Xref: utzoo sci.crypt:3547 alt.sources:2312
Path: utzoo!utgpu!cs.utexas.edu!sun-barr!decwrl!bacchus.pa.dec.com!shlump.nac.dec.com!koning.enet.dec.com!koning
From: koning@koning.enet.dec.com (Paul Koning)
Newsgroups: sci.crypt,alt.sources
Subject: Re: padrand(); /* random numbers from one-time pads */
Keywords: one-time random noise pad
Message-ID: <1990Sep14.094334@koning.enet.dec.com>
Date: 14 Sep 90 13:47:43 GMT
References: <578@hhb.UUCP>
Sender: newsdaemon@shlump.nac.dec.com
Reply-To: koning@koning.enet.dec.com (Paul Koning)
Followup-To: sci.crypt
Organization: Digital Equipment Corporation
Lines: 25

|>...
|> The C source of the padrand() routine posted here, is hereby placed in
|> the Public Domain.  A primitive driver (main) is enclosed for convenient
|> testing.  The verbal description of the algorithm immediately below, is
|> "Copyright 1990, Istvan Mohos, All Rights Reserved".
|> ...
|> Although the text of one-time pads is non-cyclic, the byte stream is
|> subject to regularities of character distribution as the function of the
|> language.  
|> ...
|> The program is somewhat wasteful of pad text, and is intended for Unix
|> environments where on-line text is abundant (as evidenced by directories
|> /usr/dict, /usr/man, ~TeX/TeXdoc and the like) but hardware random
|> number generators are absent.  

It seems to me that you have missed the one most crucial part of the
definition of "one time pad": not only must the one-time pad be non-cyclic,
but the individual bytes must be random.

When you're talking about using text files as a source of key data, you
aren't describing a one-time pad at all.  Instead, what you have is a
"book code" or "running key cypher".  Those are easy to solve; the method
for doing so dates back to the 19th century.  (See D. Kahn, "The Codebreakers")

	paul