Re: [rtcweb] Use cases - recording and voicemail

[Harald Alvestrand <[e-mail]>]

On 08/20/11 01:52, Paul Kyzivat wrote:
> Randell,
>
> Many of the issues you are bringing up are things we have been 
> addressing in the siprec WG. RTCWEB can either reinvent all that, or 
> we can find some way to reuse that work. At the moment, siprec assumes 
> that the agent that is initiating the recording is in the signaling 
> path of a sip call, and that it is capable of establishing another sip 
> session to the recorder. Those assumptions would seem to be wrong for 
> RTCWEB.
Hm.

RTCWEB doesn't seem to going in the direction of mandating SIP, but 
still, I would think that it is reasonable to say something along the 
lines of "the remote recording case is handled by connecting to a 
SIPREC-capable recorder" (with the usual degree of gatewaying help from 
our signaling proxies).

That will, of course, require that the SIPREC recorder is capable of 
participiating in an RTCWEB session (that is, support ICE and the 
mandatory codecs), that the RTCWEB implementation be capable of copying 
incoming media streams to an outgoing interface, and that negotiation 
can down-negotiate to something that is supported by both call 
participants and the recording device. Does SIPREC envision establishing 
minimum requirements for codecs and profiles?

Once we can satisfy ourselves that we have all the pieces required to 
send media off to a remote API, we should see if we can do something 
very similar for sending media off to some kind of local recorder; it 
seems less likely that we'll get into trouble with locking ourselves 
into a wrong model if we do things in that order.

My $0.02.

               Harald