The difficulty with giving a 6xx response is that it can muck up forking 
*upstream* of the element that generates the 6xx response.

Consider typical example:

AOR sip:[email protected] is forwarded to AOR sip:[email protected] and then serially 
forks to 1234's voicemail, sip:[email protected].

At proxy B, AOR sip:[email protected] is forwarded (incorrectly) to 
sip:[email protected], with no other targets.

(And while we're at it, let us note that "invalid-domain.net" is *NOT* reserved 
by RFC 2606.  Indeed, whois reveals that it is registered by Ludger Keilig of 
Bonn, Germany.)

The desired behavior is that a call to sip:[email protected] will eventually 
arrive at sip:[email protected].

When an INVITE is sent to sip:[email protected], it will first be forked by proxy 
A to sip:[email protected].  Proxy B will attempt to fork the INVITE to 
sip:[email protected].  The problem to be answered is:  What response code should 
proxy B give to proxy A?

If proxy B gives a 6xx response, then by RFC 3261, section 16.7, item 5, page 
109, proxy A may not serially fork the INVITE to sip:[email protected].  but if 
proxy B gives a 4xx response, proxy A continues the forking of the INVITE to 
sip:[email protected].

This pattern generalizes to all situations where 6xx responses have been 
proposed, which shows why 6xx responses should never be used.  See 
draft-worley-6xx-considered-harmful-00 (dated Feb. 2007).

Dale

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

Reply via email to