[dvblast-devel] 6 patches for DVBlast

[Marian Ďurkovič]

On Mon, 30 Mar 2015 00:02:07 +0200, Christophe Massiot wrote
> > 3: This one is really strange, as the RTP timestamp originally calculated
> > may perturb some receivers RTP input buffer management as the counter can be
> > too volatile for some highly variable bandwidth streams. So, it's better to
> > use a local timestamp.
> 
> Although I am personally in favour of using a local timestamp for RTP 
> timestamp (and I have done so in Upipe), this is in contradiction with 
> RFC 2250, section 2, "Each RTP packet will contain a timestamp derived 
> from the sender's 90KHz clock reference.  This clock is synchronized 
> to the system stream Program Clock Reference (PCR) or System Clock 
> Reference (SCR) and represents the target transmission time of the 
> first byte of the packet payload."
> 
> However DVB-IPI states that it is not mandatory but "recommended". So 
> does anyone object to changing the behaviour of DVBlast ?

I definitely support changing it to local timestamps. This way each packet
contains the exact timestamp when it was sent to network, so any variances at
receiving site are clearly contributable to network jitter.

When the timestamp is set to PCR, it no longer corresponds to exact packet
sending time - and with VBR streams or low bandwidth streams (radio stations)
it could be considerably different from the real packet sending time.


   With kind regards,

       M.