[Asterisk-Dev] SIP / H.323 Gateway with Asterisk

[Aaron S. Joyner]

We are using SIP phones to make calls through Asterisk, out an H.323 
connection.  Depending on the end point, some times these calls do not 
pass audio.  The call setup appears normal, everything about the 
negotiation appears to be correct, but no audio is passed out of 
asterisk, in either direction.  From a packet dump we can see that audio 
is coming in from both sides, but no audio is being sent out from Asterisk.

The various iterations are as follows:
- If we place a call to another Asterisk server, running identical 
versions Asterisk, the h323 channel driver, and OpenH323 - we get no audio.
- If we call a Cisco PSTN gateway, no audio
- If we place a call to a Linux workstation running GnomeMeeting - the 
call works just fine!  Audio is passed successfully, and everything is 
precisely as expected.

The dial strings in Asterisk for all of these incantations are identical 
(although we have of course experimented with many variations).  As per 
a trace of the h.323 debugging, the codecs being negotiated are always 
G.711u.  As an example, the dial strings we've been using are often 
formatted like this (some IPs and numbers have been changed to protect 
the innocent):

> exten => 2999,1,Dial(H323/XXXXXXXXXXX at 111.222.333.444/9195735483)

What is different when calling another Asterisk server, or a Cisco 
gatekeeper, vs calling a GnomeMeeting workstation?  Why does one 
function and not the other?  For reference purposes, the GnomeMeeting 
client can call the same remote Asterisk server, with exactly the same 
number (for example: 600 at hostname.of.machine) and get a valid 
extension.  The GnomeMeeting client can also call a SIP extension on 
either Asterisk server, so I'm certain that Asterisk is capable of 
gatewaying between SIP and H.323.  I've heard the audio go through 
myself -- but only when originated by GnomeMeeting, as opposed to 
originated from Asterisk.

A colleague of mine discovered this 
(http://bugs.digium.com/bug_view_page.php?bug_id=0000562) over the 
weekend, which we think may be directly related.  We're going to be 
implementing this patch to test it this morning.  From the sounds of it, 
his problem may be directly or indirectly related to ours, and although 
it's not an identical problem, his fix may actually encompass both problems.

If anyone has any insights or thoughts as to why this doesn't work, or 
how to go about fixing it, it would be greatly appreciated.

-- 
Aaron S. Joyner
System Administrator
Intrex.net Internet Services
(919) 573-5488 x102