Subject: Re: arc4random(9)
To: None <>
From: Perry E. Metzger <>
List: tech-kern
Date: 05/29/2002 11:04:09
Thor Lancelot Simon <> writes:
> > > Well, then, let's just call it "random()".
> > 
> > Yup -- modulo the fact that I think (I may be mistaken) that Posix may
> > specify the algorithm.
> For the random-number generator in the *kernel*?

Well, admittedly in the kernel it need not be the same, but remember
there is also a userland function to consider. I think it would be
nice to have a better RNG in userland.

> Incidentally, have a look at the way the reseeding works -- it seems
> needlessly complex, and that worries me (why not a simple rekeying with
> a new random key?).  Do you agree that it ought to toss the first N bytes 
> every time it reseeds, if it should even reseed at all?

It should certainly be tossing a bunch of data after a reseed. The way
reseeds work should also be reassessed.

Perry E. Metzger
NetBSD: The right OS for your embedded design.