[Asterisk-Dev] T.38 in asterisk.. need did

Steve Underwood steveu at coppice.org
Fri Apr 1 22:04:58 MST 2005


Preston Garrison wrote:

> My next project after finishing completely rewriting chan_sip.c is 
> t.38.  I am looking for someone who can provide me a did that goes 
> direct to someone like Global Crossing that supports t.38.  I need 
> this for testing to of the codec.  Looking for someone who has them 
> cheap or even would be willing to donate them to help get t.38 support 
> in asterisk.   I know there have been a few people that have worked on 
> this from one time to another, but I have yet to see any forward 
> progress.  If i am wrong let me know :)
>
You will find this of limited value. I have much of T.38 over UDPTL 
working with * now. Each T.38 box you try does something different, and 
usually something rather nasty and troublesome. This is to be expected. 
The T.38 spec is very open ended and leaves much of how it works up o 
the implementor - a sure receipe for trouble. I have found development 
requires working with several T.38 implementations, to try to find a 
common ground implementation that will work with them all.

What have you done so far? I have only done the minimum possible to 
chan_sip up to now. I have concentrated on the T.38 path itself. I have 
rxfax and txfax talking T.38 as well as analogue lines. I have a 
udptl.c, which behaves like rtp.c for a UDPTL stream. It does all the 
UDPTL packaging and upackaging, and error correction. I am adding full 
error correction to this at the moment. It works, but I am stripping it 
down from a generic GPLed ASN.1 PER implementation, to a leasn and mean 
one which can go into Asterisk. I have code that will interface an 
analogue line to a T.38 channel, using the modems in spandsp. I haven't 
done anything about T.38 over RTP or TCP yet. Hardly anything seems to 
support those, so it is a low priority.

So far this biggest time consumption of my time has been trying to deal 
with the poor implementation which are available to test with.

Regards,
Steve




More information about the asterisk-dev mailing list