12:03
<Ms2ger>
<!-- next section refers to previous section as "previous section" -->
12:06
<Velmont>
Glad we got that cleared.
12:12
<Ms2ger>
Feel like reviewing some tests?
13:14
<MikeSmith>
heycam|away: anything's possible
20:07
<Hixie>
heycam|away: see https://www.w3.org/Bugs/Public/show_bug.cgi?id=20496
20:09
<Stevef>
35,730 home pages (from top 50,000 sites) gathered yesterday, in 1 large zip file get it from http://www.webdevdata.org
20:21
<Stevef>
me/ useful for revealing cowpaths and undermining arguments from authority
20:22
Stevef
useful for revealing cowpaths and undermining arguments from authority
20:25
<tantek>
SteveF - I prefer to think of evidence as outweighing (rather than undermining) arguments from authority. :)
20:57
<Stevef>
tantek: yeah right, I used 'refuting' here https://twitter.com/stevefaulkner/status/285489160985382914 which I think is more fitting
21:00
Stevef
thinks it would be great if a web company that has easy access to the sort of raw data that I have been collecting would make it available on a regular basis...
21:02
<tantek>
ooh a CG too. (clicks Join)
21:52
<annevk>
Hixie: if you want to treat Nan/Infinity different it seems you might need to do that in prose
21:53
<annevk>
Hixie: unless we're going to introduce Nan and Infinity as values somehow so you can do (NaN or double)...
22:19
<annevk>
Hixie: the video player thing, do you mean the way it seeks? It was annoying me to no end yesterday trying to find those bugs you referred to...
22:35
<Hixie>
annevk: that has to be prose, but my point is just that the return value is always "double", not "unrestricted double" which is what the setter value has to be
22:35
<Hixie>
annevk: yes, the way it seeks in particular
22:36
<annevk>
Hixie: if the setter is more generic, it doesn't really matter
22:36
<Hixie>
we could make all the attributes "any", too :-P
22:36
<annevk>
Hixie: for readonly attributes, that's basically true :)
22:36
<annevk>
(unless I'm missing something)
22:50
<annevk>
arv_: re es-discuss: does this mean the ArrayClass experiment has failed?
22:50
<annevk>
arv_: should we pursue something else instead? e.g. Hixie's toArray() suggestion?
23:31
<annevk>
heycam: maybe we should have getattribute {type} {name}; and setattribute {type} {name}; for cases where they behave different? although that still does not help with PutForwards I guess...
23:31
<heycam>
annevk, maybe, I guess I'm not that fussed about the syntax particularly
23:32
<heycam>
annevk, I see the feature as "the attribute actually has type X, but additional some values of other types are accepted when assigning and get converted to the appropriate value of type X, or handled in some other way"
23:32
<heycam>
not really sure why it should be that controversial
23:33
<annevk>
I think it's mostly smaug____ that considers it controversial
23:34
<annevk>
though for everything but PutForwards one can just loosen the type as the prose already determines what is returned
23:34
<smaug____>
annevk: bz and IIRC also sicking have been agains PutForwards
23:34
<smaug____>
against
23:36
<gsnedders>
annevk_: the ArrayClass experiment?
23:37
<annevk_>
heycam: btw https://mail.mozilla.org/pipermail/es-discuss/2012-December/027688.html (unclear per logs whether this got true)
23:37
<annevk_>
s/true/through/
23:37
<heycam>
smaug____, I think asymmetric get/set types for attributes is dissimilar enough from PutForwards to consider it on its own merits
23:38
<heycam>
annevk_, oh interesting, will read
23:38
<smaug____>
I'm not sure I understand its merits
23:38
<heycam>
I think it just makes it clearer in the IDL that certain values will not be returned when getting the attribute
23:38
<heycam>
but it also gets around restrictions where we say e.g. sequences can't be the type of an attribute
23:38
<smaug____>
it feels a bit odd if setting an attribute needs to run ctor of something taking the value as param
23:39
<heycam>
you could have it be allowed to be assigned to the attribute, as long as you return something different when getting
23:39
<smaug____>
it is the creation-of-new-objects which is odd
23:39
<heycam>
I see
23:40
<heycam>
so you'd rather have the author do an explicit `new FileList(sequenceOfBlobs)`?
23:40
<annevk_>
values get converted all the time
23:41
<annevk_>
e.g. 1 -> "1"
23:41
<smaug____>
that is very different
23:41
<annevk_>
I don't see why
23:41
<heycam>
the restriction on sequence types for attributes is the only thing stopping writing `attribute (FileList? | sequence<blob>) files;`
23:41
<heycam>
and then having prose only ever return a value of type FileList
23:42
<smaug____>
annevk: you certainly aren't creating new objects that way
23:42
<smaug____>
is there any case when a new object is created automatically
23:43
<annevk>
I don't really see value in differentiating primitives and objects
23:43
<gsnedders>
smaug____: "foo".indexOf("a") creates an object, if that counts
23:43
<annevk>
you'd want the same kind of behavior to apply to both
23:44
<heycam>
I think anne's point of 1 -> "1" is that this is the more general issue -- assigning one value, and getting a different value back if you get the attribute later
23:44
<smaug____>
heycam: also, are there other use cases for this conversion?
23:45
<heycam>
Hixie pointed out https://www.w3.org/Bugs/Public/show_bug.cgi?id=20496 earlier as another case that could use it
23:45
<heycam>
so you want to have some particular processing of Infinity/NaN when assigning, but you never want to return those values
23:45
<heycam>
you could just have `unrestricted double` as the type and in prose never return non-finite values
23:46
<heycam>
but it might be easier to understand for readers of the spec if it is in the idl
23:48
<annevk>
gsnedders: the email is about that, yes, and about how maybe we should introduce something like Array, but readonly, to solve it
23:52
<heycam>
do we actually have any interfaces defined with indexed property setters?
23:53
<heycam>
or are all our array-like interfaces read only
23:53
<heycam>
?
23:54
<smaug____>
storage stuff
23:55
<smaug____>
er, hmm
23:55
<smaug____>
no
23:55
<smaug____>
it is key-value