Hi, consider INVITE was cancelled before getting _any_ reply to it (i.e. even 100 wasn't got) and top logic level timeout requires to stop this call. RFC3261 doesn't list cancelling as reason to make any change to INVITE transaction logic; so TU needs to continue resendings of initial INVITE until timer B explodes or any reply to INVITE is got (and here cancelling can start, sending CANCEL if provisional reply is got and BYE if 2xx is got). This is excessive, confuses admins and network analyzers, create race conditions... My question is: what are side effects of solution to stop retransmission of initial request (i.e. timer A), while keeping all rest of logic (timer B, state machine with processing responses...)? I see only advantages: if remote party is down, no useless packets will be generated; if there were network failure, as top-level logic already switched to another destination or even stopped the whole call, so the probability to connect call which is already useless is reduced. But maybe I missed something?
-- Valentin Nechayev PortaOne Inc., Software Engineer mailto:[EMAIL PROTECTED] _______________________________________________ Sip-implementors mailing list [email protected] https://lists.cs.columbia.edu/cucslists/listinfo/sip-implementors
