Content-Type: text/plain;
  charset="utf-8"
Content-Transfer-Encoding: 8bit
Organization: SipXecs Forum
In-Reply-To: 
<CAMgKNJWQao0SctwLUTHgeO610cdARSWkxW22LTgU8n=uyay...@mail.gmail.com>
X-FUDforum: 08063afcdd00a6e76393c5b9527381e8 <63116>
Message-ID: <[email protected]>



Hi Tony,

   So, in reading, re-reading, and re-evaluating things
based on what you and others here have said, I think I've
realized that the scope of affected transfers is smaller
than I thought (although, still critical for us).

   The issue only occurs when a call comes in through the
patton gateway, to a SipXecs extension, which is
subsequently transferred/forwarded to an extension that is
only reachable back via the Patton gateway.  I believe I've
seen this referred to as a 'hairpin' call.

   As you mentioned, you thought the difference (routing
looking, and therefore the URI transition) should happen
during the INVITE.  This certainly works in all cases where
the call is initiated on the SipXecs server (ie. using one
of the SipXecs registered endpoints), but in this case, the
PATTON is responsible for creating the INVITE in response to
the transfer, not the SipXecs server.  You can see this
behaviour at line 521 of the 'External to VoIP DID
transferred to Nortel extension.txt' file I attached
earlier.

   The Patton is doing exactly as it is instructed in the
REFER-TO request, and attempting to transfer the call to
mailto:'[email protected]'.



   On the somewhat separate topic of preceeding 4-digit
extension numbers with a '99' when dialed, you're right, in
hindsight, we could have done without it. This was as much
to help with our Nortel dialplan as anything.  It does not
affect the current issue (the transfer doesn't work if I try
to transfer to 994495 either), and I simply pointed it out
to give another indication as to where the trouble existed. 
I should also point out that this works quite well, and is
well within the realm of 'correct configuration', done via
the standard interface, regardless of whether it's how you
would do it or not.  It is completely transparent to the end
user.

   What I'm hearing from you is, the SipXecs server does NOT
lookup/translate/modify numbers in REFER-TO packets, because
it is believed that any translation can be handled during
the resulting INVITE.  I believe that, in the case of
'hairpin transfers', where the SipXecs server is NOT
responsible for creating the INVITE resulting from a
REFER-TO, this approach falls short.

...Steve...
_______________________________________________
sipx-users mailing list
[email protected]
List Archive: http://list.sipfoundry.org/archive/sipx-users/

Reply via email to