- From: Boris Zbarsky <bzbarsky@MIT.EDU>
- Date: Tue, 03 Apr 2012 10:12:07 -0400
- To: Glenn Maynard <glenn@zewt.org>
- CC: Cameron McCormack <cam@mcc.id.au>, public webgl <public_webgl@khronos.org>, "public-script-coord@w3.org" <public-script-coord@w3.org>
On 4/2/12 8:42 PM, Glenn Maynard wrote: > On Mon, Apr 2, 2012 at 7:26 PM, Boris Zbarsky <bzbarsky@mit.edu > <mailto:bzbarsky@mit.edu>> wrote: > > I think removing the Float32Array overload will make passing JS > Arrays > of Numbers and Float32Array objects work. > > > But require copies in the process, no? That seems highly > undesirable in this case. > > > If it matters as an optimization, implementations could still create > separate overloads for this as long as it's transparent to the user, > right? There are tons of edge cases here, and more being added by the JS folks all the time, so doing anything other than exactly what the spec says is a good way of introducing bugs... As an example, if the spec says to get the elements from the input thing one by one in order to construct the IDL array, then grabbing a float* from an input typed array would have observably different behavior in situations where the page has defined a property getter for an indexed property on the typed array object in question. Unless the WebIDL indexed access takes priority over user-defined properties, which I think it might for the moment. I'd really rather not have the performance characteristics of the platform depending on such fragile interactions if we can avoid it. :( -Boris
Received on Tuesday, 3 April 2012 15:18:58 UTC