Re: [AVTCORE] RTP/TFRC + FEC ?

[Olivier Crête <[e-mail]>]

On Sat, 2011-05-07 at 11:47 -0700, Ali C. Begen (abegen) wrote:
> > I've implemented the RTP/TFRC draft and it seems to work pretty well.
> > That said, it also introduces some level of packet losses. Also, with
> > some wifi connections, we have a relatively high rate of non-congestion
> > packet loss. For both of these cases, I'd like to implement some kind of
> > FEC, since losing a single packet tends to give nasty artifacts with
> > video (especially using the free video codecs).
> > 
> > It seems the current tendency in recent RFCs related to FEC is to send
> > the FEC data as a separate stream from the main data. This seems to go
> 
> That has a lot good features. Please see the drafts in fecframe wg.
> The fec framework draft will also tell you that both the source and
> fec packets can be sent in their own RTP streams (with different
> SSRCs) but within the same UDP flow, if that is what you are worried
> about.

I saw vague allusions to using different SSRCs, but I couldn't find
anything to signal that using SIP/SDP ?


> > against the intent of RTP/TFRC since using a second stream may
> > effectively double the bandwidth share used by the video. Mixing both
> 
> Huh? It does not matter at all for the overhead whether you send the
> fec packets separately or in-band. Also, FEC does not need to double
> the bandwidth.

It's not about overhead, it's about the fact that it will be two
streams, each with its own TFRC state. And with TFRC I would assume that
two streams would get roughly twice the bandwidth of one stream (and I
guess that's not appropriate).

-- 
Olivier Crête
[e-mail]
Collabora Ltd