Flags: 000000000001
From: wrs@pupthy.PRINCETON.EDU (William R. Somsky)
Subject: Re: computer follies
Date: 20 Oct 88 00:49:31 GMT
Organization: Physics Dept, Princeton Univ

In article <482@solaris.UUCP> wyle@ethz.UUCP (Mitchell Wyle) writes:
> ... Luckily, our solar-powered, diesel-backed, quantum-predicting energized
> deflecto-modulators kicked in femtoseconds BEFORE the power-failure
> (elements of the system move faster than light), ...

  That reminds me of the Sidrat 4200 machine we used to have.
(Came in a big blue box, about 1 meter x 1 meter x 2.5 meters.)
It had a tachyonic line-voltage monitor that would generated a
"power-failure-immenent" interrupt two seconds before the actual
failure to allow time for the machine to bring itself down gently.

  One day, one of our system programmers got the idea to make the
power-interrupt interrupt-handler REALLY rock-solid.  He designed
everything into it he could think of to protect the system:  core
dump for later retrival, full disk backup onto holographic media,
etc., etc., etc.  He even included a final disconnect_power_mains
routine ("to protect the power supply from line transients").

  Since the machine had full tensor architecture, this could all
easily be done in the alloted two seconds warning time.  In fact,
it took just 0.978 seconds.  So, a full second before the power-
failure, the computer was backed-up, shut-down and POWERED-DOWN.

  Of course, the line-voltage monitor detected this power shutdown
two seconds before it happened and initiated a power-failure interrupt
THREE seconds before the real power failure.  But the power shutdown
>From that one was detected as well, and ...

  The net result was that the machine shut itself off a full hour
before it was even turned on!

------------------------------------------------------------------------
William R. Somsky                          Physics Dept ; Princeton Univ
wrs@pupthy.Princeton.EDU                 PO Box 708 ; Princeton NJ 08544