Although it dates from four years ago, MIT's obituary for Corbató was
still interesting to reread. It couldn't bring itself to mention
Unix--only the latecomer Linux. It also peddled some mythology about
Whirlwind from the decade before timesharing.

"Whirlwind was ... a rather clunky machine. Researchers often had
trouble getting much work done on it, since they had to take turns
using it for half-hour chunks of time. (Corbató said that it had a
habit of crashing every 20 minutes or so.)"

"Clunky" perhaps refers to Whirlwind's physical size. It occupied two
stories of the Barta Building, not counting the rotating AC/DC
motor-generators in the basement. But it was not ponderous; its clean
architecture prefigured "RISC" by two decades.

Only a few favored people got "chunks" of (night) time on Whirlwind
for interactive use. In normal business hours it was run by dedicated
operators, who fed it user-submitted code on punched paper tape.
Turnaround time was often as short as an hour--including the
development of microfilm, the main output medium. Hardware crashes
were rare--much rarer than experience with vacuum-tube radios would
lead one to expect--thanks to "marginal testing", in which voltages
were ramped up and down once a day to smoke out failing tubes before
they could affect real computing. My recollection is that crashes
happened on a time scale of days, not minutes.

"Clunky" would better describe the interface of the IBM 704, which
displaced Whirlwind in about 1956. How backward the 60-year-old
uppercase-only Hollerith card technology seemed, after the humane full
Flexowriter font we had enjoyed on Whirlwind. But the 704 had the
enormous advantages of native floating-point (almost all computing was
floating-point in those days) and FORTRAN. (Damn those capital
letters!)

Doug