Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site lasspvax.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!mhuxt!houxm!mtuxo!mtunh!mtung!mtunf!ariel!vax135!cornell!lasspvax!norman
From: norman@lasspvax.UUCP (Norman Ramsey)
Newsgroups: net.math,net.micro
Subject: Data compression and information theory
Message-ID: <417@lasspvax.UUCP>
Date: Tue, 30-Jul-85 20:38:22 EDT
Article-I.D.: lasspvax.417
Posted: Tue Jul 30 20:38:22 1985
Date-Received: Thu, 1-Aug-85 22:31:00 EDT
Distribution: net
Organization: LASSP, Cornell University
Lines: 35
Xref: watmath net.math:2152 net.micro:11217

I have been thinking for some time about data compression, both for
archiving and for transmission over a serial link. Forgive me if I post
this both in net.math and net.micro, because I'm interested in both
information theoretic aspects and in software.
 
I have read somewhere that the information content of English text (such
as one might read in the New York Times) is roughly one bit per character.
I have also worked out (it is fairly simple) that a single-character 
Huffman encoding of English text requires about five bits per character.
One assumes that larger savings could be realized by using digraph or
trigraph Huffman encoding, but such schemes rapidly become unweildy for
a computer implementation (unless I am missing something).

So what can we do to get closer to that 1-bit-per-character theoretical
limit.
 
 
On a more computational note: wouldn't it be nice if we could compress
text to be sent over a serial line? If I could write a TTY driver that
would encode everything sent by a host machine to my home computer, I
could read the news many times faster than I can now (which, at 300
baud, makes quite a difference).
 
What can people tell me about data compression or encoding schemes other
than HUffman encoding?
 
What are the possibilities for implementing such a scheme for mainframe->
micro communication and speeding up throughput by a factor of four?
 
 
Norman
...allegra!cornell!lasspvax!norman
 
P.S. I am new to the net so if I've goofed send me some mail and let me 
know.