The draft rtp specification for text conversation that is discussed in the IETF AVT group is now available in the following URL.
www.omnitor.se/english/standards/rtp/draft-hellstrom-avt-rtp-text-01.txt
It is referenced in the H.323 Annex G Text Conversation draft that is entered as a contribution to the ITU-T Q13/16 meeting next week as document APC1561. I would like to receive comments on this draft. I am aware of the following points that might need discussion.
1. Timestamp clock frequency. I had put a low figure 10 Hz in the specification. It was questioned and proposed that a figure of some thousands should be selected. I am willing to accept that, and propose that 1000 Hz is used. A millisecond per tick is a handy figure. This will however lead to the timestamp offset in the redundancy data headers that use only 14 bits, flip around in only 16 seconds. If correct timing information is wanted, the data flush procedure described in next section should be used.
2. Retransmission of text before a pause. Text does not flow continously, but only when a user enter text. Therefore, a recommended procedure could specify that for the redundancy scheme, it is advisable to send the ISO 10646 synch character FEFF after some inactivity timer to flush out redundant data - see section 3.4.
3. Recommended procedures section. In the draft that accompanies this mail, section 3 documents the recommended procedures for using the payload format. Does the group want it to be kept there or shall it be moved to the specification specifying use of this RTP Payload? ( The first user being H.323 Annex G)
Any other comments are welcome.
Gunnar Hellström
----------------------------------------------- Gunnar Hellstrom LM Ericsson
E-mail gunnar.hellstrom@omnitor.se Tel +46 751 100 501 fax +46 8 556 002 06