i wrote the generator.
dmr or rhm wrote the comment.
it came about after one of the first
drafts of a graphical pool game.
the balls were points and the test
was the bouncing off the edge of
the pool table. the balls were placed
at "random" places on the table,
they were started with "random"
directions and "random" velocities.
frictionless it ran forever.

after many minutes, from a mess
of dots, they form a line, later a couple
lines, later several points, and finally
after a large fraction of an hour, all the
balls would converge on a single dot.

that version of the program was saved
with the name "wierd" (spelling on purpose).
i have no idea if it exists now.


On Tue, Mar 12, 2024 at 7:38 AM Douglas McIlroy <douglas.mcilroy@dartmouth.edu> wrote:
 > The author of this routine has been writing 
>   random-number generators for many years and has 
>   never been known to write one that worked.

It sounds like Ken to me. Although everybody had his
own favorite congruential random number generator,
some worse than others, I believe it was Ken who put
one in the math library.

The very fact that rand existed, regardless of its quality,
enabled a lovely exploit. When Ken pioneered password
cracking by trying every word in word lists at hand, one
of the password files he found plenty of hits in came from 
Berkeley. He told them and they responded by assigning
random passwords to everybody. That was a memorable
error. Guessing that the passwords were generated by
a simple encoding of the output of rand, Ken promptly 
broke 100% of the newly "hardened" password file. 

Doug