00:11
<AryehGregor>
IPv6 not working reliably for you? :)
01:01
<Hixie>
AryehGregor: what would <script>alert('&%dGVzdA==;')</script> alert, if anything?
01:02
<AryehGregor>
Hixie, I don't know, it's not my proposal.
01:02
<Hixie>
(and how about <script>alert(&%Jyk7YWxlcnQoZG9jdW1lbnQuY29va2llICsgJw;)</script>)
01:02
<Hixie>
k
01:03
<Hixie>
ah, anne answered my question
01:10
Philip`
notes that some pages write <a href="foo%20&%20;bar.html">
01:11
<Philip`>
so it's not entirely clear the proposed syntax is safe
01:12
<Hixie>
well the character can be changed, certainly
01:12
<Hixie>
doesn't have to be *%
01:12
<Hixie>
&% even
01:12
<Hixie>
(i was wondering about that myself)
01:13
<AryehGregor>
It could be a multicharacter string, like "xn--". :)
01:13
<jcranmer>
just make it My
01:14
<jcranmer>
who'd use that? ;-)
01:15
<Philip`>
Seems a bit strange to me to design something based on the assumption that people are too uninformed to follow a few simple rules (like always quote attributes and use some particular escape function based on current syntax), but will be able to use a new syntax and escaping function without getting horridly confused
01:15
<Philip`>
particularly since for the next five years people will write articles saying not to use this syntax yet, since it's not backward-compatible with IE6
01:16
<Philip`>
and after that people will continue reading those articles
01:17
<AryehGregor>
The only possible utility I see is easy escaping in <script>, and I'm far from sure that that justifies it.
01:17
<Philip`>
(Also everybody will get the character encoding wrong, and it's terrible for 'view source')
01:17
<AryehGregor>
"View source" can translate it, in principle.
01:17
<Philip`>
Then it wouldn't be 'view source'
01:17
<Philip`>
It'd be 'view something that's a bit like the source but not quite'
01:17
<Philip`>
and anyway my text editor won't translate it when opening an .html file
01:18
<AryehGregor>
Isn't that what "view source" already is, at least in Firefox?
01:19
<Philip`>
No
01:19
<Philip`>
as far as I'm aware
01:19
<Philip`>
unless you mean 'view selection source'
01:19
<Philip`>
which seems fancier
01:19
<AryehGregor>
I think it mangles the source. Maybe I'm wrong.
01:20
<Philip`>
Wouldn't this new syntax break blacklist-based XSS filters because you could write <span style="&%XhwcmVzc2lvbjphbGVydCgnb2ggbm8nKQo=;"> etc and sneak in forbidden properties?
01:21
<AryehGregor>
Probably.
01:21
<Philip`>
(Uh, I got that syntax all wrong)
01:27
<Hixie>
blacklist-based XSS filters are already broken
01:29
<Philip`>
People seem to design them to be sufficiently complex and conservative to successfully work in practice
01:30
Philip`
thought MediaWiki had something like that
01:31
<wirepair>
really?
01:32
<wirepair>
every site i've ever tested has failed
01:32
<wirepair>
;)
01:33
<wirepair>
especially when developers are taking user input and assigning it inside of <script> </script> blocks
01:33
<wirepair>
it's impossible to create a blacklist for that
01:34
<Philip`>
http://svn.wikimedia.org/viewvc/mediawiki/trunk/phase3/includes/Sanitizer.php?view=markup#l744 - that kind of thing
01:35
<wirepair>
input validation/filtering is different than black listing
01:35
<wirepair>
when i think of blacklisting i'm thinking like lookign for <script> tags
01:35
<wirepair>
or whatever
01:36
<Philip`>
preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value )
01:36
<Philip`>
That's the blacklisting
01:36
<Philip`>
and first it does some filtering to remove escaping and spaces and comments and other potentially dangerous things
01:37
<Philip`>
but it looks like it probably will accept &%...; input unchanged
01:37
<Philip`>
which lets the input circumvent the blacklist
01:38
<wirepair>
yeah
01:38
<wirepair>
where is this new proposal anwyays
01:38
<Philip`>
WHATWG list
01:39
<wirepair>
got it
01:50
<wirepair>
hu, i mean i think it makes sense and protects elements in some cases
01:56
<mpilgrim>
othermaciej: yt?
