[asterisk-users] Digium Fax Driver

[Steve Underwood]

Lee Howard wrote:
> Tilghman Lesher wrote:
>   
>> On Sunday 07 June 2009 19:39:50 Lee Howard wrote:
>>   
>>     
>>> Tilghman Lesher wrote:
>>>     
>>>       
>>>>> What's the use case for the Digium
>>>>> driver? Am I missing something by not using it?
>>>>>         
>>>>>           
>>>> While they accomplish the same goal, the commercial driver is based upon
>>>> a different codebase,
>>>>       
>>>>         
>>> Ok.
>>>
>>>     
>>>       
>>>> provides support for patented fax protocols,
>>>>       
>>>>         
>>> Really?  V.34-fax (33,600 bps) is supported?  I had understood differently.
>>>     
>>>       
>> I would research the patents involved, but I am prohibited by employment
>> contract from exploring patents granted.
>>     
>
> Due to said employment contract prohibitions you can't tell me whether 
> or not Digium's Fax Application supports V.34-fax (33,600 bps)?
>
>   
>> My understanding is that there are
>> certain aspects of fax that are still under patent,
>>     
>
> Yes.  Specifically V.34.  If my understanding is correct the relevant 
> patents expire in a few years.
>   
There are actually 3 aspects of FAXing still under patent protection:
   - Numerous patents relate to V.34, and the last to expire will still 
be several years away
   - There is still one patent in-force related to JBIG compression, 
which expires next February
   - The TIFF/FX appears to possibly have Xerox and Adobe's claws in it, 
but the position is not very clear.

The Digium FAX driver is clearly stated to not support V.34. I find this 
odd, as Commetrex, who supply the FAX engine Digium use, are supposed to 
have a V.34 engine, and other people (e.g. Pika) say they use it. V.34 
would have been a real value add over the free options, as a free option 
can't provide V.34 for several years.

Since Commetrex support JBIG, I assume the Digium FAX driver also does. 
However, JBIG is not that big a win. Support for it in FAX machines 
seems patchy. Maybe the makers don't want to pay patent royalties. I 
intend to add JBIG support to spandsp next year when the last patent 
expires.

TIFF/FX seems to be in limbo, with limited support in day to day usage.
>> and those are provided
>> (along with indemnification) by the commercial driver.  
>>     
>
> Understood.  But it was my understanding that V.34-fax was not supported 
> by Digium's Fax Application.  And if that's correct, then there are no 
> patents for which indemnification is necessary.  That's not to say that 
> a commercial fax driver does not have its place with some customers.  I 
> only want to clear up any misrepresentations about possible patent 
> infringements by spandsp to which you alluded.
>   
See above.
>> I'm not suggesting that the commercial driver is more reliable,
>> only that it enjoys far more testing.
>>
>>     
>
> Again, regardless of your knowledge of how much testing goes into your 
> employer's product, I question your ability to know with any degree of 
> certainty as to how much testing has been involved with competing 
> products.  I certainly know that *I* have no clue with regards to 
> spandsp other than the testing to which I've been witness.  So I am 
> curious to know how you are able to make such assertions.
>   
The Digium FAX driver is based on the Commetrex engine, which is widely 
deployed and presumably robust. How reliable the overall package might 
be is another matter. For a long time, the limiting factor in FAX 
reliability with Asterisk has been the inability of Asterisk and/or 
DAHDI to provide a clean audio stream. In commercial FAX servers using 
spandsp (either in iaxmodem or on its own) the reliability is mostly 
limited by Asterisk. Recently, with the launch of the Digium FAX driver, 
Digium has done some fudging in chan_dahdi to try to mitigate these 
problems. So far, they only seem to have reduced the issues a bit, and 
not solve them.
>> That said, hours of use in production do not speak to the amount of testing
>> done.
>>     
Right, but hours of production use on instrumented servers, which save 
the audio from every failed call for later analysis can do wonders. 
We've done this with iaxmodem+HylaFAX and with spandsp in Asterisk and 
Callweaver systems. Some people have handling hundreds of thousands of 
FAXes a day, so you can quickly build an interesting library of weird 
behaviours. It was very time consuming to find and understand all the 
weird stuff real world equipment throws at you, but we got the 
unexplained call failures to well below 1% about 3 years ago with 
iaxmodem, and spandsp is now about there too.
>
> Scrutiny of production use exposure does not constitute testing?  Well, 
> I would argue that you cannot possibly test real-world conditions 
> without actually placing the test system into the real-world with 
> real-world use (thus, production).  I cannot think of a better way to 
> test software than to eventually put it into real-world production use 
> and then have the developers monitor those systems closely.
>
>   
>> IAXmodem is a completely different ball of wax, and I think you would agree
>> that if the builtin FAX support in spandsp provided excellent support, there
>> never would have been a reason for IAXmodem to be developed.
>>     
>
> I'm interested to know how you understand my intent in developing 
> IAXmodem differs from what I recall.  I developed IAXmodem because I 
> needed to interface HylaFAX through an Asterisk PBX without purchasing 
> additional hardware (other than the T1 cards that were already involved).
>   
As Lee said, HylaFAX offers something different. There are numerous 
tools that work with HylaFAX, and iaxmodem allowed people to tap into 
those from an open source telephony platform. People asked for a FAX 
modem to use with available FAX servers, so I responded when Lee said he 
would work with me. Although most often used with HylaFAX, iaxmodem is 
actually in use with a variety of FAX server software.

When spandsp is used as a complete FAX machine its cut off from the 
HylaFAX ecosystem. I have considered developing a HylaFAX compatible 
program to interwork with the HylaFAX world, but HylaFAX already does a 
good job when used with iaxmodem. The biggest drawback has been from the 
use of IAX, because it means iaxmodem can only do audio FAXing. I've had 
a kinda-working-but-not-production-ready SIPmodem for ages, which does 
allow audio and T.38 from the same HylaFAX system, but I haven't found 
the time to complete it.

Regards,
Steve