On 06/11/2011 09:48 AM, Joegen E. Baclor wrote: > On 06/10/2011 11:09 PM, Iñaki Baz Castillo wrote: >> 2011/6/10 Brez Borland<[email protected]>: >>> Not clear to me though. If proxy has received a CANCEL from Alice it should >>> terminate the transaction. >> This is incorrect. The proxy does not terminate an INVITE server >> transaction when it receives a CANCEL. It just cancels pending >> branches and UAS's are supposed to termiinate the transactions (487) >> or maybe local timeout occurs (not responding branch so local 408 is >> generated). > Which begs the question why then do we accept CANCEL UAC rules to apply > to proxy transactions and not UAS rules? Shouldn't they always be a > pair to ensure end-to-end transaction states are synchronized?
I guess, it might be because it is better to terminate dialogs that may be created after CANCEL has been sent by a BYE from the UAC rather than an implied one through ACK timeouts. This is, however, deduced and not so clear in 16.10. IMHO, it is better to forward CANCEL for non-forked transaction and only respond to it and enforce CANCEL UAC rules if the proxy forked the INVITE transaction. _______________________________________________ Sip-implementors mailing list [email protected] https://lists.cs.columbia.edu/cucslists/listinfo/sip-implementors
