Huffman and Crypto

Sorry if these questions have been answered in posts I recently
deleted, but...

1) Instead of modifying Huffman coding, isn't it more sensible to
   just use arithmetic coding, which is more efficient anyway, and
   doesn't use fixed bit patterns?

2) I would assume there's a branch of cryptography devoted to studying
   the protections or lack thereof, afforded by compression schemes, both
   by themselves and in addition to other types of cryptography.  I would
   guess that Huffman and LZ-family codes have been studied a lot...right??
   But also more general studies of the weakness of (even secret) 
   compression algorithms as crypto methods?  I mean, compression methods
   would tend to use methods (like table lookup) that simpler cyphers
   have used for a long time, only without having been designed with
   crypto expertise...right??

3) Does anyone know of an easy-to-get-sources, easy-to-use arithmetic
   coding compression program?

quote me
-fnerd

> 3) Does anyone know of an easy-to-get-sources, easy-to-use arithmetic
>    coding compression program?
> 

I have a book titled "Data Compression" or something like that, which
explains Huffman and Huffman-type coding, arithmetic coding,
dictionary compressers (sliding window LZ* types), and JPEG
compression.  The book comes with source code, so if anybody else has
it and has a scanner...

As I recall, arithmetic compression works well, but is really cpu
intensive, even with a math-coprocessor.  Plus, you need to have a
rough idea of the statistical breakdown of the plaintext for
arithmetic compression to work as well as it can.  The Zimpel-Lev type
sliding window compressors are popular because they work well on most
inputs, with little or no pre-computation or statistics.

Isn't Phil Karn on this list?  I'm sure he can tell you everything you
want to know about compression :-)

/-----------------------------------\
| Karl L. Barrus                    |
| elee9sf@menudo.uh.edu             | <- preferred address
| barrus@tree.egr.uh.edu (NeXTMail) |
\-----------------------------------/