[AVT] Potential overhead issues for RTP over DCCP

All,
	At present there is some discussion on the Datagram Conegestion Control
Protocol mailing list (DCCP, also known as DCP, see references below) with
regards to the use of DCCP as an underlying transport for RTP.  A central
concern is that although one of the objectives of DCCP is to insure minimal
overhead in terms of packet size and end node processing time, using RTP
over DCCP will result in about 4 bytes of additional overhead per packet
for redundant representation of a sequence number.  This overhead is in
addition to the potentially redundant information transmitted in RTCP RR's,
which may be the same as that piggybacked on DCCP receiver reports.
Redundant data in this case includes metrics for estimating the RTT, and a
receiver loss metric.  
	The tradeoff currently being considered is the 4 byte per packet overhead
for a simplified API in which sequence numbers need not be represented to
the application layer by DCCP on a per packet basis.  In my opinion, the
RTCP RR can be easily resolved when we assume that the problem domain of
real-time metric reporting, and the use of these metrics by an RTP
application is rather different  that the use of reciever reports by DCCP.
When this is taken into account, the consolidation of potentially redundant
data in this case may not be appropriate. 
	What are the group's thoughts on these issues?

thanks,
damon

DCCP (DCP) references:

http://www.icir.org/kohler/dcp/draft-floyd-dcp-problem-00b.txt
http://www.icir.org/kohler/dcp/draft-kohler-dcp-03.txt
http://www.icir.org/kohler/dcp/draft-padhye-dcp-ccid3-03.txt

_______________________________________________
Audio/Video Transport Working Group
avt@ietf.org
https://www1.ietf.org/mailman/listinfo/avt