[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [MMUSIC] ICE, TURN, and QoS interactions



Really forking is the main case, in fact.

So, the guidance I think is that an implementation would set active destination immediately (i.e., prior to the first media packet) when ICE completes if it is having only one peer. If there are more than one, it sets active destination for one of them and then ideally allocates a second turn address and does a reinvite with an ICE restart on the other peer, so that you can then set active destination on that one too.

Nothing in ICE says to do this, and for purposes of interop it doesn't need to. The client can implement anything it likes. Its probably not a bad idea to clarify this in the forking section.

In the Packetcable case is more troubling since you need to know whether this behavior is followed or not in order to compute the QoS parameters for the CMTS. It may be that packetcable specs need to mandate this behavior.

-Jonathan R.

RÃmi Denis-Courmont wrote:
On Friday 07 September 2007 22:26:24 ext Kevin Johns wrote:
While looking at ICE, TURN, and QoS interactions I am struggling with a
scenario that I would like to see if others have any thoughts on.

The driver for this is a desire to ensure necessary overhead for STUN
Data/Send Indications are accounted for in QoS allocations prior to an
active destination being set by the UA.

ICE does not appear to provide any guidance on when the UA should set
the active destination. Section 12.1.1 of ICE talks about interactions
with SIP and requires that no media be sent till a valid candidate for
each component has entered the valid list. If the valid candidate is a
relayed candidate and media begins to flow, should the UA first set the
active destination or send the media wrapped in a STUN Send Indication
and only set the active destination at some later point in the call
flow? I would think you want to set the active destination as soon as
possible to avoid the extra overhead, but was unsure what the
implications of such an approach might be. If the expectation is to
always set the active destination prior to sending media, then the QoS
calculations would not need to account for any additional overhead.

Is there a case where media would begin flowing but the UA would be
unable to set the active destination?

If it keeps changing its "primary" peer.

Otherwise, you mostly have send/data indications while forking.


-- Jonathan D. Rosenberg, Ph.D. 600 Lanidex Plaza Cisco Fellow Parsippany, NJ 07054-2711 Cisco Systems jdrosen at cisco.com FAX: (973) 952-5050 http://www.jdrosen.net PHONE: (973) 952-5000 http://www.cisco.com

_______________________________________________
mmusic mailing list
mmusic at ietf.org
https://www1.ietf.org/mailman/listinfo/mmusic