like it works fine in IE8
00:18
<Hixie>
dunno why though
00:20
<Hixie>
the page i get doesn't have a misspelt charset decl
00:21
<Hixie>
wfm in firefox too!
00:21
<Hixie>
(trunk)
00:21
<Hixie>
the decl is:
00:21
<Hixie>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
00:21
<Hixie>
did they just fix this or am i going crazy?
00:22
<ap>
Hixie: I'm still seeing "contet" with a typo
00:22
<ap>
Hixie: even when spoofing UA string in Safari
00:22
<Hixie>
oh wait
00:22
<Hixie>
i didn't follow the rest of the intstructions
00:22
<Hixie>
don't mind me
00:23
<Hixie>
IE8 is broken too
00:23
<ap>
Hixie: oh. I guess we shouldn't care then
00:23
<Hixie>
I vote that we don't fix this. Every browsers is going to have this page broken in the same way in the coming months anyway.
00:23
<Hixie>
they'll fix it.
00:23
<Hixie>
interop ftw?
00:24
<zewt>
heh the key to breaking broken pages: get every browser to expose the bug simultaneously
00:26
<jennb>
I'll update the bug.
00:28
<jennb>
Hixie: thanks for checking all the browsers.
00:28
<Hixie>
np
00:29
<Hixie>
thanks for bringing it to my attention :-)
00:29
<Hixie>
it's nice to see all this work we're doing is actually bringing browsers together, even if we're having to break a few eggs in the process
00:29
<AryehGregor>
:)
00:30
<zewt>
breaking rotton eggs--at least do it outside
00:30
<zewt>
also rotten
00:32
<AryehGregor>
maxlength=30 on an address field.
00:32
<AryehGregor>
That is just lame.
00:33
<Hixie>
live somewhere shorter dude
00:33
<Hixie>
wait, you're in NY
00:33
<zewt>
heh
00:33
<Hixie>
aren't your addresses like optimally compressed already
00:33
<Hixie>
23rd NE 5th Ave #123
00:33
<Hixie>
or some such nonsense
00:34
<zewt>
worse is "verify your email address" fields with onpaste="return false"
00:34
<AryehGregor>
Street names are short, avenue names can be long.
00:34
<AryehGregor>
Apartment number adds to the length.
00:34
<zewt>
i get annoyed enough at those that I open the DOM and remove the attribute before typing my address twice for no reason
00:34
<AryehGregor>
I often remove annoying maxlength attributes from the DOM, on the theory that anyone stupid enough to add such a useless check is also stupid enough to not do a server-side check.
00:34
<Hixie>
zewt: i do that too (though i just always c/p :-) )
00:34
<AryehGregor>
Usually I'm right.
00:35
<Hixie>
hear hear
00:35
<zewt>
that's the point--it prevents copy and paste on your email address
00:35
<Hixie>
i mean i remove the attribute and type it once and copy it, not type it twice :-P
00:37
<zewt>
one time I pasted my phone number into a phone box, and the scripts on the page took the "123-123-1234" i pasted, took "123-123-12" (presumably it tried to remove hyphens in key events), and then refused to let me focus the field again--every time i clicked the field it went "this field's already done!" and focused the next field
00:37
<zewt>
sometimes pages feel like they were written by the three stooges
00:38
<AryehGregor>
On a somewhat related note, does anyone else find Chrome's autofill feature incredibly haphazard and confusing?
00:38
<AryehGregor>
For me, it stores like ten different versions of the same info, with varying levels of completeness, which it often puts in random places when I select one of them.
00:38
<Hixie>
yes but no worse than any other browsers'
00:38
<AryehGregor>
I don't think I've used any other browsers'.
00:38
<Hixie>
i've never had them work for me
00:39
<zewt>
firefox is always remembering my credit card information. heh
00:39
<zewt>
"uh ... thanks?"
00:39
<AryehGregor>
But surely you could make it *slightly* saner. Like attempting to merge the ten autofill entries that are all "Aryeh Gregor".
