00:13
gsnedders
wonders why Google.com is being so slow
00:14
<jcranmer>
gsnedders: a lot of networks just went down?
00:15
<gsnedders>
I mean, taking a minute to connect to "www.google.com" is wrong.
07:06
<heycam>
Hixie, ping?
07:07
<Hixie>
here
07:08
<heycam>
so i'm thinking about your [XXX]es that are meant to be like "partial" interfaces
07:08
<heycam>
and i see that they all only have attributes on them, no operations
07:09
<heycam>
so that means what you really want is to avoid having a new object in the prototype chain for objects implementing these interfaces
07:09
<heycam>
e.g. you don't want a new, empty prototype object in the chain that corresponds to the interface with the applets attribute on it
07:10
<heycam>
but, i don't think "extending" the interfaces, at the IDL level, is the right thing to do
07:10
<heycam>
you just want to control the JS aspects of it
07:10
<heycam>
so e.g. if these extended interfaces did have operations, then they would go on to the HTMLDocument prototype object (in this case)
07:13
<heycam>
well actually i see you have two kinds of XXXes now
07:13
<heycam>
on WindowModal, where you separate it because sometimes you want this interface to be implemented and sometimes not, but if it is, then there's no extra object in the prototype chain for WindowModal
07:14
<heycam>
and for the things in the Obsolete section, where they're separate just so you don't see them in the earlier sections
07:14
<heycam>
but they are actually part of the same interface
07:16
<heycam>
ok maybe i'm talking myself into allowing "partial" interface definitions
07:16
<heycam>
for that second case
07:16
<heycam>
even though it's only because of how you want to organise the spec
07:16
<heycam>
but for the WindowModal case i think your XXX does need to be something like NoPrototypeObject
07:19
<heycam>
maybe partial should also exist to support adding new constants from different specs, like DOMException codes
07:22
<heycam>
also, i'm thinking of moving a bunch of things from extended attributes into the "proper" syntax
07:22
<heycam>
like optional arguments, variadic operations
07:23
<heycam>
i'm gonna give up pretending web idl is a superset of omg idl
07:26
<Hixie>
(still here, hold on)
07:27
<Hixie>
heycam: basically my desire is to not have the spec separation affect the prototype chain
07:28
<Hixie>
heycam: so we should be able to take, e.g., HTMLDocument, or Window, and split it amongst multiple specs and have zero effect on implementation requirements
07:28
<Hixie>
heycam: basically i'm trying to avoid the problem hsivonen mentions regularly where you end up splitting the tech into silos per wg
07:30
<heycam>
so, having a [NoPrototypeObject] could do that
07:30
<heycam>
especially if the interface only has attributes, since they never go in prototype objects
07:30
<Hixie>
they'll have methods sometimes
07:30
<Hixie>
i'm surprised that they don't already
07:31
<heycam>
maybe something like [AugmentPrototype=Window] instead
07:31
<Hixie>
yes
07:31
<Hixie>
that would work
07:31
<Hixie>
and then the "actual" interface name wouldn't matter?
07:32
<Hixie>
i.e. it'd be an editorial concern only?
07:32
<heycam>
yes, except for other language bindings
07:32
<heycam>
which don't have prototypes
07:32
<Hixie>
well for bindings that have classes i'd want the same thing
07:32
<heycam>
they'd be proper, separate interfaces still
07:32
<heycam>
yeah?
07:32
<Hixie>
the idea is for there to be zero differences
07:32
<heycam>
how should it matter?
07:32
<heycam>
if you say "all HTMLDocument objects also implement HTMLDocumentObselete"
07:33
<heycam>
oh casting would matter
07:33
<Hixie>
don't you have to explicitly cast from one to the other?
07:33
<Hixie>
right
07:33
<heycam>
i mean, you'd need to cast
07:33
<heycam>
hmm
07:33
<heycam>
so it sounds like you exactly want a "partial" modifier on interface
07:34
<Hixie>
the point is we should be able to take HTML5, 10 years from now, when it's all perfectly implemented (hah), and make 15 specs out of it and be able to release that as an Edited Recommendation that has zero changes required to any test suites in any languages
07:34
<Hixie>
with all the same code working unchanged
07:34
<Hixie>
that's basically what i'm doing, btw -- taking DOM2 HTML, and splitting it into the "obsolete" and "today" sections in html5
07:35
<Hixie>
10 years after DOM2 HTML was published
07:36
<heycam>
ok i'm gonna go for a "partial" keyword then, see how it goes
07:36
<Hixie>
cool
07:36
<heycam>
partial interface HTMLDocument : ... { /* now stuff */ }; partial interface HTMLDocument { /* old stuff */ };
07:37
<Hixie>
do i put it on all interfaces?
