From: "Sweeney, Andrew \(Andrew\)" <[EMAIL PROTECTED]> I am trying to determine if transport=tcp must be added to a request when the user is going to run over TCP.
First, make sure you've read RFC 3263. The general philosphy is "A SIP agents has a SIP message that it needs to send to a particular destination -- what should it do to send the message?" Among other weird possibilities, different requests in a dialog can be sent via different transports. In particular, if the host-part of the URI is described by DNS SRV records, which SRV records exist constrains what transports can be used. To pick a real example, "sip:[EMAIL PROTECTED]" *must* be contacted via TCP. (See http://interop.pingtel.com, item 18.) A remote device initiates and establishes a connection over TCP but doesn't include transport=tcp in the uri or the contact. The local device wants to do a transfer with re-invite. Can he do this over udp or must he use tcp since the original dialog is on TCP? If you mean, "How does the local device send the re-INVITE?", that depends on the Contact address of the far end, or the first Record-Route address, interpreted according to RFC 3263 (as modified by local policy, especially an outbound proxy). Dale _______________________________________________ Sip-implementors mailing list [email protected] https://lists.cs.columbia.edu/cucslists/listinfo/sip-implementors
