From: eli+@GS160.SP.CS.CMU.EDU
Date: Tue, 20 Feb 1996 03:07:43 +0800
To: cypherpunks@toad.com
Subject: Re: True random numbers
In-Reply-To: <+cmu.andrew.internet.cypherpunks+kl9Y:mm00UfAI100wG@andrew.cmu.edu>
Message-ID: <9602191812.AA07312@toad.com>
MIME-Version: 1.0
Content-Type: text/plain


Tim May writes:
>4. I'd avoid altogether phrases such as "generates random numbers," unless
>your method uses radioactive decay or Johnson noise measurements, for
>example, and maybe not even then.

Persi Diaconis gave a talk here last week on pseudorandom generation,
during which he was asked by people didn't use hardware RNGs.  He said
that he wasn't aware of any that passed the standard battery of
statistical tests.

(He also mentioned that nobody had thoroughly tested the
complexity-theoretic generators such as Blum-Micali, and got into a
rather vigorous discussion with a professor who argued that testing
was superfluous.  Diaconis: "You theorists always take that tone!")

-- 
. Eli Brandt                                        usual disclaimers .
. eli+@cs.cmu.edu                                  PGP key on request .
. arrest me:                       violation of 18 U.S.C. 1462: "fuck".