Re: [MMUSIC] Offer/Answer PT Questions - text proposal

[Paul Kyzivat <[e-mail]>]

On 2/26/16 12:52 PM, Roman Shpount wrote:
> On Fri, Feb 26, 2016 at 12:20 PM, Paul Kyzivat <[e-mail]
> <mailto:[e-mail]>> wrote:
>
>
>     The session remains, and so the PT bindings remain. But the O/A
>     rules are negotiating which of those are to be used going forward.
>     If the new offer no longer mentions some PTs that were previously
>     bound, then it is saying that it doesn't want to receive them any more.
>
>     It is becoming clear to me that it will be necessary to define PT
>     bindings - what they are, when they are established, when they are
>     forgotten, and the scope within which they live. Each participant
>     will need to know these bindings, and they must agree on them,
>     except for well defined signaling delays. I guess this is a lot like
>     an RTP session, except that it is tied to O/A signaling.
>
>
> One other thing to keep in mind is that RTP session creation and tear
> down is not time synchronized with O/A. End point might still be
> receiving media packets for the previous RTP session after O/A
> completes. If O/A happen often enough, several O/A exchanges can
> complete while old RTP sessions established by old O/A exchanges are
> still running. PT should uniquely identify codec profile for all of
> them. In case of SIP, O/A exchanges typically take more time then RTP
> session setup/tear down, so keeping track of the last two O/A exchanges
> is typically enough.

Yes, I understand that. I don't see how to solve it. It presents a hard 
problem. The new session may include a new participant that doesn't know 
the old bindings. If we require it to know them then that will require 
us to write more rules, and potentially break some existing usages.

	Thanks,
	Paul