Residential line emulation (RLE) is tough. (a.k.a. POTS 
emulation or keyed system)  There have been some suggestions
to have SIP full mesh signaling, but audio which does
otherwise.  I would like to discuss this.

Forgive me, but how does the audio gets plumbed if
the SIP signaling is full mesh?  I am interested in
the audio using unicast.  I don't see how the clients
agree on who mixes.  

Actually, maybe I do know.  Some "smart" call controller 
sends out a bunch of REFER messages, which the phones hide 
from the user.  This, I will argue, is worse than B2BUA.
I have many issues. 

Agreed, that SIP full mesh does appear to spread
the state, such that if any participant
dies, the others may have enough knowledge to recover.
This is at great expense of UA complexity.
I question whether implementations will be strong
enough to handle it.  The unsuspecting far end
may not be capable.  More importantly, the fault
condition may not be noticed because senders
can arbitrarily stop sending audio.  Recovery
detection would be tough to write and harder
to test.  Ultimately, I expect the smart controller
to have essential state.

Is there some rule of thumb I am missing? ...like, "All
two-party call legs utilizing the same call-id should be
presented as the same line light to a PSTN acquainted
human."   I am NOT yet ready to accept this.

Would someone please clue me in?  Does anyone care
to discuss this?  Maybe we are not yet ready for this
can of worms.

Sadly, I do not yet have an elegant solution
to this problem.  Using a B2BUA doesn't yet
solve the problem of shared line light presentation.
My current drafts rely on phonectl:  A B2BUA device 
(which maybe inside one of the phones, the primary phone) 
owns the call state.  It manages the audio. 
One or more secondary phones is a phonectl master and 
presents the line statuses as his own to the human.  

The phonectl syntax would need modification.  Specifically
the primary phone would need to support 
multiple call-legs/media-streams per line light, and
the phonectl syntax reflect this.  I am still undecided 
whether phonectl sent down an established call-leg should refer
to the call-leg as a device or a line.

Pardon the alpha nomenclature (argot? :-).

Cheers,
Rick
3Com

_______________________________________________
Sip-implementors mailing list
[EMAIL PROTECTED]
http://lists.cs.columbia.edu/mailman/listinfo/sip-implementors

Reply via email to