Hello, On 11/2/06, Kite Flyer <[EMAIL PROTECTED]> wrote: > I need to write a one-way conference server where always a single user > speaks and all remaining users listen. > > Assuming that all conference users use the same codec, I want to improve > performance of flowgraph by removing decode/ encode overhead. If I remove > MprDecode, MprEncode resources from MpConnection and connect MprFromNet and > MprToNets directly to MprBridge resource, will the bridge resource be able > to replicate raw RTP frames? This is possible, but I think it will take a bit more changes. E.g. you should remove RTP packetizer. But current RTP packetizer produce buffers in different format from RTP reader. This mean you could not just short-circuit RTP reader with network writer. I believe the right way is converting RTP packing and network writer code to use same buffer format (MpRtpBuf) as RTP reader does.
Then you may want to totally disable mixing code in MprBridge. It may be done depending on incoming packets type (MP_BUF_RTP or MP_BUF_AUDIO) or depending on value passed with construcor (some bool doMixing or bool oneWayOnly param). This work may be useful to other users and good to sipXtapi itself. So I hope you'll contribute your code back to SIPfoundy. :) Regards, Alexander Chemeris _______________________________________________ sipxtapi-dev mailing list [email protected] List Archive: http://list.sipfoundry.org/archive/sipxtapi-dev/
