🚀 go-pugleaf

I remember waiting for a job to run on the Bell Labs Murray Hill GCOS
system (A Honeywell 6000 machine) back in 1973.  One of the other users
hanging around at the center said something like the following:

" You think you have it bad.  These days, our turnaround time is rarely
  worse than three hours.  Last year, it was six hours or worse.

I asked what the center had done to improve things.

" Nothing.  Well, not quite nothing.  They announced that the turnaround
  time would be halved by changed operational procedures, and that did
  the job.

This didn't make any sense to me.  How could the mere announcement by
high muckymucks at the computer center change the time it took for a job
to run?

" Simple, if you expect to wait six hours for a job to run, you'll spend
  that six hours trying to guess what errors might be in your program,
  and submit a bunch of other jobs with changes made in anticipation of
  those errors.  Everyone knows that, shotgun debugging.

" The people at the center guessed that if they announced the changed
  system performance, people would submit fewer jobs in anticipation of
  getting their results sooner, and it turned out they were right!
  People submitted enough fewer jobs that there was a real performance
  improvement.

Almost 20 years later, I still find this story to be wonderful.  Those
old systems may have been clunkers by today's standards, but the people
who ran them were good!

And yes, all the cards I punched at Bell Labs were punched on O29 punches.

				Doug Jones
				jo...@herky.cs.uiowa.edu