Richard Town wrote the following to Craig Ford, and I quote (in part):

-=> Note: Copied from USR_MODEMS by WIMM/2 1.31

 -=> Quoting Craig Ford to Richard Town <=-

   CF> CF> Just hold on Richard, you'll eat crow again.
   RT>  Symbol  Expected data   Preemphasis carrier
   RT>  rate    rate
   RT>  2400    1001            0011        0
   RT>  2743    1010            0011        0
   RT>  2800    1010            0011        0
   RT>  3000    1011            0100        0
   RT>  3200    1100            0100        0
   RT>  3429    1100            0100        0
 CF> You're on the right path, now keep looking. It is staring you in the
 CF> face but you can't see it.

 RT> USR chooses the 3200 symbol rate when answering a call from a modem
 RT> that doesn't support the 2743 symbol rate and that indicates the
 RT> same maximum projected data rate for both the 3200 and 3429 symbol
 RT> rates.  This does not match what other V.34 implementations do.  By
 RT> choosing the 3200 symbol rate instead of the 3429 symbol rate USRs
 RT> are giving up about 2dB of SNR.  On a lot of lines that is the same
 RT> as giving up one data rate. 

Before you quote Dan, you need to find out what he's been up to lately.

1. Why do you project a data rate for a symbol rate you don't support?
2. Why do you weight the 3200 and 3429 symbol rates equally, if 3429 is what
   you deire to run at?   
3. Whose V.34 implementation are you refering to, that doesn't do this.
4. Why was't there any interop testing done _before_ you released this?
 
    RT> 'Cept that it don't work, coz your favourite modem is deficient 

 CF> It isn't part of any ITU recommendation.

A modem which supports all of the recommendation is deficient, and one that
doesn't is? Take your blinders off Richard.

 RT> Neither is HST nor V32terbo.  In thruput terms, given that on good
 RT> links MNP/V42bis outperformed pre-SREJ LAP-M, this was yet another
 RT> method to deliberately strike a performance difference between
 RT> budget modems and USRs by design.

There is _no_ difference between LAP-M and MNP4 with equivalent data frame
sizes. Try again when you actually _know_ what you're talking about.

 RT> It's called Modem Apartheid.  ie USRs are deliberately configured
 RT> to recognise the colour of callers' money spent on modem in
 RT> deciding what performance to achieve.

It is called _fully_ participating in the design and development of
internationally agreed upon procedures instead of pursuing alternatives
with the intent of confusing the unknowing. 

There is a concept know as "accepted practice", when it comes to
implementing ITU-T protocols. Everything isn't spelled out in black and
white, and it takes a bit of testing with other implementors to arrive at
common implementation strategies. That is what interop testing is all
about. Perhaps you might convince the manufactuer of the datapump you use
to do so before it embarases you further.

Regards....

Craig
aka: cford{at}ix.netcom.com

--- timEd/2 1.10+