On Thu, 2006-03-23 at 10:19 -0700, Barker, Mark wrote:
I believe that a structure like you describe cannot be implemented.
> Q2. How will the Request-URIs be constructed for outgoing REGISTER
> requests from an endpoint?
The correct way to look at the system starts with "What is the Address
Of Record that the endpoints will have?" They should be something like
"[EMAIL PROTECTED]", "[EMAIL PROTECTED]", etc. The "domain" part needs to
route to
the registrar/proxy when the rules of RFC 3263 are applied. In the
simplest cases, that means that "domain" is either the IP address of the
registrar/proxy, or a DNS host name that maps to the IP address of the
registrar/proxy.
Once you've settled on that, you have to arrange for each endpoint to *a
priori* know the "domain" and its "user" (EP1, etc.).
When the endpoint starts, it constructs the REGISTER request thusly:
REGISTER sip:domain SIP/2.0
To: <sip:[EMAIL PROTECTED]>
From: <sip:[EMAIL PROTECTED]>
Contact: <sip:[EMAIL PROTECTED]>
The endpoint can now carry out RFC 3263 on "domain" to determine where
to route the message, or it can simply forward the message to its pre-
configured "outbound proxy" (which will carry out RFC 3263 to further
route the message).
As you can see, as you've outlined it, it is impossible for the
endpoints to determine the "domain" to use -- it is not constant, and
there is no way to convey it to the endpoints, because the *first* SIP
message they send requires that they know it. Conceptually, they have
no idea *which* registrar/proxy in the Internet they should be talking
to.
One way to solve this would be to add a custom extension datum to DHCP
which carries the registrar/proxy's IP address. That might be workable
if you have a closed network.
Dale
---
interop.pingtel.com -- the public SIP phone interoperability test server
_______________________________________________
Sip-implementors mailing list
[email protected]
https://lists.cs.columbia.edu/cucslists/listinfo/sip-implementors
