Comment #4 on issue 4997 by
[hidden email]: isPrime is broken
http://code.google.com/p/pharo/issues/detail?id=4997Not sure you can say the PRNG in itself is poor on those grounds (though it
has been shown to have flaws), but it's API in Squeak/Pharo surely begs to
be misused in the manner you describe.
1) The PRNG has a period of 2^32-1, any integer above that will have
numbers which will never appear from atRandom.
2) Internally it stores the seed as a Double, so atRandom will at least
multiply the number with the smallest possible Double (ie why you see so
few variations of the low bits in the 97bit-number).
3) Assuming a perfectly bit-random algorithm of any size n, it is still
impossible to get a *perfectly* uniform distribution of any range with size
other than 2^n -1. In other words, however you divide the outcome, some
numbers will be more heavily biased than others.
4) Using it for cryptography, as suggested by nextInt: comment, is a *very
bad idea*. Both because of 1) and 2), but also because a single call to
next will expose the entire seed to an attacker. (Just multiply with m)
TLDR; The Random API and implementation is a bad fit for mostly any use
other than fast generation of numbers of a small range/value.
_______________________________________________
Pharo-bugtracker mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-bugtracker
Locked