01:56
<othermaciej>
mpilgrim: good day sir
01:56
<mpilgrim>
i'm having reports that http://diveintohtml5.org/detect.html crashes the iPad browser
01:57
<mpilgrim>
do you have access to an iPad to verify that?
01:57
<othermaciej>
sure I can test
01:58
<othermaciej>
cannot reproduce
01:58
<othermaciej>
it loaded fine for me
01:58
<mpilgrim>
well shit
01:59
<mpilgrim>
source of the bug report claims it happens for others too: http://twitter.com/al3xandru/status/22220332275
01:59
<mpilgrim>
the only recent change i made was the font-face declaration, it now includes an SVG font (generated by fontsquirrel's awesome font-face generator)
02:00
<mpilgrim>
i'm waiting on reports about whether http://diveintohtml5.org/test.html (which is the same page, but without the SVG font-face declaration) also crashes
02:00
<paul_irish>
i can repro the crash on iPad
02:00
<mpilgrim>
aha
02:01
<mpilgrim>
when does it happen? on load?
02:01
<paul_irish>
test.html does not crash.
02:01
<mpilgrim>
hooray!
02:01
<mpilgrim>
er,
02:01
<mpilgrim>
you know what i mean
02:01
<paul_irish>
(yeah on load)
02:01
<mpilgrim>
ok
02:01
<mpilgrim>
it's definitely the SVG fonts then
02:01
<paul_irish>
the only difference is the url('f/essays1743-italic-webfont.svg#webfontmRs1TiYo') format('svg') and such?
02:01
<mpilgrim>
yes
02:02
<mpilgrim>
5 or 6 instances -- two fonts with different styles/variants
02:02
<paul_irish>
Yeah sounds like ripe territory for a bug.
02:03
<othermaciej>
I tried a few more times and still no crash
02:03
<mpilgrim>
well, crashing is definitely a bug -- dunno if it's exploitable, but still
02:03
<mpilgrim>
i'd hate to be responsible for the next major jailbreak exploit :-D
02:03
<othermaciej>
I'm running iOS 3.2.2
02:03
<othermaciej>
(7B500)
02:03
<paul_irish>
3.2 (7B367)
02:04
<othermaciej>
would love to hear if it still happens for you after updating
02:04
<othermaciej>
I think odds are good that this is a fixed bug, as I recall hearing about a bug with svg fonts
02:05
<mpilgrim>
i wonder if google's font API team knows about this
02:05
<paul_irish>
i'll ping them about it.
02:06
<mpilgrim>
not sure how we would detect it
02:06
<mpilgrim>
UA sniffing, i guess :(
02:06
<mpilgrim>
hard to detect a browser crash ahead of time
02:06
<mpilgrim>
what with the halting problem and all that
02:06
<paul_irish>
yup. would have to.
02:07
<othermaciej>
people tend to update iOS pretty quickly, though maybe a bit less for for minor updates
02:07
<mpilgrim>
are you willing to upgrade your iPad for science?
02:07
<mpilgrim>
(@paul_irish)
02:07
<paul_irish>
dont have a cable handy at the moment, but as soon as we do, certainly will.
02:08
<mpilgrim>
ok
02:08
<mpilgrim>
i'll leave those SVG fonts up at their current locations, but i'm going to update my HTML to stop referring to them
02:09
<mpilgrim>
in lieu of doing minor-version UA sniffing, which is the kind of extreme optimization that my employer might be willing to do, but I personally am not
02:21
<variable>
Hixie, around ?
03:08
<wirepair>
Hixie do you have any license for your live DOM Viewer? I'd like to include it in my browser testing kit
03:33
<Hixie>
wirepair: what license is your browser testing kit?
