Gould computers

[ndd at duke.uucp]

Recently, someone posted a request about Gould computers.
I feel that the following is of general enough interest
to send to the net. We have been running a PN6005 (virtual
memory) computer for about 3 months now. The machine is
fairly fast, the UTX/32 operating system seems to be a
reasonable unixoid (SYS V + 4.1c), and the service has been
pretty good, although strictly at a board swap level. Plugging
boards into the back of the bus is a pain, because there are
wire-wrap pins sticking out of the back-plane which are
supposed to mate to sockets on the board. However, Gould
puts as few sockets on the board as is possible, which
causes alignment problems and not a few bent pins.

BUT, the real kicker is the 8-line async interface that
they provide. The original version that we received
would simply not accept data from a terminal (or line to
another machine) faster than 1200 baud (under UTX). The
input buffer quickly overflows and loses data. So Gould
brought out a new interface and driver. The new board is
supposed to have bigger input buffers, but the real
'solution' is to use DTR and DSR. The new board seems to
be a little faster, but it will not run even at 4800 baud
without using DTR/DSR for control flow, which says to me that
it really can't run at 4800 baud, period. Sure, you can
set your baud rate to 4800, but the throughput is going to
be way down, and what are you going to do if all your lines
are 4-wire, or if the machines you talk to don't do DTR/DSR?

This may not be a problem for some, or even most, sites,
but I would be sure to discuss it with Gould before making
a purchase.

Ned Danieley
Basic Arrythmia Laboratory
Duke University Medical Center
duke!dukebar!ndd