JT> I wouldn't be the least bit suprised if it is buggy firmware
JT> not being able to deal with many retrains and re-sends.

RS> Yeah, almost certain, it shouldnt actually be retraining at all.
RS> Just negotiating the session params properly in the first place
RS> and say falling back a bit speed wise if the session deteriorates.

JT> It seems very much that my modem gets "over zealos" and
JT> tries to connect at a too higher speed in the first place.

Its worse than that coz that doesnt require a full retrain and
hence the possibility of a dropout if the retrain doesnt succeed.
If it was JUST doing what you say, it should just fall back properly.

JT> The down-shift bit works (some of the time), I just wish it would
JT> connect at the lower speed in the first place, as it often drops
JT> carrier before it has had a chance to shift to the slower speed.

That just shouldnt be possible on the drop carrier stuff. If the line
speed is too high, that should just result in an increased error rate.
It drops the call coz it trys a full retrain and that cant negotiate
a viable set of session params. The negotiation phase is a bit fucked.

Thats assuming that you dont have a line which is mostly not too bad
and it periodically goes completely pear shaped, badly enough that no
modem could negotiate a usable session in that condition. That sounds
rather unlikely tho, and the way to prove that is with say another copy
of the same modem the BBS uses on that line you have a problem with.

JT> Another area that I would have liked to have a longer play with, is
JT> we had a customer bring in an ACER PC, with its own internal supplied
JT> modem, (dunno which brand but I *can* tell you they are shit).

JT> Anyway, I tested it calling my system.  Using any comms program
JT> on their modem, after the connect, all I get is a bunch of
JT> repeating characters. "-?-?-?-?-?-?-?" if I remember correctly.

Thats normally the result of a failure to negotiate the EC properly,
so you see the modem at the other end trying to negotiate the upper
level of the EC protocol, and that is visible with patterns like that.
The modem you are calling out with should recognise that stuff as being
part of the EC negotiation and not pass it out the cable to the comms prog.

JT> On my end at home, it has no connect
JT> message, just a NO CARRIER after it answers.

Coz your modem was aware that it was still trying to negotiate EC and it
never managed to see much sense on that from the other end and gave up.

Corse its quite possible your modem got seriously confused trying
to decide if it was a fax or data all as well as the EC negotiation.
There are some significant weirdnesses in that level of negotiation
of fax and data and you can get some significant warts in implementations.

JT> I tried changing a few options in the limited time
JT> that I had it in, but didn't come up with anything.

JT> Mind you, it had no trouble connecting to other BBS's

Yeah, that Banksias do have some blemishes when you have fax and data enabled.

JT> I've just come back and tested a new modem, a tossup between the hayes,
JT> USR and the Netcom.  I have the USR first, so far, one failed call to
JT> 3:632/348, and an aborted call (half way through the file) to 3:712/311.

I'd be trying an external Dynalink myself. I've installed a few of the
28.8s for various people and havent had any real problem at all with them.
They now have the 33K external but I havent tried any of those myself.

I'm not that impressed with USR myself, and when you add the price to
that, I wont be bothering with them. The Sportsters particularly have
had some massive fuckups like soldered in roms and that glitch with
the internal and EC and the Couriers are just too fucking expensive IMO.

JT> 311 doesn't bother me as much, as I used to get 3-8
JT> failed calls a month (out of 30) anyway.  I think he
JT> *may* have changed his modem in that time though.

JT> I have to call again and do some more testing with 632/348,
JT> it seems very intermittant, probably just line noise.

You shouldnt be getting those particular symptoms tho. An indifferent
line should just produce a good solid connect at a lower speed.
@EOT:

---