RE: [rohc] Video and header compression

[Rolf Hakenberg <hakenberg@panasonic.de>]

Hi all,

I agree with Khiem's statement below. If any additional information (e.g. TS
jump) needs to be transmitted to the decompressor a FO packet should be
used.

As suggested by Krister in his mail before, we should work on the
requirements of the FO packets and start designing these. I think one
requirement should be, that FO should be able to be used  for
updating/changing the context and for just sending information for one
packet (non-updating). With this means we could solve many of the problems
of some of existing AND future payload formats for RTP very efficiently
without going through the process of defining specific RoHC profiles.

Best Regards,
Rolf


> -----Ursprüngliche Nachricht-----
> Von: owner-robhc@sm.luth.se [mailto:owner-robhc@sm.luth.se]Im Auftrag
> von khiem.le@nokia.com
> Gesendet am: Mittwoch, 9. August 2000 21:35
> An: zhigang.liu@nokia.com; krister.svanbro@lu.erisoft.se;
> anton.martensson@era.ericsson.se; rohc@cdt.luth.se
> Betreff: RE: [rohc] Video and header compression
>
> Hi,
>
> The IR/FH, FO and SO header classification is very general, and does not
> depend on whether the media is audio or video. It would be a mistake to
> attempt to identify/define new kinds of headers specifically because of
> video. If we take that path, God knows what other kinds of headers we will
> have to introduce in the future, as new media are identified.
>
> SO is simply defined as a header where the decompression can be
> done just by
> knowing the RTP sequence number. Cases which require more than
> the sequence
> number information to decompress are either FO or IR/FH. It's that simple.
>
> Khiem
>
> -----Original Message-----
> From: EXT zhigang.liu@nokia.com [mailto:zhigang.liu@nokia.com]
> Sent: Wednesday, August 09, 2000 10:14 AM
> To: krister.svanbro@lu.erisoft.se; anton.martensson@era.ericsson.se;
> rohc@cdt.luth.se
> Subject: RE: [rohc] Video and header compression
>
>
> > the primary video application target for ROHC should be one where
> > there are more than one RTP packet per video frame.
>
> I think there is no doubt about it.
>
> > ... The timestamp and the Marker will change often, which the SO cannot
> > handle... Perhaps we can get this from the FO format...
>
> As I said in my previous email, using FO is the way to go. It's
> conceptually
>
> consistent (with audio) and straightforward in implementation.
>
> Just consider the FO format for audio: it needs to have SN field, M-bit
> field, TS and/or IP-ID field. So, what are the additional things needed
> for video? None. The major differences between audio and video are NOT
> in the header format. They are TS_STRIDE (e.g. 3000 for video instead
> of 240 or 160 in the case of audio) and the frequency of FO headers
> (90%, or even 100%?). Please see my comments on them in my previous
> email.
>
> Zhigang
>
> > Cheers,
> >
> > / Krister
> ---
> Mailing list for Robust Header Compression WG
> Archive: http://www.cdt.luth.se/rohc/
> ---
> Mailing list for Robust Header Compression WG
> Archive: http://www.cdt.luth.se/rohc/
>