Gorry,
On 3 Jul 2006, at 11:54, Gorry Fairhurst wrote:
Thanks, we should definitely discuss this in the meeting next week.
Some comments from a read of the new rev of this draft are below,
...
* In Section 4.2, you note that the action depends on the actual
time an
RTCP packet is sent, how do you envisage the API handling this?
I was envisaging a blocking send, with the RTCP implementation noting
the time at which the send call returned as the actual time the
packet was sent. An API which queues packets and paces them out
according to the congestion control might have implementation
advantages -- the kernel could probably space packets more accurately
than a user process -- but has problems with the user process not
knowing how much data has been sent, or when.
* In Section 4.5, "The only potential conflict occurs when" -- is this
really so, or would it better to say the only "known"?
It might be better to say known, although of the possible profile-
specific modifications listed in RFC 3550, this is the only one that
conflicts with DCCP.
* In Section 4.5, "changes the RTCP reporting interval or RTP
reporting
rules" - I don't understand what exactly this means.
The RTP specification allows senders considerable flexibility in when
they send RTP data packets, provided the result is not TCP
unfriendly. It also defines the RTCP reporting interval such that
reception quality reports are sent at low rate in the background (an
average of once per 5 seconds, in the typical case). An RTP profile
can change these rules, for example by increasing the RTCP feedback
rate, or mandating that RTP data packets are scheduled according to
some algorithm.
* At the end of section 5, you seem to say don't use TFRC at the
RTP level, which seems fine. Would it be good to point here to the
relevant CCIDs that implement TFRC within DCCP?
Yes, will do (at the end of section 4.5).
* The draft mentions DCCP partial checksum coverage, late it also
discusses SRTP - are these orthogonal?
Probably not. Let me think about this some more.
* I wonder what the relationship of this should finally be to the
TFRC media guide and the DCCP user guide - should these both be
cited as Refs?
Probably as informative references, yes.
Thanks,
Colin
