<div><div>Hello,</div></div><div><br></div><div>I'm experiencing an issue similar to <a href="http://jira.freeswitch.org/browse/FS-4669">Jira FS-4669</a> on 1.3.14b+git~20130306T182328Z~3e4fb4b0a2 (git 3e4fb4b 2013-03-06 18:23:28Z). The scenario at my customer's site is:</div>
<div><br></div><div> XMediusFAX -> FS -> PSTN (usually Sonus GSX)</div><div> or</div><div> PSTN -> FS -> XMediusFAX</div><div><br></div><div>The XMediusFAX software and the PSTN are on two different sofia profiles with the call bridged between them. Either side can send the T.38 reinvite. If the sofia profiles are configured for proxy-media, 100% of calls are successful. If the sofia profiles are not configured for proxy-media but rather only t38-passthru, approximately 50% of the calls are successful. There are no config changes between a successful and a failed call.</div>
<div><br></div><div>On the failing calls the reinvite completes successfully, but FS appears not to pass the UDPTL properly. Wireshark shows only egress "Unknown RTP Version 0" where it would show UDPTL if the call were successful. In other words, the ingress UDPTL appears correct, but on failing calls the egress media does.</div>
<div><br></div><div>I'm already yelling at myself for not filing a Jira yet. That's the plan, but I can't get it to fail in the lab, and the customer pcaps have too much data I'm not allowed to share. I'm wondering if anyone might have a thought about what *might* cause this so I can make sure I reproduce it in the lab accurately and file the Jira with sufficient data.</div>
<div><br></div><div><br></div><div>- Jeff</div><div><br></div>