[asterisk-users] Ast/Hyla/IAX Scalability?

[Steve Underwood]

David Backeberg wrote:
> On Tue, Mar 17, 2009 at 10:34 AM, Steve Davies <davies147 at gmail.com> wrote:
>   
>> While we have your attention Steve (Underwood) do you have a
>> high-level changelog available for spandsp-0.0.4 to 0.0.5 to 0.0.6? We
>> currently use 0.0.4 with a very high success rate. Is there any
>> benefit in moving up to a newer library? I looked at the Changelog in
>> the source, but it stopped at 0.0.4.
>>     
>
> I'm not Steve, but I can tell you what I've found.
>
> 0.0.5 seems to be required for building asterisk-1.6.0.6
> 0.0.6 introduced some API changes, and trunk has been updated, but
> asterisk-1.6.0.6 does not have those changes, and you won't be able to
> compile unless you replace app_fax.c with the version in the 1.6.0
> subversion branch.
>
> I suspect (since it's in 1.6.0 branch) but cannot confirm that the
> next 1.6.0. release will have the changes to build / use spandsp-0.0.6
>
> the 5 and 6 have progressively better support for misbehaving faxes,
> whereas faxes that behave according to spec work well on all versions
> I've tried.
>   
That's about right. The changes to make 0.0.6pre6 and beyond work with 
Asterisk 1.6.0 are tiny - well, you can see them in 1.6.1, as it uses 
the date stamp of the spandsp version to select how to build.

quite old versions of spandsp work well with clean FAXes. 
spandsp-0.0.6pre7 works well with some pretty messed up ones. I believe 
the results with 0.0.6pre7 should be comparable to spandsp + iaxmodem + 
HylaFAX, as long as the timing within your Asterisk is OK. The use of 
IAX introduces the possibility of dropped packets, but also adds some 
timing elasticity.

For the kind of results you can expect see 
http://www.soft-switch.org/spandsp-soft-fax-performance.html

If you have more than 1% FAX failures which cannot be explained, you 
have a problem :-)

Regards,
Steve