> > > In testing labs I am seeing a worrying number of UA
> > > implementations making the same mistake of sending requests
> > > to a Proxy where the Request-URI is in the form
> > > sip:user@<proxy-address>. I think some of the confusion may
> > > be traced back to early SIP message flows which showed this
> > > behaviour. I urge implementers to rectify this situation as
> > > we are seeing this cause a great deal of confusion to
> > > the first commercial adopters of SIP technology. These people
> > > are not SIP protocol experts and just expect this technology
> > > to work.
> > >
> > > It is often seen when using a local out bound proxy. The
> > > situation is that the UA takes a configuration parameter
> > > for the local proxy. You can then dial by keying in just the
> > > user part of a SIP URL. The UA then builds a complete URL by
> > > adding the Proxy's address and sends the message to the proxy.
> > > The Proxy then receives an INVITE of the form
> > > sip:user@proxy-ip - according to the protocol spec the Proxy
> > > will route this request to the server at the given IP address.
> > > This is obviously itself so an illegal loop is created. A
> >
> > An outbound proxy for userA can be configured
> > to translate upon a received outgoing message
> > from known userA however it wishes. UserA
> > dialing *69 on a phone can be sent to outbound
> > proxy with Request-URI of *69@outboundProxy-Ip.
> > Assuming the outbound proxy knows userA, it can
> > change the Request-URI into the last "known"
> > user@host that called userA.
>
> According to the SIP spec sip:*69@outboundProxy-Ip
> does not map onto the last incoming caller's sip URL
> this is specialised logic. It may exist but a truly
> interoperable UA must still be able to work if it
> doesn't.
>
> The point I was really trying to make is that a UA
> should not write Request-URIs out to be the destination
> of where they are being sent in the hostport when using
> a local outbound proxy.
>
> Say I want to dial sip:[EMAIL PROTECTED] but
> the UA configured with a local outbound proxy setting
> of proxy.this-domain.com actually sends the INVITE
> out as:-
>
> INVITE sip:[EMAIL PROTECTED] SIP/2.0
> To: sip:[EMAIL PROTECTED]
> From: sip:[EMAIL PROTECTED]
> ....
>
A proxy receiving the above request assumes
that the user is trying to reach the Request-URI.
By becoming aware the need to act as an
outbound proxy, the Request-URI can be changed
by the outbound proxy. When any proxy or user
agent receives the initial INVITE, the user@hostport
of "To" should only really mean what was supposedly
dialed/entered to reach you (and call-leg identification).
If the user agent was trying to reach
sip:[EMAIL PROTECTED],
it should be in the Request-URI. And the
message should be sent to
proxy.this-domain.com if it wants to use it
as an outbound proxy.
Note: Recent threads on outbound proxies also
define the use "Route" to allow for an outbound
proxy chain. However I don't think that discussion
has been finalized.
> This will either result in a 404 response, possibly
> 482 if the Proxy sends on to itself, or go to the
> wrong jsmith. If a UA can only send out INVITEs
> with Request-URIs like this it is broken.
>
> Cheers,
> Neil.
> --
> Ubiquity Software Corporation, UK http://www.ubiquity.net
>
> _______________________________________________
> SIP mailing list
> [EMAIL PROTECTED]
> http://lists.bell-labs.com/mailman/listinfo/sip
>
