On Thu, 2006-05-04 at 10:58 -0400, Bob Penfield wrote: > Non-success final responses (3-6xx) and the ACK to those responses are > hop-by-hop. Regardless of whether or not the proxy forwards the non-success > final response upstream toward the UAC, it MUST always send an ACK.
> This applies to proxies as well as UAs. That's true in theory, but there are devices that behave mostly like proxies but sometimes act more like B2BUAs. The "who eats the response" test works correctly in all sorts of unpleasant cases. Though in most cases, the device is proxy-like in that it can (in principal) fork any incoming request, so it's really eating any response (in order to chose the best response of the responses it got (even if it got only one resonse)). But there are cases where the device is not intended to fork, but only to execute an elaborate transformation of SIP messages on their way through, and in those cases it's acceptable (possibly preferable) to pass non-success final responses through and let the other agents generate the ACKs. Dale _______________________________________________ Sip-implementors mailing list [email protected] https://lists.cs.columbia.edu/cucslists/listinfo/sip-implementors
