Date   : Tue, 17 Jan 1984 01:15:00 EST
From   : Robert L. Plouffe <PLOUFF@mit-mc>
Subject: mdm716/mdm717 bugs?

1. "Mea Culpa" on stack imbalance at label MONIN:  Thanks to
Dick Mead for finding that one.  I don't know how that got in there.

2. Data capture buffer is NOT affected by file transfer buffer size.
It uses all of memory up to the CCP or optionally can even overwrite
the CCP.  The 2k size recommendation has to do with the slowness of
floppies.  16k of buffering does not buy anything with respect to file
transfer speed.  Try it both ways and you will see for yourself.

3. The reason for SECTOR_RESEND_ON_VALID_NAK only has to do with
not upsetting the protocol when a file transfer is being made FROM
a remote that is under BYE and the 'Q' switch is not set so that full
progress reporting is seen at both ends. Thus is the reason for killing
the ACKNAK flag and has nothing to do with ARPANET.  I realize that
this is slightly different from the original CHRISTENSEN protocol,
but it is completely compatible with it.  So in MDM716, repeats are
done only if a valid NAK is received or upon a 10 sec timeout.

4.  Sorry for delay in responding to all of the discussion of the
past week or so, but I was out of the Country for 10 days and just
returned.

5. In view of the above discussion, and if you use MDM717 as released
by HOFF, I recommend the you modify the source code to give a 2k
file transfer buffer AND that you set the ACKNAK flag to permit
resending of sectors upon receipt of valid NAK only (or timeout)..