00:40
<zewt>
this page makes me lose all hope: http://www.gmacmortgage.com/
00:41
<zewt>
("spot the problem in under five seconds")
00:42
<AryehGregor>
"Flash"
00:42
<zewt>
well, that's the underlying problem, but not the important one
00:42
<gsnedders>
AryehGregor: "Saviour of the universe"?
00:43
<zewt>
it's a login page that isn't https ... and you can be sure no regular users notice
00:43
<hober>
gsnedders AryehGregor: Terminator X it!
00:43
<AryehGregor>
Not using HTTPS is a noteworthy problem in your books?
00:43
<zewt>
for a login page?
00:44
<AryehGregor>
What percentage of login pages actually use HTTPS?
00:44
<AryehGregor>
Like 0.2%?
00:44
<AryehGregor>
The large majority of sites don't support HTTPS for anything at all.
00:44
<zewt>
for pages where you pay your mortgage?
00:44
<zewt>
i've never once seen an online store or anything else that deals with payment/financial information that isn't https
00:45
<zewt>
now, the login mechanism underneith is flash, and presumably *that* part is still encrypted ... but as it's http, you could MITM it trivially
00:47
<AryehGregor>
If the site's main page doesn't use HTTPS, you can MITM it anyway, unless you expect the user to notice that the login page is missing the lock icon.
00:47
<AryehGregor>
(example: http://www.amazon.com)
00:48
<zewt>
well, normally you do do that--if your main page isn't http, you redirect to https to authenticate, so the https indicators are on
00:48
<zewt>
but that's just the problem--https depends on users noticing that, and they don't, ever
00:48
<AryehGregor>
Which is why everyone needs to use STS instead.
00:48
<Hixie>
there have been studies on this
00:49
<Hixie>
users do check for the lock
00:49
<Hixie>
iirc about 25% of people check for the lock in the content area
00:49
<Hixie>
(and ignore hte browser chrome)
00:49
<Hixie>
some proportion check for the "https" and ignore the broken-lock warnings
00:49
<Hixie>
etc
00:49
<zewt>
Hixie: if that was the case, I'd expect this page to have been fixed long ago--the fact that they don't use http suggests that they don't receive many complaints about it
00:50
<Hixie>
i don't have any data on this specific case
00:50
<zewt>
maybe they get lots of complaints and just ignore them all; don't know
00:52
<AryehGregor>
Look, a useful and detailed bug report from Microsoft: http://www.w3.org/Bugs/Public/show_bug.cgi?id=12091
00:52
<AryehGregor>
I guess they've done those before, to be fair.
00:52
<zewt>
of course, this particular case is broken and insecure, which implies massive incompetence--so I suppose the "receives and ignores complaints" possibility isn't a small one
04:03
<wirepair>
starting to think CORS was a bad idea.
04:04
<wirepair>
i've been finding more problems with old code that expectes xhr.open('GET', user_input_url, true); to be 'safe'
04:04
<wirepair>
:/
07:42
<estellevw>
Does anyone know if there is any browser support for image() fallbacks. The 2009 spec reads body {background-image: image(wavy.svg, wavy.png 150dpi, wavy.gif or blue);}
07:44
<estellevw>
http://www.w3.org/TR/css3-images/#image-notation
08:56
<zcorpan>
wirepair: should we all do the ie way with a different constructor?
08:57
<zcorpan>
XDomainRequest or what it's called
09:19
<mikeke>
Hello all, I have a quick question about the history API
09:21
<mikeke>
is there something like the same domain policy on it?
11:54
<annevk>
on the train in Sweden
11:54
<annevk>
it has the wireless
11:54
<annevk>
MikeSmith, you around?
11:54
<annevk>
MikeSmith, was wondering about Web Notifications
11:55
<jgraham>
Sweden?
