Wirt writes:
> While no one could begin to count the number of computers that will be
> used on tomorrow's flight, the number that were used on his first is well
> known. The total count was zero.

Of course how much guidance do you need on a cannonball :-)

There was a quite interesting article many years ago in the Journal of the
ACM that described the 5 main computers that controlled Space Shuttle
flights at that time (and knowing NASA's conservatism, probably still do).

Four redundant computers running the same software, plus a fifth backup
running software with essentially the same specification but it was
required to have been written by a different company.  These systems
were/are best described in units of KB and KHz rather than the MB/GB and
100s of MHz typical even of today's cheapest PCs.  Different programs
for different phases of flight are loaded from a tape drive under manual
operator control.  Each computer drives separate servos for the flight
control surfaces, so that if one goes berserk and starts trying to push
flight controls the wrong way, the other two systems will be able to
overpower to actions of the malfunctioning one, which eliminates the
need to combine the outputs of the multiple systems or even to accurately
detect the malfunction of one computer.

> TTL, the first truly successful small-scale integration series, didn't
> appear until about 1964-1965, and these circuits only had a few dozen
> components per chip, and yet TTL circuits prospered well into the 1980's
> (and you can still buy them).

In fact the first PA-RISC systems that shipped to customers (the HP3000
series 930 and HP9000 series 840) did not use microprocessors as their
CPUs.  These first PA-RISC processors were implemented as five boards
full of discrete AS series TTL logic chips, still fundamentally the same
as the ones Wirt describes (the 950 and later models did of course have
microprocessor implementations).

G.