On 11/17/2010 03:00 AM, Kristian Kielhofner wrote:
Dave,
It's a better parameter for T.38 in general...
When you look at a T.38 trace (presumably without ECM) you'll see that the T.38 data stream is entirely unidirectional. The sending device sends a stream of data - that's it. Er, no. There is is two way exchange between each page, or between each chunk of a page in ECM mode. You should require a single page that takes longer than 600s to send for this timer to expire. I first caught on to this with large faxes or faxes with high complexity that took a while to send. I finally found one fax page with multiple images that failed 100% of the time when being sent from FreeSWITCH to the device. Once I set the rtp-timeout-sec parameter to 600 it succeeded 100% of the time. This change was (and has been) the single biggest change I've made to my configuration to improve my T.38 success rate.
Ideally FreeSWITCH could be updated with an "enhanced understanding" of the unidirectional flow of T.38 and modify/tweak/remove the RTP timeout parameter as necessary (I imagine it does something similar with silence suppression/CNG now). Or maybe I'm completely wrong and this change had no effect (I seriously doubt that).
On Tue, Nov 16, 2010 at 11:14 AM, Dave Redmore <dave.redmore@spigotsystems.com> wrote:
I'm curious about your recommendation to use a long (600 sec) media/RTP timeout value in the sofia profile. Is this a specific need for the Audiocodes device or just generally better for T.38? I'm assuming you are referring to the "rtp-timeout-sec" parameter?
I'm using FreeSwitch as a media proxy to a T38 provider and have had decent success, but usually have problems with longer faxes (8+ pages) and am wondering if this setting might help?
Thanks,
Dave Redmore Steve