On Tue, 08 Mar 2011 19:47:47 +0100, Arun Ranganathan <a...@mozilla.com>
wrote:
On 1/31/11 11:04 AM, Simon Pieters wrote:
Hi,
Why is FileList defined as
typedef sequence<File> FileList;
rather than as an interface similar to NodeList? It seems that in Gecko
and WebKit FileList is more like NodeList than as a sequence.
(In the process of going through feedback to prepare FileAPI for Last
Call Status, I realize I overlooked this message. My apologies for
that.)
So this goes back to the earliest versions of the specification, when
WebIDL itself was in flux. Sequence parametrized types were considered
a superior choice owing to having a lot of the hard work of defining
these kinds of datatypes "in-built", and I was urged by Hixie and others
to adopt it for FileList. I *originally* had NodeList (or something
else akin to the classic array model -- can't remember now :) ).
Do you disagree that FileList should be a sequence parametrized type?
Would you rather I revert back to earlier WebIDL constructs? I'm happy
to do so, *especially* if these stand as LC blockers for any
implementer, but I personally don't feel a lot of reason to change it,
since I think WebIDL accounts for the FileList use case. Do you
disagree strongly?
Other tasks remain before LC, obviously -- notably, an event queue
overhaul of the spec. to make it closer to what HTML5 does -- but I've
continued to update the Editor's Draft:
http://dev.w3.org/2006/webapi/FileAPI/
Personally I don't think one is more superior than the other. However, I
will note that Opera has implemented (but not shipped yet) NodeList-like
FileList. That makes three browsers that implement NodeList-like, and none
that follow the spec. It seems easier to change the spec than to change
the implementations.
--
Simon Pieters
Opera Software