[FFmpeg-devel] GSoC 2009: FFmpeg is in

Luca Abeni lucabe72
Fri Mar 20 22:02:56 CET 2009


Luca Barbato wrote:
> Luca Abeni wrote:
>> I very much like the idea of having a "solid" rtp/rtsp demuxer that can
>> arrive to replace live555. But if I look at the current rtp demuxer, I
>> am not sure if it must be improved or re-designed. The current RTP code
>> gives the idea that it has been designed with certain goals and assumptions
>> in mind, and over the years people added features removing and or changing
>> some assumptions in subtle ways. As an example, let's look at the dynamic
>> payload handlers... Why should a payload parser mess with the RTP timestamp?
>>
> 
> Because you may have aggregates in packets (N frames in a single rtp 
> packet) and you have
> 
> RTP packet (aggregate) ->> Demuxer ->> many frames

Well, the problem is that the current rtp_parse_packet() function is not 
supposed to return a frame... A parser is used later for splitting the 
stream in frames. So, the payload parser should not really touch the RTP 
timestamp (with the current design... This is something that might be 
redesigned).

>> Only after that the student can start adding new functionalities (but
>> before this, some known bugs have to be fixes - for example RTCP RR
>> packets...).
> 
> We should try to come up with a list and cut the task for the time of 
> the SoC.

I think I posted a list one year ago (or something similar). In the 
meanwhile, I fixed one of the bugs but most of them are still there. 
I'll search for this list and add it to the wiki (I can post it first if 
  you prefer).


			Luca




More information about the ffmpeg-devel mailing list