[M4IF Technotes] Use Rtp to transport Mpeg4 video

[Dave Singer]

At 00:23 +0200 7/4/03, Frederic Bouilhaguet wrote:
>Dear Dave,
>
>I would like to know what you think of the hinters that produce Hint Tracks
>that contain 'stts' and 'ctts' tables that are equal to the 'stts' and
>'ctts' of the media video track...
>Shouldn't these Hint Track contain *only* one 'stts' table with delta time
>stamp values inside it that guarantee RTPTimeStamp = CTS WITHOUT any 'ctts'
>table ?

Yes.  Somewhere in the standard it says explicitly that hint tracks 
are not 'composed' and have no 'ctts' table.  A server shouldn't have 
to do composition re-ordering;  it doesn't make sense.  The hinter 
should have computed the right time-stamps for the (e.g.) RTP packets 
in a hint track.
>
>Frederic
>---
>Resonate MP4
>Tél: +33 1 42 41 16 65  - Mobile: +33 6 63 40 99 12
>
>>  -----Message d'origine-----
>>  De : technotes-admin   lists.m4if.org
>>  [mailto:technotes-admin   lists.m4if.org]De la part de Dave Singer
>>  Envoyé : jeudi 3 juillet 2003 18:31
>>  À : technotes   lists.m4if.org
>>  Objet : Re: [M4IF Technotes] Use Rtp to transport Mpeg4 video
>>
>>
>>  Li -- You have it upside-down.  One would only use the (ugly)
>>  in-stream timing of MPEG-4 video when the external system does not.
>>  In this case, in order to achieve sync, and obey the general nature
>>  of RTP, one would believe the RTP time-stamps and ignore what is
>>  in-stream.  The same goes for an MP4 file.
>>
>>  The config information in RTP is normally carried out of band, in
>>  SDP, just as it is in MP4.
>>
>>  In summary, the timestamp of a vop in RTP is determined by the RTP
>>  timestamp;  vop_time_increment, modulo-time-base, and all that stuff,
>>  are irrelevant.
>>
>>
>>
>>  At 07:00 +0200 7/3/03, Hans-Juergen Bardenhagen wrote:
>>  >On Thursday, 03.07.03, 09:14 (received 03.07.03, 06:49) Li wrote:
>>  >
>>  >>  Now I'm trying to transport Mpeg4 video stream over RTP, and coming
>>  >>  against a problem. The RTP says that the configuration information,
>>  >>  such as Visual Object Sequence Header and other headers, should be
>>  >>  bond into a rtp packet. Then How to determine the time stamp of such
>>  >>  a rtp packet? The time stamp of the rtp packet consisting of a vop
>>  >>  can be easily determined by inspecting its vop-time-increment syntax.
>>  >
>>  >As far as I know, the elementary streams in a MP4 container must be
>>  >"hinted" for this, i.e. creating individual time stamp tracks that are
>>  >added to the the container. At least this is how the open source MPEG4IP
>>  >project handles RTP streaming e.g. with Apple's Darwin Streaming Server
>>  >(DSS). See http://sourceforge.net/projects/mpeg4ip/ for more, they also
>>  >have a web forum which you can search through and ask Bill May from
>>  >Cisco if you need further information.
>>  >
>>  >ZZee ya, Hans-Jürgen
>>  >
>>  >_______________________________________________
>>  >Technotes mailing list
>>  >Technotes   lists.m4if.org
>>  >http://lists.m4if.org/mailman/listinfo/technotes
>>
>>
>>  --
>>  David Singer
>>  Apple Computer/QuickTime
>>
>>  _______________________________________________
>>  Technotes mailing list
>>  Technotes   lists.m4if.org
>>  http://lists.m4if.org/mailman/listinfo/technotes

-- 
David Singer
Apple Computer/QuickTime