11:55
<jgraham>
Also, lucky you :) They're not all that nice
11:56
<annevk>
Stockholm for a few days
11:56
<jgraham>
Unless by "the wireless" you just mean radio ;)
11:56
<annevk>
will come visit you early March :)
11:56
<jgraham>
Yes, it is written on out whiteboard
11:56
<jgraham>
*our
11:56
<annevk>
haha
11:57
<annevk>
hg push works
11:57
<annevk>
and I saw they have 4G here too
11:57
<annevk>
up to 80mbit
11:58
<annevk>
at home I've 120mbit, but that's wired
12:00
<jgraham>
Only in Stockholm though
12:00
<jgraham>
afaik
12:03
<jgraham>
Oh, Malmö, Göteborg + Karlskrona
12:03
<jgraham>
And more polling out this year
12:03
<jgraham>
*rolling
12:03
<annevk>
neat neat
12:03
<annevk>
guess Linkoping has too much students for it to be interesting as a starting point
12:05
<jgraham>
550 SEK/month is not cheap
12:05
<jgraham>
And it looks like the dongle things have the same crappy build quality (and probably the same crappy drivers) as the 3G ones
12:06
<annevk>
seems we have arrived
13:39
<beowulf>
anyone know which browsers implement w3c widgets?
13:50
<zcorpan>
i know opera does
13:51
hsivonen
isn't aware of any non-Opera browser supporting them
13:51
<hsivonen>
I hear Apache has a non-browser widget runtime
13:52
<hsivonen>
and darobin had at least a prototype-level Gecko-based runtime
13:58
zcorpan
kills the 5 on http://forums.whatwg.org/
14:00
<beowulf>
cool, thanks
14:12
<jgraham>
So if I have <table><form><input> and I remove/reinsert the <input> from the DOM, what form (if any) should it end up associated with?
14:20
<zcorpan>
jgraham: i think the spec has an algorithm for removing/inserting form controls
14:20
<zcorpan>
jgraham: which iirc kills the parser's form pointer association and just checks form="" and ancestors
14:21
<jgraham>
zcorpan: That's what I think too
14:21
<jgraham>
But I am being held to account for not thinking that two years ago
14:22
<jgraham>
So I wonder if the spec changed or I am not reading it right now
14:22
<jgraham>
Also, that's not what Gecko/WebKit do
14:39
<MikeSmith>
wonder if anybody has taken a look at http://www.w3.org/TR/2010/NOTE-test-methodology-20100128/
14:42
<hsivonen>
documents whose title starts with "A Method for" make me think "patent! run away, run away"
14:42
<hsivonen>
well, at least it's not "Apparatus and Method"
14:42
<MikeSmith>
heh
14:51
<jgraham>
MikeSmith: I have scanned it before
15:50
<MikeSmith>
anybody know how much hardware support OpenSL ES has?
15:50
<MikeSmith>
http://www.khronos.org/opensles/
15:50
<MikeSmith>
"OpenSL ES™ is a royalty-free, cross-platform, hardware-accelerated audio API tuned for embedded systems"
15:52
<MikeSmith>
jgraham: any opinion on the http://www.w3.org/TR/2010/NOTE-test-methodology-20100128/ idea… it seems to mostly just amount to putting IDs on all testable assertions in a spec
15:53
<jgraham>
MikeSmith: I think my opinion is roughly "I'm glad I'm not trying to convince Hixie to do that"
15:53
<MikeSmith>
heh
15:53
<MikeSmith>
indeed
15:56
<Philip`>
HTML5 doesn't have a 1:1 correspondence between assertions and elements
15:57
<Philip`>
or between assertions and continuous sequences of text
15:58
<Philip`>
e.g. it often says "If X, you must Y. Otherwise, you must Z." where the testable assertions are identifiable as "If X, you must Y" and "If X, [...] Otherwise you must Z"
15:59
<Philip`>
(You don't want the assertion to just be identified as "Otherwise you must Z" because that loses a critical part of the context)
16:00
<Philip`>
With the canvas tests I have assertion patterns like
16:00
<Philip`>
"When the UA is passed an empty string<^> <...> then it *must* return null."