03:41
<wirepair>
haven't decided yet ;)
03:41
<wirepair>
whatever license says you can use it / modify it, but can't just sell it without giving back source
03:46
<Hixie>
if you use MIT, BSD, GPL 2 or 3, or Apache 2, feel free to just use it under the same license
03:47
<wirepair>
thanks, is there any server side code? or what you see is what you get at http://software.hixie.ch/utilities/js/live-dom-viewer/
03:54
<variable>
wirepair, just use the BSD licence - simple. easy. free. ;)
03:54
<Hixie>
there's some server-side code for the saving/loading, iirc
03:54
<Hixie>
but should be trivial enough to reimplement
03:54
<variable>
Hixie, about the type="username" proposal. If I understand correctly the issue was lack of any UA difference?
03:56
<Hixie>
variable: the main issue was that we already have solutions in place that people aren't using, so there's no reason to believe anyone will use new features
03:56
<Boogyman>
html5 discussion?
03:56
<variable>
Hixie, you don't think that browser vendors have more of a focus on HTML5 over other standards?
03:57
<Hixie>
like which?
03:57
<Hixie>
by "people" i meant authors, not browser vendors
03:58
<variable>
Hixie, or document authors. I'm fairly knolegeable of various web standards - but I've never heard of RFC3106
03:58
<Hixie>
well, you've heard of it now :-)
03:59
<variable>
Hixie, only cause I followed that specific proposal before it made it to HTML5. For the document authors that care about semantics and use the HTML5 spec - but don't follow the development of the spec - they won't ever hear of it
04:00
<variable>
and IMHO a document author should not be expected to follow the WHATWG mailing list
04:00
<variable>
in order to find out how to mark up things properaly
04:00
<variable>
*properly
04:02
<Hixie>
variable: i think plenty of other specifications have managed to get plenty of exposure without being in the technology of the day (be that html5 or whatever)
04:03
<Hixie>
variable: if peoplereally want a solution, they'll use the one that exists
05:40
<MikeSmith>
brianleroux++
05:40
<MikeSmith>
for lots of reasons
05:42
<MikeSmith>
but at the moment, for http://github.com/brianleroux/wtfjs
08:05
<annevk>
hmm, charset registration is also failing it seems
08:05
<annevk>
3 months without answer
08:05
<annevk>
actually 4
08:06
<annevk>
geez
09:07
<virtuelv>
Neat; http://joshduck.com/periodic-table.html
09:08
<annevk>
added responseBlob to XMLHttpRequest Level 2
09:09
<annevk>
somewhat tempted to also add followRedirects
09:11
<virtuelv>
annevk: unconditionally follow them, or with limitations?
09:12
<annevk>
xhr.followRedirects would be a boolean
09:12
<annevk>
true by default (as it is now)
09:12
<annevk>
when set to false you simply get the response
09:12
<annevk>
e.g. 302 BLAHDIEBLAH\n\nLocation: somewhereelse
09:13
<annevk>
euh \r\n but you get the idea
09:13
<annevk>
it's low-level but allows people to implement everything they want
09:15
<virtuelv>
annevk: are there any security implications?
09:15
<annevk>
not as far as I know
09:17
<virtuelv>
off-site redirects and CORS?
09:17
<annevk>
virtuelv, off-site redirects are required to carry CORS headers
09:17
<annevk>
already
09:18
<virtuelv>
so no additional implications. good
09:18
<annevk>
i.e. cross-origin redirect responses without CORS headers result in a network error today
10:19
<jgraham>
(btw whatwg.org seems to be down)
10:26
<hsivonen>
jgraham: wfm
10:29
<jgraham>
Oh
10:31
<jgraham>
hsivonen: Fails for me from two different locations
10:35
<hsivonen>
http://lists.w3.org/Archives/Member/w3c-wai-pf/2010JulSep/0000.html (Member-only)
10:37
<annevk>
hsivonen, wtf
10:38
<annevk>
must be some fatal misunderstanding
10:38
<annevk>
also sad why they make all kinds of assumptions instead of simply sending an email with the question
10:40
<jgraham>
Yeah, WTF
10:40
<gsnedders>
Um, okay.
