From: ichudov@algebra.com (Igor Chudov @ home)
Date: Sun, 1 Dec 1996 11:11:13 -0800 (PST)
To: ichudov@algebra.com
Subject: Another problem with IPG algorithm
In-Reply-To: <no.id>
Message-ID: <199612011844.MAA03510@manifold.algebra.com>
MIME-Version: 1.0
Content-Type: text


Don and others,

At the heart of IPG algorithm there is a pseudo-random number generator
which generates values of A(JV). (see http://www.netprivacy.com/algo.html)

       DO
         JV=JV+1
         IF JV=53 THEN JV=0
         A(JV)=(A(JV)+B(JV)) MOD C(JV)
       UNTIL A(JV)<16384

Note that if B(JV) and C(JV) in a triplet (A(JV), B(JV), C(JV)) are not
mutually prime, they will generate very few numbers and not a whole set
0-16383. For example, if C(JV) is 20000, and B(JV) is 10000, and initial
A is (for example) 57, the only two numbers that this triplet will
generate will be 57 and 10057.

This refutes Don Wood's claim that the distribution of results
approaches even. Even if only ONE triplet is such as I described (and it
is VERY likely to happen statistically), the distribution will be
skewed.

Don, what do you think about it?

igor