[For anyone on Sip-Implementors, sorry for crossposting this here. However, it occurred to me that viewed in a certain light, this could be seen as more of an operational issue than a theoretical one, and I wonder if any of you have encountered it too.] Hi everyone, I am faced with what strikes me as a genuine mystery. In the last few days, I have encountered several completely unrelated commercial SIP UAC implementations that all exhibit the same erroneous behaviour. This behaviour, while certainly incorrect as far as I can tell, is consistent across three or four completely unrelated stacks, leading me to wonder what I am missing. The scenario is that UAC A places call to UAS B through stateful proxy P: A ---> P ---> B P inserts Record-Route header on the initial INVITE in order to stay in the loop on sequential requests. Nothing eccentric in that inherently. When final 200 OK response to the initial INVITE transaction is relayed from P back to A, it contains a normal Contact URI which should be used as the remote target URI for sequential requests (absent any requests that would refresh it): Contact: <sip:xxx at a.b.c.d:5060> It also contains a Record-Route header which has been copied into the reply by B, of course, which looks something like this: Record-Route: <sip:ip.of.proxy.p;lr;param1=blah1;param2=blah2> But bizarrely, the end-to-end ACK sent by A to this 2xx response has a request line formed not from the Contact in the 2xx reply, but from the content of the Record-Route header: ACK sip:ip.of.proxy.p;lr;param1=blah1;param2=blah2 SIP/2.0 Naturally, the proxy is quite bewildered and does not propogate this. 3261 scripture seems quite clear--at least, to me--on why this would be wrong. 12.2.1.1 ("Generating the [in-dialog] Request") says: If the route set is not empty, and the first URI in the route set contains the lr parameter (see Section 19.1.1), the UAC MUST place the remote target URI into the Request-URI and MUST include a Route header field containing the route set values in order, including all parameters. Other parts of the spec make quite clear that this kind of ACK is an in-dialog request, and that the Contact in the 2xx response establishes the "remote target URI" aforementioned. In other words, pretty commonsensical stuff. I think anyone can agree the request line of the ACK should be: ACK sip:xxx at a.b.c.d:5060 SIP/2.0 However, the fact that I have encountered this in completely unrelated situations with an improbable frequency leads me to believe that there must be some element of this I am failing to grasp. Is it some sort of legacy behaviour and/or RFC 2543 backward compatibility mode that is causing this, or perhaps something else? Thanks, -- Alex Balashov - Principal Evariste Systems LLC 1170 Peachtree Street 12th Floor, Suite 1200 Atlanta, GA 30309 Tel: +1-678-954-0670 Fax: +1-404-961-1892 Web: http://www.evaristesys.com/