[AVT] Summary changes draft-rey-avt-3gpp-timed-text-01.txt and Timed Text presentation in Minneapolis

Hi all,

this is what I would like to discuss on Monday:

- draft-rey-avt-3gpp-timed-text-01.txt presents a payload format for the
3GPP Timed Text media format. With respect to the -00 version in general
all sections were completed:
* completed empty sections from -00 draft.
* abstract and introduction re-arranged. Moved section "Basics of the
3GP File Structure" to end of the document as Annex B.
* SLEN, SIDX and SDUR lengths fixed to 16, 16 and 24 bits, respectively.
* New OPTIONAL header, SPLDESC, added to transport sample description
in-band.
* Section 4 on payload format expanded: text header, fragment header and
sample description header are fully specified.
* SMIL usage section added.

- Generic vs. Specific transport: in Vienna the issue was raised whether
it should be more generic. After the publication of -01 version, little
feedback has been heard about this; we would like to hear evereyone's
opinion so the work can continue without surprises in the future.

- fragmentation headers: the draft also defines complex fragmentation
headers (FHDR, MTSF) that enable to get the most out of each fragment
even when previous fragments were lost.  What is the opinion of the WG
on this? This may be an overkill as it increases complexity for just a
couple of cases in which they will be used...

- the PT-multiplexing will be abandoned because it is complex and gives
problems, for example in SDP and ROHC. Instead additional fields in
fixed header wil be used.

- the sizes of length, index and duration fields of the text samples
(SLEN, SIDX and SDUR) are currently 16, 16 and 24 bits.  We are not
clear if this is enough/too much.  They allow for text samples sizes up
64KBytes, 65536 different indexes and a duration of samples up to four
hours.  When thinking on reducing these lengths one could argue to
reduce SLEN to 1 byte but 256Bytes is too small . Taking a non-multiple
of 8 is not an option because of parsing complexity, so 16bits is OK.
For SIDX, 8 bits, should be enough I don't think there would be more
than 256 sample descriptions in a text sample. And for SDUR, 16 bits
would mean @1Khz ~1 minute, which is too short for streaming, so we
propose to keep 24bits; also to cope with the case of different
clockrate values. Comments?

- resilient transport: when a packet is fragmented into several pieces,
we think *at least* the fragments containing plain text strings shall be
playable regardless of previous arrived or not (i.e. self-contained).
Making the modifier boxes also resilient  introduces complexity, as seen
with the MTSF headers in the -01 draft. Therefore the tradeoff here
would be to keep/adapt the necessary fields in the fragment header
(FHDR) to display isolated text strings but abandon the MTSF headers for
the modifier boxes. A solution for the latter could be to protect this
through retransmissions, FEC or else.

- the -01 draft does not specify how to reassemble fragments. This will
be specified.

- all other comments on the list noted.

That's all, see you in Minneapolis.

thanks,

Jose

____________________________________
Jose Rey
Mobile Multimedia Team -  MCOM Group
Panasonic European Laboratories GmbH
Email: rey@panasonic.de
Tel:   +49(0)6103-766-134
Fax:   +49(0)6103-766-166
http://www.pel.panasonic.de
____________________________________

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