07:37
<heycam>
hmm
07:37
<heycam>
you want any interface to be extendable?
07:37
<Hixie>
i'm sure there will be specs that add to Window if there aren't already
07:37
<heycam>
right
07:37
<Hixie>
and it would have been nice to be able to add to HTMLElement instead of having the side interfaces, for some cases
07:38
<Hixie>
but i dunno
07:38
<heycam>
yeah i'm not sure
07:38
<Hixie>
i mean, if we have to have it to extend, you're basically saying it's the opposite of "final"
07:38
<heycam>
well, not really
07:38
<heycam>
you can extend with ":"
07:38
<heycam>
but it's a different kind of extending
07:38
<Hixie>
yeah
07:39
<Hixie>
":" and also with "side" interfaces
07:39
<Hixie>
like how EventTarget is on all Nodes
07:39
<heycam>
right
07:39
<heycam>
depends on how you want the prototype chain to look
07:39
<Hixie>
but for some cases like Window that seems... wrong
07:39
<Hixie>
yeah
07:39
<heycam>
yeah Window is a bit special in that regard, in that you think of it as just a pool of properties that you add to
07:39
<Hixie>
i guess in practice people would just say that the interface they want to extend had "partial" on it
07:39
<Hixie>
even if it didn't
07:40
<heycam>
heh perhaps
07:40
<Hixie>
so in practice even if you say we have to put it on both, it won't matter
07:40
<Hixie>
it's just nice to annotate one of them (the "not primary" one, there's usually a clear one that is the "primary" one) as being an extention of another
07:42
<heycam>
it's kinda weird for languages like java to have an interface whose definition changes depending on which specs you decide to implement
07:43
<Hixie>
why?
07:43
<Hixie>
how is that different from when you decide how much of one spec to implement?
07:43
<Hixie>
it's not like spec boundaries reflect actual implementation boundaries, like, ever
07:44
<heycam>
it's different if you want to use two libraries together, say
07:44
<heycam>
let's say a spec first defines interface A
07:44
<heycam>
and a later spec defines an extension to A
07:44
<heycam>
and both of these specs used this "partial" keyword to allow this
07:45
<heycam>
one library was compiled with the first A, and another library was compiled with the extended A
07:45
<heycam>
now both libraries have different A.class files
07:45
<heycam>
whichever one the class loader finds first is the one that will be usable
07:45
<heycam>
you need to be sure to get the extended one loaded first somehow
07:46
<heycam>
but then it's impossible if there were two extensions to A, and each library chose the different extended one
07:46
<heycam>
you couldn't get both libraries working together in the one program
07:48
<Hixie>
what about if you use two libraries, and they both implement a part of the same spec?
07:48
<Hixie>
same problem
07:48
<Hixie>
literally the same problem, because the next day that spec might be split into two
07:49
<Hixie>
and the day after that the specs might be merged together again
07:49
<heycam>
you could at least have the same all-in-one interface implemented by both libraries, but some of those methods just throw a NotImplementedException or something
07:49
<heycam>
in fact you would need to, to get it to compile
07:50
<Hixie>
i guess in java maybe. in javascript we encourage implementations that don't support something to not expose it, so that you can test to see if it exists
07:50
<heycam>
yes
07:51
<heycam>
it's a benefit of not having to implement a whole interface i suppose
07:51
<Hixie>
well it's a benefit of easy reification
07:52
<heycam>
benefit of duck typing (sorta)?
07:53
<Hixie>
vaguely
07:53
<Hixie>
not exclusively though
07:53
<Hixie>
it's a benefit of late binding and easy reification
07:54
<Hixie>
sorry, not reification, reflection
07:54
<Hixie>
late binding and reflection
07:54
<Hixie>
because you can compile code that uses something without it needing to be supported, and then you can test to see if that code would work, and if it doesn't, you can skip it
07:57
<heycam>
so for something like WindowTimers, would you want to do that with partial instead?
