Subject: Re: replace kernel random number function
To: None , Johan Danielsson <>
From: Richard Rauch <>
List: tech-kern
Date: 10/22/2000 21:36:45
Um, something that no one has directly mentioned:

If you were to go so far as to change random() to use rnd(), you'd be
breaking a documented characteristic, wouldn't you?  I.e., the ability to
reseed random(), as with rand(), to reliably reproduce a sequence.

``True'' randomness may sometimes be desirable (as I gather is the intent
rnd()'s entropy collection).  But, simulations, for example, need to be
able to regenerate the same sequence on-demand.  (E.g., for testing.)

random()'s man-page implies that it can generally be made to regenerate
sequences, given the same seed.  (It specifically says that you can do so
if you seed with 1.)

Am I missing something in the suggestion of using rnd entropy?  I must be,
or else (IMHO), a reminder was in order.  (^&

(Yes, I realize that the original post wasn't about this.  But the
sub-thread developed, and people seemed to not even mention/consider the
reproducability that one wants with rand(), and presumably with random().)

  "I probably don't know what I'm talking about."