Re: FINAL QUESTION: your natlangs. Sorry this is the last of the survey...

-----Original Message-----
From: John Cowan <cowan@...>
To: Multiple recipients of list CONLANG <CONLANG@...>
Date: Wednesday, October 07, 1998 3:22 PM
Subject: Re: FINAL QUESTION: your natlangs. Sorry this is the last of the
survey...


>Tommie Powell wrote:
>
>> In 1962, some mathematician at NSA "proved" that a "zero-entropy"
language
>> would be theoretically possible.  In other words, if you had such a
>> language, then, no matter how you jumbled up a sentence's symbols, the
>> result would be some other sentence that made sense.
>
>Actually, the proof's trivial.  Here's a sketch:
>
>1) Start with a generator that generates all possible English-oid
>random strings (those using only a-z and space with terminal
>period, question mark, or exclamation mark) in increasing order of
>length.  You may want to chop off this process when you get beyond
>sentences that take 100 years to say.
>
>2) Apply a filter to this that passes all and only those outputs
>that are actually sensible English sentences.  This assumes that
>there really is an effective procedure for detecting sense in an
>English sentence, AFAIK an unresolved question but one which intuitively
>seems reasonable.
>
>3) Assign each of these a sequentially increasing integer.
>
>The result is zero-entropy, at least at the sentence level: there are
>a countable infinity of grammatical and sensible English sentences,
>and scrambling the digits of an integer will always give another
>integer.
>
>Once you want to make discourse, however, there will be at least some
>return of entropy:  "Passengers will please refrain from using
>restrooms while the train is in the station" is not really likely
>to be followed by "Darling, I love you!"  However, that degree
>of entropy is probably insufficient for cryptanalysis.
>
>> There's no such thing as an unbreakable code [...].
>
>Well, there is one-time encryption, which depends on destroying the
>*usable* entropy by mixing in 1 bit of truly random 1key for every bit
>of message.
You're right -- and the teletype machines that we spies used for encryption
back in the 1960's employed that principle -- but "one-time encryption"
isn't true encryption (because a truly random key cannot be determined by a
code).  Because random keys could not be determined by a code, their
selection had to be a built-in feature of the encryption machine -- and that
meant that, if a single such machine fell into the hands of the enemy, the
enemy would be able to easily read all messages sent on all such machines of
the same model (and, in the 1960's, all U.S. military intelligence units
used the same model of such a machine, so that was a very serious problem).

In the 1980's, we were still relying on such machines when Oliver North gave
some to the Contras -- who then sold some of them to the enemy!  For that,
IMO, he should have been tried for treason and shot.

-- Tommie