Feb 28 15:13:04 <jmspeex> pwk: you really want to spawn a new jitter buffer for each and every talk event?
Feb 28 15:14:27 <pwk> spawning a new jitter-buffer for each and every talk event has worked fine so far, do you advise against it? There are an awful lot of potential talkers so that keeping jitter-buffers open for every
client seems not to be an option
Feb 28 15:16:27 <pwk> what I was thinking of (but have not implemented yet) was to keep jitter-buffers alive for a certain "cool-down" period (e.g. 30 seconds) in order not to loose the adaptation in case the client conti
nue talking "soon"
Feb 28 15:16:37 <jmspeex> Well, I guess it depends on how often any given speaker starts talking.
Feb 28 15:16:55 <jmspeex> pwk: yes, that would definitely be good.
Feb 28 15:17:39 <jmspeex> After a while, it's not much use because the timebase will drift away -- unless clients send keepalive packets to keep the sync.
Feb 28 15:18:12 <pwk> no keepalive packets, when a client is not talking he does not transmit anything.