Russ Daigle <[EMAIL PROTECTED]> wrote:
> Thanks for you response.
>
> Do you per chance mean "can't put for example port 5060 in the
> CONTACT header..."?

The contact header has no bearing of the routing of responses to 
requests you send.

I thought your question was only about response routing, but no - you 
should not put something invalid (like a port where you don't listen) in 
the contact header of the requests you send, if you want requests sent 
in the other direction to work.

> You mentioned Via: header,  but I thought the
> Via header should have the actual source port for responses of the
> same transaction.  For TCP it seems redundant to put source port in
> the Via header, since responses should come back on the same socket.

The Via header doesn't have to include a port number. If it does, then 
the port number should make sense. The Via header _might_ need to 
include a port number, or the "rport" parameter.

> I presume Contact: header should have TCP/TLS port that we are
> listening on for subsequent requests that use a new TCP connection
> (and not TCP source port of the current TCP connection).  However, I
> presume it is harmless to put TCP source port of current socket in
> the Via header.

Should be - unless there is a NAT in between you and the party you are 
testing with, that works in a way that changes the port number. Say that 
you send the request from port 1234, and a NAT changes that to port 
2345 - a response sent to port 1234 would probably not reach you. This 
is what "rport" is for.

/Fredrik

_______________________________________________
Sip-implementors mailing list
[email protected]
https://lists.cs.columbia.edu/cucslists/listinfo/sip-implementors

Reply via email to