[asterisk-users] OT - "T.38 unreliable on a LAN" : truth or obscurantism ?

[Adam Moffett]

On 02/15/2012 05:33 PM, James Sharp wrote:
> On 02/15/2012 03:03 PM, Olivier wrote:
>> Hi,
>>
>> When someone says "T.38 is not reliable on a (normally loaded and
>> managed) LAN", would you rather agree or disagree ?
>> In this case, fax calls are coming in through an analog gateway,
>> passing trough Asterisk and then going out to ISDN through a digital
>> gateway.
>>
>> Comments ?
>
> While I can't speak for Asterisk's T.38 performance (it was barely 
> past the point of "okay, it compiles" at the time of this datapoint), 
> T.38 in general can handle nasty network conditions without a problem 
> as long you enable some sort of error correction (either FEC or packet 
> redundancy).  Case in point, I ran several hundred SIP-based T.38 
> calls a month over VSAT links.  The links ran anywhere from 550 to 
> 750ms latency and would average around 1-2% packet loss (averaged over 
> a 5 minute period).  Those were with a Quintum ASG400 at the far end 
> and a Quintum CMS960 going into PRIs at the VSAT hub.
>
> So if T.38 can handle that, it can handle just about anything.

It seems like the bigger issue is interop between different gateways and 
fax machines.  It seems to work 99% of the time, but that 1% is 
impossible to fix if you can't control all the variables.  If you 
*could* control all the variables you'd probably not be faxing. :)