[whatwg] Peer-to-peer communication, video conferencing, and related topics (3)

[Ian Hickson]

Apologies, I missed these two e-mails in my previous round just now.

On Fri, 25 Mar 2011, Leandro Graciá Gil wrote:
>
> I have two more questions about the Stream API:
> 
> 
>    1. Are GeneratedStream objects supposed to stop when they go out of
>    scope?

Assuming by "stop" you mean "stop generating the underlying stream", then 
insofar as there is any way to test this, no, but insofar as there isn't, 
yes. That is to say, if you have (for example) created a URL for such an 
object before it becomes elligible for garbage collection, and are playing 
that URL using a <video> element, then the stream will keep going even 
after the object has been garbage collected. (The lifetime of these blob: 
URLs is defined by the File API spec.) Similarly, if you have a 
StreamRecorder, then the stream will still be recorded even though the 
Stream object itself will have gone away.

(This is intended to be consistent with the rest of the Web platform.)


>    2. Should StreamRecorder objects keep alive the Stream objects that
>    created them? Is there a use case where this should not be the case?

They don't keep alive the Stream object (how could one tell anyway?), but 
they do keep the stream alive.


On Sat, 26 Mar 2011, Harald Alvestrand wrote:
> 
> My work was started before the revised PeerConnection API came out, and 
> I discussed the proposal with Ian before he published PeerConnection, so 
> I felt that it was fairer to publish both proposals at this time than to 
> bury the fact that I had been working on something else.

Indeed. For the record, alternative proposals are always welcome on this 
list; in the past we have several times thrown away an earlier design in 
favour of a new proposal brought forward in such a manner. The WebSocket 
design (before it went to the hybi working group) is one such example.


> As I wrote in my first message on the thread, I do not favor embedding 
> this API in the HTML5 tome, but am willing to live with that if the 
> community decides that this is the proper path forward.

Note that the PeerConnection API isn't in the HTML5 spec (and will 
presumably never be given the timetable for HTML5, though that's up to the 
W3C who are publishing the HTML5 spec). The WHATWG HTML standard is 
unversioned. (Also, though this is perhaps a little "inside baseball", if 
it wasn't for the interaction with <video>, I'd probably just have this in 
the Web Apps 1.0 standard rather than both the HTML standard and the Web 
Apps 1.0 standard, since we removed the <device> element.)

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'