10:40
<hsivonen>
annevk: I wonder how many 'requirements' come about in a comparable way
10:41
<hsivonen>
but see also 0008
10:46
<annevk>
keep getting surprised
10:57
<hsivonen>
Why does WebKit trt to clone IE where IE disallows innerHTML setter?
10:57
<hsivonen>
when html5 makes it work on every html node
10:57
<hsivonen>
s/trt/try/
11:11
<micheil>
morning Rik`
11:20
<Rik`>
hi micheil
11:47
<annevk>
hsivonen, do they still do that with the new HTML parser?
11:56
<hsivonen>
annevk: looks like it but I'm not sure
13:47
<annevk>
xhr.followRedirects defined
13:47
<annevk>
including support in CORS
19:39
<Hixie>
should i just s/float/double/ ?
19:39
<Hixie>
people keep asking for changes along those lines?
19:40
<micheil>
float as in 1.5567633
19:40
<micheil>
yeah?
19:43
<Hixie>
yeah
19:44
<micheil>
makes sense I guess
19:44
<micheil>
doesn't bother me either way
19:47
<Philip`>
What's the value in using float?
19:47
<Hixie>
no idea
19:47
<Hixie>
to be honest
19:47
<Philip`>
I assume the harm is that it doesn't match JS Numbers so you need rounding
19:49
<Philip`>
What APIs using floating point values?
19:49
<Philip`>
s/using/use/
19:50
<Hixie>
anything that uses floating point numbers except valueAsNumber
19:51
<Hixie>
except the latter is currently buggy because of http://www.w3.org/Bugs/Public/show_bug.cgi?id=9886
19:51
<Hixie>
i guess we'll change to double and see who complains
19:52
<Hixie>
all the canvas stuff is float
19:53
<MikeSmith>
micheil: have you had a chance to try http://github.com/aredridel/html5 at all yet?
19:53
<micheil>
umm.. no
19:53
<Hixie>
Philip`: would there be a compat problem if i changed canvas s/float/double/?
19:55
<Philip`>
Hmm, there's some thing some Opera people mentioned
19:55
Philip`
tries to find it
19:57
<Philip`>
CORE-32111
19:58
AryehGregor
infers Opera probably uses JIRA
19:59
<micheil>
MikeSmith: I'll follow up in a min, just talking to someone about some audio stuff..
19:59
<MikeSmith>
micheil: no problem
19:59
<Philip`>
Hixie: http://krijnhoetmer.nl/irc-logs/whatwg/20100818#l-737
20:00
<MikeSmith>
micheil: was just asking out of curiosity
20:00
<Philip`>
Hixie: (If you don't round to float then the value is too large and it throws)
20:01
<micheil>
MikeSmith: yeah, listening to a podcast about audio mixing, and then I managed to get talking to a rep from a company I'm looking at getting some gear from, which was pretty cool
20:01
<MikeSmith>
great
20:01
<micheil>
MikeSmith: as for the html5 parser..
20:01
<Philip`>
Hixie: (It could be made safe by not making drawImage throw exceptions (and treat things as transparent black) but I don't know whether that kind of behaviour change has other problems)
20:02
<micheil>
MikeSmith: if it uses JSDom, then it's probably going to be pretty good
20:02
<Hixie>
Philip`: yeah whether to fire an exception or not based on what you're dividing seems suboptimal
20:03
<MikeSmith>
micheil: cool, thanks… I'm just starting to take a look at some of this stuff today
20:03
<micheil>
MikeSmith: sure thing, doing node knockout?
20:03
<MikeSmith>
nope
20:03
<MikeSmith>
I lack the chops
20:03
<MikeSmith>
at this point
20:04
<MikeSmith>
looking forward to seeing the results though
20:04
<hober>
A bug in jsdom has been driving me crazy for weeks
20:04
<MikeSmith>
spectator sports
20:04
<MikeSmith>
hober: what kind of bug?