08:01
<Hixie>
yeah
08:01
<Hixie>
actually no
08:01
<Hixie>
well
08:01
<Hixie>
gah, i dunno
08:02
<Hixie>
WindowTimers is also implemented by the WorkerUtils object
08:02
<heycam>
ah
08:02
<Hixie>
but i'd want it to appear on the prototype of WorkerUtils (which itself should just appear on the prototype of two other interfaces), and i'd want it to appear as if it was part of Window also
08:02
<Hixie>
and it's specced in the same spec as Window but not the same spec as WorkerUtils
08:03
<heycam>
do any of these use ":"?
08:03
<heycam>
or is it [ImplementedOn]
08:04
<Hixie>
none of these use :
08:04
<heycam>
btw i think i like the "Node implements EventTarget;" kind of syntax
08:04
<Hixie>
oh actually that's a lie
08:04
<Hixie>
the WorkerUtils thing is implemented on interfaces that do use inheritance
08:05
<heycam>
[AugmentPrototype] Window implements WindowTimers;
08:05
<heycam>
or if you're doing something with ":"
08:06
<heycam>
interface ExtendedWindow : [AugmentPrototype] Window ...
08:06
<heycam>
but have Java interfaces change depending on which specs you choose to implement seems bad to me
08:07
<Hixie>
well i don't really mind how you end up doing things for java really
08:07
<Hixie>
i'm mildly professionally interested in how what you do ends up for C++, because of Native Client's use of WebIDL
08:07
<Hixie>
but really my interest lies in JS
08:09
<heycam>
having c++ classes with different members seems as bad as with java
08:09
<heycam>
mixin abstract classes (just like different java interfaces) seems ok though
08:12
hsivonen
raises eyebrows at www-tag
08:15
<hsivonen>
s/www-tag/www-archive-
08:20
<heycam>
Hixie, i'll think about partial/[AugmentPrototype] a bit more, gotta head home now
08:51
<Hixie>
heycam: cool, thanks
11:40
<zcorpan_>
gsnedders: http://gsnedders.html5.org/outliner/process.py?url=http%3A%2F%2Fsimon.html5.org%2Fsandbox%2Fhtml%2Fgoogle-chrome-comic seems wrong
14:40
<annevk42>
http://troogle.wordpress.com/2009/06/17/google-is-trying-to-set-us-up-breaking-news-the-new-web-project-revealed-before-its-release/ wtf
14:40
<annevk42>
especially since http://whois.domaintools.com/thenewwebproject.com points back to them
14:41
<annevk42>
some kind of elaborate joke that did not cover its tracks?
14:42
<Dashiva>
"we believe the GOVERNMENT is involved too"
14:42
<Philip`>
"Google has plans to work together with Microsoft ..." - clearly this whole thing is just a work of fiction
14:43
<Dashiva>
It looks like a parody of that British guy who walked around with visible secret papers
14:44
<annevk42>
http://cocaman.ch/wp/2009/06/the-new-web-project-its-a-movie-and-plot-against-google/
14:46
<annevk42>
they could have at least used a proxy for the whois
15:55
<Philip`>
http://asyourworldchanges.wordpress.com/2009/05/31/hear-me-stumble-around-white-house-recovery-and-data-gov-web-sites/ looks potentially informative
20:08
<ojan>
I know arun's not around, but maybe others here know the answer to my question...why does File API need getAsDataURI and getAsText?
20:09
<ojan>
as in, can't you replace getAsDataURI with a few lines of code wrapping getAsText?
20:18
<annevk2>
i guess it's just a convenience thing
20:18
<annevk2>
not sure it's needed if we get the getAsLocalURL()
20:19
<annevk2>
ojan, seems best to comment on public-webapps
20:20
<ojan>
annevk2: will do. just wanted to make sure i wasn't missing something obvious
20:20
<ojan>
annevk2: yeah, getAsLocalURL does seem very useful to me
20:20
<annevk2>
oh wait, I guess PNG and such won't get you text but will get you data URLs
20:21
annevk2
wonders why he did not think of that
20:21
<annevk2>
but then getAsText might work anyway, it's not very well defined
20:22
<ojan>
i was thinking that if you passed base64 as the encoding to getAsText it ought to work for binary content
20:24
<annevk2>
base64 is not a character encoding, however
20:24
<annevk2>
but then getAsText is underdefined
21:26
<annevk5>
oh yay, some guy is advocating software patents on www-style
21:54
<sicking>
Hixie, ping
22:15
<Hixie>
sicking: here
22:15
<sicking>
Hixie, is there a draft for File drag-n-drop?
