James Jack wrote:
>     I have one question to consult, for
> example , for a stateless proxy, according to
> RFC2543bis02, it says it can not generate 
> provisional response(but it does not say it
> can not generate final response), I think since it it
> can not generate provisional response, it can not
> generate final response. Then in implementation,
> should
> the stateless proxy check loop or not, for example, if
> one INVITE request is loop,what would stateless proxy
> deal with,one terriable situation is that loop may
> cause this message is sent from this stateless proxy
> to this stateless proxy.
> 
>   The same situation happens if the value in
> Max-Forwards is zero in one INVITE request, what would
> statelss proxy deal with this INVITE request? Discard 
> it?

The proxy could discard the message, but I wouldn't be too
impressed.

I can think of two possible approaches:
 * Respond statefully; that is, realise that statelessly
   proxying isn't viable (because the request has looped,
   or it's reached Max-Forwards, or whatever), and switch
   to "stateful mode" for this particular transaction.
 * Respond statelessly; that is, return a final response
   for each retransmission of the request.

Note that the second approach isn't so blessed by the
spec, but it seems to me that it's not an unreasonable way
to act statelessly in various error conditions (I think it
was originally proposed by Jonathan as a pattern for
stateless redirect servers).

There is a slight problem, however.  If you respond
definitively to an INVITE statelessly, you're going to get
an ACK.  Proxying this ACK, I feel, would be a little
bit poor, so I would suggest ensuring that the final
response you send contains a very &:) unique tag parameter
in the To.  Then if you get an ACK with this tag in, you
know to drop it.

What if the To already has a tag?  Well, I guess you're
outta luck -- you have to respond statefully.

HTH,


 - Jo.

_______________________________________________
Sip-implementors mailing list
[EMAIL PROTECTED]
http://lists.cs.columbia.edu/mailman/listinfo/sip-implementors

Reply via email to