20:05
<hober>
setAttribute doesn't seem to work
20:05
<hober>
http://gist.github.com/551578
20:06
<Hixie>
Philip`: maybe i should say to round to whole pixels
20:08
<Hixie>
no that wouldn't work for svg inputs or something
20:08
<Hixie>
like that
20:08
<Hixie>
hmm
20:08
<Philip`>
Hixie: That would seem a bit weird since the canvas API currently never cares about pixels (except the (get|put)PixelData aberrations)
20:08
<Hixie>
yeah
20:09
<Hixie>
i'll just make it transparent black outside and see if people complain
20:09
<hober>
MikeSmith: it's entirely likely that I'm just doing something stupid, but I haven't been able to figure out what that might be
20:11
<AryehGregor>
Does anyone know how to get e-mail notification from the IE feedback thing? I haven't figured out if there's any way.
20:11
<AryehGregor>
Do they really expect me to check manually all the time?
20:12
MikeSmith
looks through hobercode to see if he can find anything that looks funny
20:12
<Philip`>
AryehGregor: I remember getting emails from them automatically
20:12
<Philip`>
(on bugs that I filed)
20:13
<AryehGregor>
Hmm.
20:13
<AryehGregor>
Maybe they got filed as spam.
20:14
<AryehGregor>
Or maybe something choked on the address I provided.
20:15
<AryehGregor>
Interesting, this site's security certificate is revoked. I wonder how that happened: https://profile.microsoft.com/RegSysProfileCenter/Error.aspx
20:16
<AryehGregor>
Actually, the same is true for all of https://profile.microsoft.com/. Nice.
20:16
<AryehGregor>
I like how Chrome refuses to give me *any* option *whatsoever* to continue.
20:16
<AryehGregor>
This kind of security theater is asinine, no real attacker with half a brain would use a bad certificate and pop up a scary warning.
20:17
<AryehGregor>
Firefox doesn't pop up any warning at all . . .
20:18
<Philip`>
Maybe Google decided to revoke all of Microsoft's certificates for a laugh
20:19
<AryehGregor>
Also, editing my profile is completely and totally broken on Minefield/Linux.
20:19
<AryehGregor>
It has required drop-down fields that it gives me no options for.
20:20
<AryehGregor>
I used to think that Microsoft was actually competent at programming and just got lazy due to lack of competition, but sometimes I have doubts.
20:21
<Hixie>
AryehGregor: no real attacker with half a brain would use a bad certificate and pop up a scary warning, but if no scary warning popped up, they could just use any arbitrary certificate
20:21
<AryehGregor>
Like every time I use IE8, it prompts me about some inane thing I don't care about (not sure what), and the only "go away" option it gives is "Ask me later". Which it does. Forever.
20:21
<Hixie>
AryehGregor: so in this particular case, it's not security theatre
20:22
<AryehGregor>
Hixie, no, but you could just pop up a scary warning and allow easy clickthrough. Attackers still would be very unlikely to trigger that, so it's fine to allow reasonably easy clickthrough.
20:22
<AryehGregor>
Chrome in this case allowed no clickthrough *at all*.
20:22
<AryehGregor>
That's just stupid.
20:22
<Hixie>
people click through things without reading them
20:22
<AryehGregor>
Sure.
20:22
<Hixie>
so it wouldn't work
20:22
<Hixie>
didn't work
20:22
<Hixie>
anyway, lunch time, bbiab
20:22
<AryehGregor>
No, it would work fine. Because attackers are unlikely to ever trigger the error, since they can typically avoid it just as easily.
20:23
<AryehGregor>
So many users will just click through without reading, but that's okay, because it's almost certainly not an attack.
20:23
<AryehGregor>
The authors of this paper were unable to find a single case, ever, anywhere, where attackers used an invalid certificate: research.microsoft.com/en-us/um/people/cormac/papers/2009/SoLongAndNoThanks.pdf
20:24
<AryehGregor>
Hmm, looks like Chrome auto-http://-adding is buggy.
