My issue was only with pre-compression.
I am beginning to wonder nowadays to what extent does compression
help defeat cryptoanalysis and to what extent it impairs it.
Arguments for the latter abound. But compression does introduce
some regularity and quite a bit more in terms of "early error
detection." (A byte which causes a scrambled tree in a LZ78
scheme or an outrageous pointer in a LZ77 scheme is a good indicator
that you need not continue with the current key! Furthermore,
the early parts of compressed data tend to be either highly correlated
to the uncompressed data or be some kind of regular structure like
the specification of a Huffman tree...)
My current thinking is to keep the two separated. Find a reliable
encryption algorithm and keep your faith in that and that alone.
(3 pass DES is my personal favorite.)
The only recent topic of interest in sci.crypt (where this compression
+ encryption topic started to the best of my knowledge) is the
possibility that MD5 based schemes may not be secure against existential
forgery! :-)
-Ray