LD> The Courier has defined 'retrain disabled' to mean reject
LD> remote request.

 DD> Disabled does mean not enabled doesn't it?

In the case of Rockwells, it's not so binary.  Have an anotated quote:
     
   %E - Enable/Disable Auto-Retrain

     This command controls whether the modem monitors the line quality and
     requests a retrain when necessary.  This feature is commonly known as
     ^^^^^^^^
     fall-back.  Under adverse line conditions, the modem requests a
                                                          ^^^^^^^^
     retrain from the other end and falls back to a lower speed in order
     to maintain a connect.

[in fact, it more often than not requests a rate shift first, though it
should resort to forcing a full retrain if there's no useful response]

     Format:        AT%En
     Parameters: n = 0 Disables auto-retrain.

[this is truly 'disabled'.  Modem will not initiate or respond to requests]

                 n = 1 Enables auto-retrain.  Modem will "fall back"
                       to lower carrier speed if     line quality passes
                       preset threshhold.

[actually, modem will _request_ fallback, then do so if the other end
agrees.  What it doesn't mention is that the modem will also respond to
requests from the _other_ end to fallback, but it will also respond to
fallforward requests (though it won't _initiate_ fallforward requests,
differing from %E2)]

                 n = 2 Enables auto-retrain.  Modem will "fall back"
                       and "fall forward" if line    conditions permit.

[meaning modem will both respond to _and initiate_ fallback/forward requests]

                 n = 3 Enables auto-retrain.  Modem will hang up on
                       reaching preset threshold     (fast hang up).

[I'm not sure about this one, earlier models didn't have this option .. ]

     Factory default: %E0

[duh!  That's from an oldish Maestro manual, BTW; I think 'twas later fixed]

     NVRAM:      Saved.
     Result Codes: OK      If parameter `n' is accepted.
                   ERROR   If parameter `n' is not accepted.
     
LD> My modem and others has defined 'retrain
LD> disabled' to mean accept remote request but not initiate
LD> request.

 DD> What's the point of that?

Finer control, or allowing control by one end of a link, can stop
'hunting', where the two modems are madly sending each other
fallback/forward requests, from differing perceptions of the current line
quality for a given bitrate.

LD> Having the Courier hang up 'in disgust' seems to be quite
LD> illogical compared to the way my modem handles it.

That sounds the same or similar to the Rockwell %E0 behaviour, above,
though I'm not sure that's what David meant exactly.  This is just a
terminology misunderstanding, methinks ..

 DD> I guess the courier feels that if a retrain is necessary
 DD> and the other end can't comply, there is no point in
 DD> continuing the connection.

Yes, and the Rockwell would do that too, if the Courier refused its
request, after trying 'x' retrains.  Not sure what 'x' is, though I don't
think it's externally adjustable in the Rockwells .. is it in the Couriers?

 DD> I can't imagine why one would want to disable retrains (in
 DD> any form) anyway...

As I've said a few times recently, disabling _fallforward_ requests in some
Rockwells (ie, using %E1 instead of %E2) can and has solved problems on bad
lines, especially _really_ bad lines, which most 'slickers have never seen.

Tap, tap .. is this thing on? :)

Cheers, Ian

--- MaltEd 1.0.b5