16:00
<Philip`>
"When the UA is passed <...> a string specifying a context that it does not support<^>, then it *must* return null."
16:00
<Philip`>
for that kind of thing
16:00
<MikeSmith>
OK
16:01
<MikeSmith>
so you're saying your annotations-injection mechanism handles those cases correctly?
16:01
<MikeSmith>
the one you're using for the annotated canvas spec, I mean
16:01
<Philip`>
where <...> means any number of characters (within the current paragraph), and <^> means that's where the annotation should be rendered in the spec, and *must* means that's the relevant normative keyword
16:02
<Philip`>
Yeah, the <...> thing means it's possible for the assertion to include the relevant context without including irrelevant phrases in the middle of the sentence (which might be part of a different assertion)
16:03
<Philip`>
HTML sadly is tree-structured so you can't really do that in the markup
16:04
<Philip`>
(at least not without rewriting the text)
16:05
<Philip`>
(and I don't think rewriting the text of HTML5 would be a necessarily good idea)
16:05
<jcranmer>
starting over from scratch is always a good thing
16:06
<jcranmer>
just ask the developers of Duke Nukem Forever
16:06
<jgraham>
Curiously this is quite close to the first use case I heard for graph-like markup which was annotating the bible
16:07
<jgraham>
Which I assume, without having any particular knowledge, has similar problems
16:07
<Philip`>
I imagine people would object if you started over from scratch on the Bible
16:08
<jgraham>
Well it would be nice to replace the glaring factual errors</flamebait>
16:08
<jcranmer>
Philip`: it's called the Koran ;-)
16:09
jgraham
wasn't under the impression that the people who wrote the Koran started out with the Bible and decided to rewrite it from scratch
16:10
<jcranmer>
well, arguably, everyone who starts their own religion and eschews older texts and writes their own is arguable rewriting it from scratch
17:40
<MikeSmith>
jgraham: see my reply on testing requirement thread
17:40
<MikeSmith>
I just created http://www.w3.org/wiki/TestInfra/goals
17:41
<MikeSmith>
or rather, copied content we had elsewhere over to there
17:47
<wilhelm>
Cool.
17:47
<wilhelm>
Re: “allow more than one way to test functionality: … anything else?”, I'd like to know what other browser vendors are doing for tests that require some sort of human interaction. Requiring humans to validate test results doesn't scale.
17:49
<wilhelm>
There are Watir and WebDriver implementations for all major browsers now, with varying degrees of vendor support.
17:49
<gsnedders>
wilhelm: Gecko and WebKit pretty much don't do any, just rely upon bugs from users of nightlies, AFAIK
17:51
<wilhelm>
Opera has a test suite for user interaction with forms with a few hundred tests we could probably contribute. That uses the Watir API and emulates real clicks and keyboard events. The same approach might be useful for other features too.
17:53
<wilhelm>
gsnedders: Then this might be a good time to decide on something everyone can use in the future. (c:
17:55
<gsnedders>
wilhelm: WebKit also has https://trac.webkit.org/wiki/Writing%20Layout%20Tests%20for%20DumpRenderTree#DumpRenderTreeJavaScriptEnvironment
17:56
<wilhelm>
Yes, I've seen that one. Looks useful.
17:57
<gsnedders>
wilhelm: Gecko does something similar at a JS level for testing interaction stuff
19:28
<Ms2ger>
Gecko has non-automated litmus tests too
19:28
<Ms2ger>
https://developer.mozilla.org/en/Litmus_tests
19:44
<MikeSmith>
Ms2ger: odd choice of a name for a class of tests
19:44
<MikeSmith>
is the name meant to be ironic?
19:45
<Ms2ger>
Dunno, why?
