AW: [vorbis-dev] Vorbis RTP Internet Draft

[Phil Kerr]

On Wednesday, Jan 8, 2003, at 22:32 Europe/London, Tor-Einar Jarnbjo 
wrote:

> If the codebook list (or the identification header) changes during
> transmission and it is sent to the client using RTPC, will the client 
> then
> be able to tell exactly which audio packet is the first to be using 
> the new
> information when being decoded? You write "just prior to the Vorbis
> bitstream change", but as I see it, it would be impossible for the 
> server to
> be sure that the RTP and RTPC packets are received in the correct 
> order by
> the client.
>
> One solution would be to send codebook and id header changes in-stream 
> (a
> disadvantage about this would be that the client may fail to receive 
> the
> data completely), another solution would be to add a field to the RTPC
> packets with the sequence number of the first RTP packet for which the 
> new
> data should be valid. The last solution would also allow the new data 
> to be
> sent some time before the actual change is taken place, allowing the 
> client
> to request a resend before the changes take effect.

I broadly agree with a seq id marker solution, and it would be quite 
easy to implement.   If the client fails to receive the new set of 
codebooks the periodic codebook retransmission should bring the client 
back in step (but will your speaker cones be embedded in the opposite 
wall by then?).

There is, however, the possibility of a much better solution - TCP over 
RTP.  This issue was discussed again on the AVT list last month and 
there should be a I-D to cover this functionality soon (I flagged this 
as a potential delivery vector in a previous email to this list on 
16/12/02).  As there is nothing defined yet I could not specify this as 
the transport mechanism for the codebooks but I think it's worth 
pushing for this once the Vorbis RTP I-D clears the AVT submission 
queue.

This will mean more complexity for the server, and client, as they will 
have to handle UDP and TCP, but the increase in codebook transport 
reliability certainly makes up for this.

But just in case this new draft doesn't appear soon we should still 
look for a workable solution without it.

Phil

>
> Tor
>
> --- >8 ----
> List archives:  http://www.xiph.org/archives/
> Ogg project homepage: http://www.xiph.org/ogg/
> To unsubscribe from this list, send a message to 
> 'vorbis-dev-request at xiph.org'
> containing only the word 'unsubscribe' in the body.  No subject is 
> needed.
> Unsubscribe messages sent to the list will be ignored/filtered.
>

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.