22:16
<sicking>
Hixie, we have an intern that's going to start working on it pretty much today
22:16
<sicking>
Hixie, we can do a lot without a spec for now I think, but it'd be nice to have a draft
22:16
<annevk5>
there's a proposal on the WHATWG list
22:17
<Hixie>
i'm waiting for arun's draft to stabalise
22:17
<Hixie>
but basically i expect datatransfer to get some new api for files
22:18
<Hixie>
maybe make getData and setData take |any| instead of DOMString
22:18
<Hixie>
and have them return either DOMStrings for Files
22:18
<Hixie>
or alternatively, have the files be totally separate
22:18
<sicking>
Hixie, ok
22:18
<Hixie>
just a .files or something on DataTransfer
22:19
gsnedders
hits more bugs in PHP. ergh.
22:19
<Hixie>
oh hey, arun is mozilla too, i forgot
22:19
<Hixie>
go light a fire under his chair :-)
22:19
<Hixie>
i think we'll go with just .files
22:20
<gsnedders>
Hixie: But that'll mean he'll have to move away from his computer, thus making him take longer to write the spec
22:20
<Hixie>
gsnedders: it is implied that he'd be tied down first and only released once the spec was done, thus increasing the motivation to work on it :-)
22:20
<Hixie>
sicking: yeah so i imagine we'll add a .files that returns a sequence<File>
22:22
<sicking>
Hixie, ok
22:22
<sicking>
Hixie, ugh, we'll have to add support for sequence<> to gecko...
22:22
<sicking>
Hixie, the idea is that that'll map to a JS array, right?
22:24
<sicking>
annevk5, where is the proposal on the whatwg list? What's the subject line?
22:25
virtuelv
thinks "file" should be the atom here, whether that is a file or collection of files
22:26
<virtuelv>
and treat, at least, zip files as if they were collections of files
22:26
<othermaciej>
sicking: fwiw we have a patch to implement it for WebKit, and we proposed an API on WHATWG
22:26
<virtuelv>
that way, an application could easily process openoffice (and similar) documents like epub
22:27
<othermaciej>
sicking: our version is fairly agnostic with respect to details of Arun's draft
22:27
<sicking>
othermaciej, cool
22:27
<sicking>
othermaciej, do you have a link to the draft you proposed?
22:27
<gsnedders>
Hixie: But then you risk him dying before he finishes the spec, which means you need to find a new editor
22:28
<othermaciej>
http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2009-June/020302.html
22:28
<othermaciej>
sicking: ^
22:30
<Hixie>
that's basically what i just described :-)
22:30
<Hixie>
sicking: sequence<T> is just shorthand for an interface with a length property and an item() property or something like that, i think
22:30
<Hixie>
sicking: basically i just mean your regular DOM list of things
22:32
gsnedders
gives up trying to obtain Lightroom, seeming Adobe keep refusing to sell it to me
22:32
<gsnedders>
Oh well, their lost sale.
22:34
<sicking>
Hixie, ok
22:44
<sicking>
othermaciej, btw, have you had a look at the API that Arun proposed for File?
22:45
<sicking>
othermaciej, I know FileDialog is still controversial, but it'd be nice to hear how you feel about File
22:45
<othermaciej>
sicking: not yet
22:45
<othermaciej>
sicking: on my TODO list of spec review
22:46
<othermaciej>
I just did a careful review of ForceTLS so I'm in a reviewing mood
22:46
<sicking>
hehe
22:46
<sicking>
sweet
22:46
<othermaciej>
sicking: I expect to have issues with it if there's still synchronous file I/O, which I suspect there is
22:47
<sicking>
othermaciej, nope, asynch only
22:47
<othermaciej>
then maybe not so much!
22:47
<sicking>
sweet!
22:47
<sicking>
i know you guys had concerns about googles Blob proposal, which was async only
22:47
<othermaciej>
we're obviously interested in making file access better, and it would be nice to have fine DnD and our <input type="file" multi> fuly spec'd
22:48
<sicking>
not sure what that feedback was though, so not sure if it applies to File
22:48
<othermaciej>
I dunno
22:48
<Hixie>
i'll spec both of those within days of arun's proposal being stable
22:48
<sicking>
othermaciej, Agreed. I think the only controversial part is FileDialog, assuming you're fine with file I/O
22:48
<othermaciej>
right now we're just using the File interface as a dumb handle
22:56
<Hixie>
sicking: oh i got a response back from the gmail guys about the filedialog thing -- if you can get arun here that'd be great to discuss it
22:57
<sicking>
Hixie, you should mail arun and me so we can figure out a day that works
23:00
<sicking>
Hixie, what about storing File in localStorage?