19:46
<MikeSmith>
http://en.wikipedia.org/wiki/Litmus_test_(politics)
19:46
<MikeSmith>
that term has negative connotations
19:46
<MikeSmith>
seems like it's normally a pejorative term
19:47
<othermaciej>
in the context of politics it has acquired such a connotation
19:47
<othermaciej>
but the original term from science is neutral
19:47
Ms2ger
assumes chemistry is a more likely source
19:47
<othermaciej>
and it seems totally reasonable to have a binary pass-fail test for software
19:48
<othermaciej>
well, acid vs. base isn't really a good vs. bad judgment
19:55
<mikekelly>
anyone about?
19:55
<Ms2ger>
Nooooooooooooooooooooo
19:56
<mikekelly>
:-)
19:56
<mikekelly>
I've got an edge-case HTTP behaviour I was wondering if it had been considered on the browser side..
19:57
<mikekelly>
If a 200 response to /foo has a Content-Location header of /foo.html - is it feasible for browsers to adjust their address bar appropriately?
19:59
<mikekelly>
same thing for a 201 as well
20:11
<AryehGregor>
That sounds like it's exploitable to prevent users from copying URLs from the URL bar.
20:12
<mikekelly>
AryehGregor: does the history api suffer from that problem?
20:12
<AryehGregor>
Dunno about the history API.
20:13
<mikekelly>
I assume it would, don't know for sure though
20:19
<gsnedders>
mikekelly: It would be backwards incompatible with deployed content (which relies upon the header being ignored)
20:21
<mikekelly>
gsnedders: what sort of deployed content are you thinking of, are there any examples?
20:26
<gsnedders>
mikekelly: http://support.microsoft.com/default.aspx?scid=http://support.microsoft.com:80/support/kb/articles/q218/1/80.asp&NoWebContent=1
20:30
<mikekelly>
gsnedders: that's a bug, right?
20:31
<gsnedders>
mikekelly: It's a bug in IIS, yes. But until all the servers using buggy versions of IIS are gone, there's no way browsers are even going to seriously consider supported Content-Location.
20:35
<mikekelly>
gsnedders: so browsers never make changes that could break buggy applications?
20:38
<AryehGregor>
They never make changes that will cause their users to complain significantly, unless the practical benefit is really worth it.
20:38
<gsnedders>
mikekelly: Pretty much. If a site works in version x of a browser and not in version x + 1, users assume the browser is buggy and try another browser.
20:38
<mikekelly>
AryehGregor: practical to who'm ?
20:39
<AryehGregor>
To the browser implementers.
20:39
<AryehGregor>
Or, by extension, their users.
20:39
<AryehGregor>
Or, by extension, the authors of the pages their users visit.
20:39
<AryehGregor>
(since their goal is to make users happy)
20:39
<AryehGregor>
(their users)
20:39
<mikekelly>
as an author I can tell you that this CL behaviour would be greatly beneficial
20:41
<AryehGregor>
Doesn't matter, if it will cause too many users to complain to the browser implementers.
20:41
<AryehGregor>
Even if it didn't, I very much doubt your claim that this will be greatly beneficial to the author of almost any page that browser users visit, weighted by popularity.
20:42
<mikekelly>
maybe we could meet half way and look at trialling an additional X-Content-Location header with same behaviour?
20:44
<mikekelly>
gsnedders: if we did the above we have the best of both worlds
20:49
<mikekelly>
or better "X-Content-Location-Display: true"
21:34
<Ms2ger>
Reftests almost done...
21:40
<gsnedders>
Ms2ger: For what?
21:40
<Ms2ger>
For running them
21:44
<gsnedders>
Ms2ger: And there you had more all excited about more tests. :(
21:45
<Ms2ger>
Go review AryehGregor's ):
21:45
<Ms2ger>
Or :), even
21:57
Ms2ger
will upload tomorrow
22:25
<zewt>
heh, is mozilla's tracker *intended* to be impossible to search, with robots.txt preventing it from being indexed...