20:25
AryehGregor
registers and verifies his address, maybe didn't verify it before
20:27
<AryehGregor>
Anyway, the point is, certificates are obnoxiously complicated and an innocuous error is overwhelmingly more likely than an actual attack. Even if there were a tiny number of attacks, the inconvenience it causes is surely not even close to worth it.
20:28
<AryehGregor>
The concern should become basically obsolete with STS, anyway, when all e-commerce and other important sites switch to using that.
20:28
<AryehGregor>
Since that mandates hard failure, and actually closes up the gaping hole of HTTP redirecting to HTTPS.
20:29
<AryehGregor>
Of course, once DNSSEC is deployed, it will be very simple: use the cert from DNS no matter what, and if there's an error authenticating it, you can't even resolve the domain name, so people will fix it pretty quickly. :)
20:30
<AryehGregor>
SRP would be even better, though. That's insanely hard to get wrong, if you have a basic implementation. No certificates to juggle. And best of all, it completely kills phishing, which certificates can never do.
20:31
<AryehGregor>
But something like Account Manager is a prerequisite for that. Go Firefox!
20:31
AryehGregor
checks channel stats to see if they have a stat for "talks to himself", and if anyone else is even in the running for that one
20:33
<AryehGregor>
Hixie and MikeSmith beat me.
20:33
<AryehGregor>
:(
20:36
<AryehGregor>
Anyway, I tried arguing with the Chromium security mafia before, and they didn't listen to me, so I won't bother again.
20:37
<AryehGregor>
(I should say, the person who responded did respond in some detail to all of my points, but didn't back down even though the behavior in that case was clearly broken.)
20:39
<MikeSmith>
hober: fwiw, I get same result as you for your test file
20:39
<MikeSmith>
in my environment
20:42
<MikeSmith>
that is, no attributes
20:42
<MikeSmith>
at all
20:43
<hober>
hmm
20:44
<hober>
well, kindly upvote http://github.com/tmpvar/jsdom/issues#issue/9 and hopefully tmpvar will fix it. :)
20:44
<hober>
I keep meaning to carve out some time to debug what jsdom is doing myself, but never manage to
20:44
<MikeSmith>
hober: maybe worth asking on #Node.js to see if anybody else has insight
20:45
<hober>
yeah, I've asked, and have been warnocked
20:52
<Philip`>
Hixie: Did you intentionally change the drawImage behaviour without mentioning it in the commit message?
20:52
<Philip`>
(Seems like people might miss it)
20:53
<Philip`>
(Although, I suppose I should just update my test cases and they'll notice when they look into their failures)
21:03
<hober>
MikeSmith: my impl of the html->atom algorithm fails pretty spectacularly when it can't create attributes, as I'm sure you can imagine
21:03
<MikeSmith>
heh
21:09
<Hixie>
Philip`: yeah
21:53
<MikeSmith>
silly that github doesn't support display of HTML-formatted readme files
21:56
<AryehGregor>
It would have to sanitize them.
21:57
<AryehGregor>
As the people in this room should know, sanitizing arbitrary HTML is quite nontrivial.
22:07
<hober>
Why am I even having this conversation? (Re: Bug 10455) It's like talking to a brick wall.
22:12
<AryehGregor>
Talking to brick walls is much less stressful, since they don't talk back.
22:12
<AryehGregor>
Possibly more productive, too.
22:12
<AryehGregor>
I heartily recommend it, although walls made out of wood or sheetrock work fine too.
22:23
<jgraham>
Hixie: I don't suppose you have the overwhelming desire to comment in the opera bug that Philip` mentioned earlier, so you?
22:24
<jgraham>
s/so/do/
22:24
<jgraham>
So that we don't fix to the old behaviour
22:24
jgraham
can do it of course
22:25
<Hixie>
uri?
22:31
<AryehGregor>
Do ex-employees keep access to Opera's bug tracker?
