<< Previous Message Main Index Next Message >>
<< Previous Message in Thread This Month Next Message in Thread >>
Date   : Wed, 06 Apr 1983 19:59:00 EST
From   : Keith Petersen <W8SDZ@mit-mc.arpa>
Subject: Retry...

Thanks for the report on TOPS-20 and MODEM7xx, Doug.  This problem has
been addressed, and MDM706 (due out in a few days) will have code to
allow the user to set the program to accept only ACK and NAK and
ignore everything else when it is sending.  This will not only fix the
TOPS-20 (and other mainframes) problem, but should also help those who
have had problems with BBN's short-sighted use of a 64-character
buffer in TAC's.  We're hoping they may expand that someday, but in
the meantime when the TAC's buffer gets full, it sends out an 87h
(bell with high-order bit turned on), which causes MODEM7xx to resend
the sector (thinking it was a "not ACK").

XMODEM74 is also due out in a few days and will have the same option.
I've been testing it on my system and even on micro-to-micro transfers
it does offer improved operation.

Consider this possibility:  you, the "sender" have just sent a
128-byte sector and are now waiting for the ACK from the receiving
end.  Suppose there is some noise on the line during the time you are
waiting for the ACK.  If you take "not ACK" as same as NAK (the way
the old program does), you immediately start re-sending the sector.
Assuming that the receiving end had sent you an ACK, this is an extra
repeat of a sector he already got (which he will of course ignore
since he already has it).  The penalty is the time it takes to send
that sector.  Suppose the same thing happens at the end of that
sector.  If you have a noisy line, it's possible to resend that same
sector MANY times, even though the receiving end has it already and
doesn't want it again!

In my opinion, all versions of Ward Christensen's MODEM2 program
(which includes the various MODEM7xx and XMODEM programs) should be
made to require ACK or NAK after sending a sector and everything else
received while waiting for this should be ignored.

Hopefully the mainframe versions of MODEM will be changed to this new
approach and it may make life a lot easier for us all.  It's been VERY
sucessful on my RCPM's XMODEM.

<< Previous Message Main Index Next Message >>
<< Previous Message in Thread This Month Next Message in Thread >>