Re: TPC-OPER: Re: flexfax: Large number of errors with Hylafax and

[David Woolley <david@djwhome.demon.co.uk>]

> I don't think so. It lokos like there are some problems in software. I can
> see almost all mentione errors in almost totaly controled envronement
> with guaranted good lines and with control over both side faxes (be it
> standalone or fax/modem.
> 
> So if someone will came with an idea or patche, I can test things in such
> environement. 

Retraining is only any use in dealing with a bad line, not with protocol
problems.  It may be that the remote fax modem (including one in a fax
machine) is using the wrong criteria to request a retrain, but if it is
doing so, it is at fault.

(The receiving modem has to model the characteristics of the line, which
will tend to blur one bit into another (one signalling unit, which is really
multiple bits).  To do this, part of the start up of the modem is a sequence
of increasingly complex, but predictable signals, that the receiving modem
can use to calibrate its filters (basically what it is doing is working out
how much the signalling units interfere with each other and introducing
compensating corrections based on previous units).  If the line characteristics
change significantly, it may be necessary to repeat the calibration process,
which is what is called a retrain.

Retrains ought to be initiated on the basis of measurements of the analogue
signal, but might be triggered, in a simple implementation, by an excess 
error rate in the demodulated data.  It would only be in that latter case that
anything in the software or interface to the modem would trigger a retrain,
and then would only be the result of using the wrong criteria.)