14 Jan
2014
14 Jan
'14
2:43 p.m.
On 14.01.2014 12:54, Holger Hans Peter Freyther wrote:
On Tue, Jan 14, 2014 at 09:41:39AM +0100, Jacob
Erlbeck wrote:
Ok, I'm convinced that my solution makes probably no difference, since
keep-alive has only worked with the help of the BTS packets so far
anyway. And your approach is indeed simpler and doesn't change the
behaviour compared to the old inplementation.
Jacob
If (and only if) we can rely on incoming packets
from any type of BTS
when SEND_RECV is used, we could optimize these few packets away.
Otherwise I'd rather check, whether packets have been sent to the net
since the last timeout and only send a keep alive packet if there were none.
I disagree. Keep things simple and look at the connection state. if (and only
if) we start to look into DTX we can consider adding a heuristic that checks
the time of the last outgoing RTP packet.