Date   : Sun, 22 Aug 1982 18:10:00-EDT
From   : Ronald G Fowler <RGF@Mit-Mc>
Subject: Problem with MODEM221

Roger,
  I've had a very similar problem with LMODEM (using various versions
of the modem program on the micro end).  I believe that what is happening
is that LMODEM is misinterpreting the NAK from the micro modem program,
thus considering the record as successfully transfered.  LMODEM then 
transmits the following sector.  Since the micro modem program missed
the overrun sector, and the new sector number being transmitted is one
greater than the one it expects, it aborts (this is to be expected, since
the protocol doesn't provide for requesting the sender to back up a record).
  I've noticed that this problem almost never occurs when MC is lightly
loaded.  Since the problem is related to the actual number of transmission
errors occuring, I suspect that a heavily loaded MC yields more errors.
This might be due to inter-byte time delays occuring when the load is
heavy; I seem to recall that the modem program allows a rather tight time
delay between bytes when receiving a record (one second as I recall).  Ob-
viously, when the load is heavy, the odds of this timeout occurring are
higher.  This would seem to be confirmed by 230K+ of files I downloaded
today: all transfers occured while MC had 4-8 users, and in all cases, the
LMODEM log showed no retranssions necessary.
  In your case, I suspect the errors might be caused by your BIOS; some
BIOS implementations will flush the internal disk buffer when the console
output routine is entered; in that case, the flush occurs after the ACK is
sent to MC, causing a delay while the CPU writes the buffer; in the mean-
time LMODEM sends the next record, which causes an overrun, since your CPU
is too busy writing to disk to receive another record. (Note that the char-
acter output routine is entered when you're in the "view" mode, or when
the modem program puts up the expected sector number).
  The bottom line here would seem to be the problem of LMODEM treating a
NAK as an ACK; I have no idea why this would happen, but all of the symtoms
seem to point to that.
  It would be helpful if anyone reading this could confirm these symtoms
(I think I remember Bill Blue as well as a couple of others describing
essentially the same problem).                 --Ron Fowler