Ross is against using RTP Payload 14 at all though,
and with good reason.  An mp3 frame is not an ADU
because of the bit reservoir, so losing one packet
means you could lose a few frames.  And that is still
assuming one frame per packet.  So unless you are
interleaving, which also is non-trivial in Payload 14,
relatively minor packet loss will make the stream
sound terrible.

We turn off the bit reservoir when we make mp3's for our
stream, enforce one frame per packet, and hope to do
some sort of staggering/interleaving soon.

But Ross has also come up with a nice solution, see

He also has LGPL code to implement this, anybody
want to hack this into obs?

> Apologies for not responding to this sooner. My freeamp-dev mail was
> going into the wrong folder and I never spotted it.
> On 26 Feb, Dave Chapman wrote:
> > On Monday 26 February 2001 02:46, Marty Schoch wrote:
> >> A more complicated solution would be to modify Obseqium to set
> >> the Market bit in the RTP header when it's about to change format.
> >> Then just have the player reconfigure the decoder when it
> >> detects the Market bit in the stream.
> That's what I wanted to do in the first place, but then Ross Finlayson
> advised me against it. I forget what his reasoning was. The obs side of
> things is not hard -- its the FreeAmp side that is harder to solve.
> > Whenever this bit is set, Freeamp should assume that the audio type (layer,
> > bitrate, whatever) has changed, and reconfigure itself appropriately.  It
> > could also reset it's own time display to zero (if people want that - I do).
> >
> > Do people agree?
> Agreed. I'll see what I can do.
> --ruaok         Freezerburn! All else is only icing. -- Soul Coughing
> Robert Kaye   --    [EMAIL PROTECTED]   --
> _______________________________________________

Marty Schoch
Join the I-Force
Download the MCT Player

Reply via email to