From: Eric Hughes <hughes@soda.berkeley.edu>
Date: Mon, 26 Oct 92 09:53:38 PPE
To: cypherpunks@toad.com
Subject: entropy
In-Reply-To: <921024155350_74076.1041_DHJ67-1@CompuServe.COM>
Message-ID: <9210261653.AA13072@soda.berkeley.edu>
MIME-Version: 1.0
Content-Type: text/plain


Hal:
>Ascii encoding used by PGP reduces this to 6 bits per character
>(e.g. a character set with 64 printable characters) neglecting
>line separators and message beginnings and endings.  So there
>should be a little less than 6 bits per character for encrypted,
>Ascii-encoded messages.

Hal is, of course, right.  I was getting myself confused between
entropy lost in the encoding and the entropy of the encoding.  The
channel uses up two bits of entropy per character in the encoding.
What's left is six bits.

As punishment for getting this so egregiously wrong, I'm going to post
some C code for measuring entropy so that you all can play with it.

Eric