23:00
<sicking>
Hixie, I'd imagine that that's needed for offline-gmail
23:07
<Hixie>
sicking: well i'm free whenever
23:07
<annevk5>
sicking, would that do a copy or reference the file on disk?
23:07
<Hixie>
sicking: gmail uses the sql database stuff for their offline gmail
23:08
<Hixie>
sicking: but yes, they would like a way to store references to Files while waiting for an opportunity to upload them
23:08
<Hixie>
sicking: i was more talking about FileDialog though
23:08
<Hixie>
sicking: turns out they didn't use flash to avoid using <input type=file> at all -- in fact they didn't even list that as a reason for using flash until i specifically asked them about it
23:08
<Hixie>
sicking: and then they dismissed that as not a big deal
23:08
<annevk5>
if you just wanna store references the local:<id> thing would work I think, you can just store those in localStorage
23:09
<sicking>
annevk5, how its stored seems like an implementation detail. I'd imagine sqlite has blob support
23:09
<annevk5>
Hixie, did they use it for async upload and multiple files?
23:10
<Hixie>
sicking: i wouldn't expect the local:<id> thing to survive page load, but i guess that depends on how we define it
23:10
<sicking>
Hixie, yeah, my question about localStorage was orthogonal to the meeting
23:10
<Hixie>
er
23:10
<Hixie>
s/sicking/anne/
23:10
<annevk5>
sicking, it's not an implementation detail, a reference to the file is very different from copying it
23:11
<annevk5>
e.g. the user could modify the file later on
23:11
<Hixie>
annevk5: the two things they said were multiselection of photos, and file upload progress.
23:11
<sicking>
annevk5, ah, yes, we should define that it's the contents of the file at the time of the call to store it in localStore
23:11
<annevk5>
Hixie, yay for me guessing why they used Flash
23:12
<annevk5>
and boo for sicking not believing me :p
23:12
<Hixie>
annevk5: and then they added dnd of files, local display of files while uploading, and selecting or uploading a whole folder at once
23:12
<sicking>
annevk5, i prefer data over guessing
23:12
<Hixie>
and finally they mentioned the <input accept=""> feature (though not by name)
23:12
<sicking>
annevk5, except when I'm the one guessing
23:12
<sicking>
annevk5, :)
23:12
<annevk5>
sicking, so it appears :)
23:12
<Hixie>
oh and resuming uploads after browser crash
23:13
<Hixie>
though that one was more theoretical
23:13
<sicking>
Hixie, flash supports that?
23:13
<annevk5>
i always wonder whether selecting a folder should be a UI feature or some semantic thing like <input type=folder>
23:13
<Hixie>
i think that was after they transitioned from why they used flash to what they wanted most of all
23:14
<sicking>
Hixie, localStorage being able to store files would take care of that
23:14
<Hixie>
anyway their answer to "so, uh, styling of <input type=file> isn't an issue?" was basically "oh i guess that would be nice too" but they basically said it wasn't a big deal compared to the rest
23:14
<Hixie>
so i vote for dropping FileDialog for now
23:14
<sicking>
we'll see
23:14
<Hixie>
sicking: or the sql database, yeah
23:15
<sicking>
Hixie, yeah, though i'm not sure what our timeline is for sql database. I think we still want to see some other API
23:15
<sicking>
Hixie, have had some discussions about that internally recently
23:15
<Hixie>
yeah, we were talking about it in #developers recently too
23:15
<sicking>
ok
23:16
<Hixie>
basically i'm gonna do nothing on the sql front until october and then i'm gonna suggest moving to LC
23:16
<sicking>
mostly due to the pains of standardizing a SQL dialect, and then having to implement a sql parser that maps that dialect to sqlite
23:16
<Hixie>
and after that look into speccing teh actual dialect
23:16
<Hixie>
so you have until then to come up with a better solution :-)
23:16
<sicking>
i don't see how we could move into LC without a dialect
23:16
<Hixie>
which i'd likely support since i have no desire to actual spec out the dialect :-)
23:17
<sicking>
heh
23:17
<sicking>
hence i vote for File in localStorage for now :)
23:17
<Hixie>
well the dialect would be its own spec i think
23:17
<sicking>
and across postMessage
23:17
<Hixie>
i don't know what file in localStorage would really be
23:17
<Hixie>
postMessage should definitely support it
23:18
<sicking>
Hixie, what do you mean, what it would be?
