[Asterisk-Users] HylaFAX v. spandsp

[Steve Underwood]

Michael Welter wrote:

> Can someone explain where we are with spandsp?  Is it ready for a 
> production environment?  How much will one fax using spandsp load the 
> processor on an * system?
>
> Thanks,
> Mike

Some people have been using spandsp-0.0.1k in production systems for 
months. The main problems have been:

    Canon fax machines often gave problems
    HP machines are buggy, and with spandsp the bugs cause black streaks 
oon received documents.
    People get pages chopped in half.

You won't really notice the load caused by one channel of spandsp's fax 
machine on most modern CPUs. I had 100 channels running on a dual Xeon 
at one point, but I haven't tested recent versions at high loads.

With the latest spandsp-0.0.2pre4
    The bug affecting Canons seems to be resolved.
    The software used to rely on the internals of libtiff to compress 
and decompress images. Now it uses its own code which works around buggy 
machines likes the HPs, and allows better reporting of the received data 
quality.
    Of all the pages getting chopped problems I received, one turned out 
to be a stability issue in the modems. This seems to be resoolved.

Every other report of pages getting chopped, that I have investigated, 
has been due to timing problems in * or the hardware. This is probably 
about 30 examples now. There is nothing I can do about this. If spandsp 
does not work in these cases, neither will hardware modems and HylaFAX 
with the calls passing through *. Modems just cannot tolerate hiccups in 
the audio stream.

The long delay between spandsp-0.0.1k and the recent spandsp-0.0.2prexxx 
versions has been due to the fact that spandsp-0.0.1k has been working 
pretty well for a lot of people. There wasn't a need for many minor 
releases. 0.0.2 is a fairly major update, with many areas of the 
software changing in major ways. Some of these are complete, and some 
are still works in progress.

The modems have been reworked to get closer to the theoretically 
achievable performance on noisy lines. Their compute requirements are 
also a bit lower now.

Fax compression and decompression has been added.

An incomplete skeleton for class 1 modem operation, to work with 
HylaFAX, has been added.

A nearly working V.17 modem has been added. This still needs 
improvements to the receiver carrier locking to make it work properly.

A half finished V.22bis modem has been added. A couple of people asked 
for this, so allow integration of things like credit card validation 
machines with *. Most of these still sues V.22bis.

Variable speed playback of speech files. This is something a lot of 
people ask for, but which is fairly useless in practice :-). Its just to 
play with, though.

A number of little modules, such as V.8 processing, have been 
implemented, tested, and added to spandsp, but are not really being used 
right now.

Regards,
Steve