22:31
<Hixie>
some non-employees have access
22:31
<Hixie>
to some parts of the bug tracker
22:31
<Hixie>
not a direct correlation to previous employee status
22:35
<jgraham>
CORE-32111
22:35
<jgraham>
So er https://bugs.opera.com/browse/CORE-32111
22:39
<jgraham>
AryehGregor: See e.g. http://operawiki.info/TheElektrans
22:41
<jgraham>
Hixie: (URL above btw)
22:41
<Hixie>
jgraham: yes, i commented as soon as you pasted it :-)
22:46
<jgraham>
Hixie: Ah, wasn't sure if you had seen it :)
22:52
<sicking>
Hixie: with regards to .value. Is IE the only browser that changes the content attribute when .value changes?
22:52
<sicking>
Hixie: i've always liked the IE model and would love to switch to it, but it might not be safe to do so if only IE is doing that :(
22:56
<AryehGregor>
This alerts null in Minefield, Chrome dev, Opera 10.60, and "foo" in IE6 (ies4linux). <!doctype html><script>var el = document.createElement("input");el.value = "foo";alert(el.getAttribute("value"));</script>
22:56
<AryehGregor>
(if that's what you meant)
23:05
<Dashiva>
Yeah
23:06
<Dashiva>
.defaultValue assigns to @value, I believe
23:42
<Hixie>
sicking: the IE model has all kinds of problems
23:42
<Hixie>
sicking: e.g. it makes browser autocomplete mutate the DOM
23:42
<Hixie>
sicking: and it just moves the default value to an IDL attribute instead of the current value
23:43
<sicking>
Hixie: the latter sounds like a much smaller problem than the general confusion of having .value != getAttribute("value")
23:43
<sicking>
Hixie: the former i don't quite understand, please elaborate
23:43
<Hixie>
.value is the current value of the control, right?
23:43
<sicking>
yup
23:44
<Hixie>
if changing .value changes the DOM, then when the UA loads the page and autofills a control, the DOM has to change
23:44
<Hixie>
so you're making the browser mutate the DOM on page load
23:44
<Hixie>
this has all kinds of issues
23:44
<sicking>
Hixie: why is that a problem?
23:44
<sicking>
such as?
23:44
<Hixie>
well it kills round-tripping, for one
23:45
<sicking>
when does autocomplete happen during important round tripping?
23:45
<Hixie>
*shrug*
23:45
<sicking>
the parsing algorithm also kills round tripping, no?
23:46
<Hixie>
seems really bad to me to have the DOM not reflect the markup after load
23:46
<Hixie>
but in any case this is all highly academic, there's no way we can change all this now
23:46
<sicking>
well, autocomplete is basically the UA helping the author type
23:46
<Hixie>
user
23:46
<Hixie>
but yues
23:46
<sicking>
sorry, yes
23:46
<sicking>
and having the DOM change as a result of the user typing doesn't seem bad to me
23:47
<Hixie>
*shrug*
23:47
<Hixie>
tell brendan about 15 years ago
23:47
<sicking>
yeah, probably academic if all other UAs do it the other way
23:47
<sicking>
Hixie: actually, this wasn't a problem until DOM L1
23:47
<sicking>
no getAttribute before that
23:48
<sicking>
and it wasn't really a problem until much later when gecko gained marketshare
23:48
<Hixie>
ok, tell vidur about 12 years ago
23:48
<sicking>
i think it's tell me+jst+a few other about 7 years ago
23:48
<sicking>
or possibly even tell you about 3 years ago when you wrote this
23:49
<Hixie>
by the time i specced it i was just speccing current practice
23:49
<Hixie>
and really put very little thought into the logic of it
23:50
<sicking>
well, you probably at that point could have chosen either IE behavior or gecko behavior
23:51
<sicking>
but possibly too late now i agree
23:51
<Hixie>
the IE behaviour really wasn't anywhere near as logical as you're making out
23:51
<Hixie>
e.g. it threw exceptions when you set .type
23:51
<Hixie>
which i hope we are in agreement is a non-starter