On Wed, 2009-07-08 at 08:32 -0400, Dan Mongrain wrote: > RFC 3261 basically mentions that UAC can fail a transaction in the case > of no response after 64*T1 (retransmitting every 2*T1). With the > default T1 being 500ms, 32 seconds is a long time to wait especially > when there are multiple proxies available to process a request. Is > there a RFC that specifies the behaviour of a UAC when there are > multiple proxies in order to accelerate failed proxy detection in order > to rapidly fallback to another proxy?
In sipXecs, we did quite a bit of design work regarding this. Though the RFC specifies the retransmission and fallback times in great detail, there's no particular reason to follow those times that we've discovered. We've set the first retransmission at slightly greater than the round-trip-times we observe routinely, and the final timeout at slightly greater than the longest RTTs we see in practice, with each resend doubling the time. So the current resend schedule is T+100ms, T +200ms, T+400ms, and T+800ms, with timeout at T+1600ms. This allows fallback to another server in 1.6 second, which is not obnoxious to users. Dale _______________________________________________ Sip-implementors mailing list Sip-implementors@lists.cs.columbia.edu https://lists.cs.columbia.edu/cucslists/listinfo/sip-implementors
