Subject: Re: arc4random(9)
To: Jason R Thorpe <>
From: Perry E. Metzger <>
List: tech-kern
Date: 05/28/2002 20:25:51
Jason R Thorpe <> writes:
> On Wed, May 29, 2002 at 07:20:29AM +0900, wrote:
>  > 	or, call this wrapper random() and retire truely-crappy random()
>  > 	in the kernel.
> We could do that, although being able to predict the output of random()
> is necessary in some testing/research scenarios.

By definition, rc4's output is predictable if you know the seed. It is
just a PRNG -- in my opinion, though, a much better one than random()
although not quite up to strong cryptographic use.

I'm a big fan of RC4 for things like monte carlo work and such -- much
better results than Linear Congruential produces. I wouldn't generate
RSA keys using RC4 but that's a totally different story.

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