Re: Encripted huffman-like compression

[Peter Wayner <pcw@access.digex.net>]

I wrote something on this in Cryptologia several
years back. I believe it is the April Issue of 
1988. 

It describes how to scramble the tree of the Huffman
compression to achieve more cryptographically useful
compression. Why is this necessary? Because people
often assume that compression removes many of the
redundancies of the language. Well, it only does this
in a theoretical sense. The patterns are still there.
If the Huffman encoding maps "T" to "01", "H" to "1001"
and "E" to "11", then the pattern "01100111" is going
to be very common in English text, but "10010111" is
going to much less common. 

-Peter Wayner