23:18
<gsnedders>
Oh dear…
23:18
<Hixie>
sicking: well localStorage up to now has only been unstructured name-value string pairs
23:18
<sicking>
Hixie, can't we just make the value <string-or-file>?
23:18
<gsnedders>
If you follow the rules for comparing charsets in UTS #22, you end up with iso-ir-9-1 and iso-ir-91 being identical, but they shouldn't be.
23:19
<Hixie>
sicking: i don't think gmail wants an unstructured answer here; and extending DOMString to DOMString-or-File seems weird when you don't allow anything else
23:19
<sicking>
nono, not touch DOMString
23:19
<Hixie>
gsnedders: do browsers support those?
23:19
<sicking>
that would be, um, crazy
23:19
<gsnedders>
Hixie: No idea.
23:19
<Hixie>
sicking: that's what i'm saying :-P
23:19
<Hixie>
sicking: anyway, send mail
23:19
<Hixie>
sicking: i'm swamped right now
23:19
<sicking>
ok
23:20
<Hixie>
sicking: and waiting for arun to finish his draft before i think of this
23:20
<gsnedders>
Hixie: On an HTML 5 note, it'd be nice if the charset override table used the same case as the IANA registry (i.e., lowercase "w" for "windows-*")
23:20
<sicking>
ok, i'll try to spend some time with him touching up the current draft
23:21
<sicking>
it needs a lot of editing, but I'm not sure that the actual APIs will be much different from what we have now
23:21
<sicking>
apart from Annes bikeshedding ;)
23:21
<Hixie>
gsnedders: send mail
23:21
<sicking>
and possibly FileDialog since you're all a bunch of pansies ;)
23:21
gsnedders
was expecting that answer :)
23:21
<othermaciej>
Hixie: the reasons cited for using Flash to upload sound just like what the MobileMe people told us
23:22
<othermaciej>
which is why we added multi-file selection and file upload progress (via XHR) to Safari
23:22
<annevk5>
sicking, arun already accepted all my comments :)
23:22
<annevk5>
sicking, well, apart from FileDialog, but it seems I'm not alone
23:23
<sicking>
annevk5, see above re 'pansy' ;)
23:24
<annevk5>
sicking, says the man who doesn't wanna do redirects :p
23:25
<sicking>
annevk5, redirects are the root of all evil
23:25
<sicking>
annevk5, together with money
23:25
<sicking>
annevk5, and premature optimizations
23:25
<Hixie>
scripting modal dialogs are the root of more evil than redirects :-P
23:26
<sicking>
there's going to be a dialog no matter what (in the non-dnd case)
23:26
<Hixie>
a user-invoked one is not a big deal
23:26
<sicking>
The difference between having it triggered by an arbitrary click, and on a JS call, is very small
23:26
<sicking>
Hixie, UAs are free to warn instead of putting up a dialog if the JS-call happens outside a click
23:27
<sicking>
Hixie, a'la popup blockers
23:28
<Hixie>
sicking: any feature that is admitted to need popup-blocker-like treatment _at the design stage_ is imho a non-starter
23:35
<sicking>
Hixie, i don't think it needs a popup blocker
23:35
<sicking>
Hixie, IE has had this API for ages. Granted, that's not the best indicator ever
23:36
<sicking>
Hixie, also, geoid is doing the same thing
23:39
<Hixie>
IE doing something is not a good indicator that it is secure.
23:39
<sicking>
i know, but i've never heard or people being annoyed by it
23:39
<Hixie>
and the geoloc stuff isn't a modal dialog, it pops up an asynchronous UI that can then later cause the script to get a callback
23:39
<sicking>
or attacked by it
23:39
<Hixie>
anyway
23:39
<sicking>
FileDialog is async too
23:39
<Hixie>
we don't have a need for this, and people aren't really happy about it
23:40
<Hixie>
so why do it
23:54
<Hixie>
hsivonen, others: i'm writing a section for the parser spec talking about some error handling cases and basically explaining how they end up as they do
23:54
<Hixie>
if you have any specific error cases you would like me to discuss in particular, let mek now