[Webkit-unassigned] [Bug 68460] Add WebCore platform interfaces needed by updated MediaStream and PeerConnection design

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=68460





--- Comment #7 from Adam Bergkvist <adam.bergkvist at ericsson.com>  2011-09-21 07:10:41 PST ---
(In reply to comment #3)
> (From update of attachment 108042 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=108042&action=review
> 
> > Source/WebCore/platform/mediastream/PeerHandler.h:46
> > +struct PeerHandlerConfiguration : public RefCounted<PeerHandlerConfiguration> {
> 
> Shouldn't this class be called PeerConnectionConfiguration?
>

Perhaps PeerConnectionConfiguration would better explain that it's created from the PeerConnection serverConfiguration argument, but it's the object that configures the PeerHandler and it's never used by PeerConnection. It's also stored as a member on PeerHandler.

> >> Source/WebCore/platform/mediastream/PeerHandler.h:51
> >> +        TypeTURN
> > 
> > I'm not quite clear which concepts are going to be in WebCore and which concepts are only in the underlying library.  I was hoping that STUN would be a library-internal concept.  Is there some reason WebCore needs to be aware of STUN?
> 
> I agree with you here, but "unfortunately" the spec specifies exactly how the configuration string should look like. The alternative, which I prefer, is to just pass the string to the platform specific code.

The serverConfiguration argument string must be specified in order to achieve interoperability between implementations. I think the parsing should be done in the shared WebCore code to make the behavior consistent between ports (and to avoid duplicate code). The PeerHandlerConfiguration struct could easily be serialized back to a string if it's easier to pass it to chromium that way.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.