Path: utzoo!utgpu!watserv1!watmath!att!rutgers!usc!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!ira.uka.de!smurf!urlichs
From: urlichs@smurf.sub.org (Matthias Urlichs)
Newsgroups: comp.dcom.modems
Subject: Re: Data Compression
Message-ID: <%6ele2.0_1@smurf.sub.org>
Date: 16 Jul 90 21:04:51 GMT
References: <3547@hsfmsh.UUCP>
Organization: University of Karlsruhe, FRG
Lines: 29

In comp.dcom.modems, article <3547@hsfmsh.UUCP>,
  tnixon@hsfmsh.UUCP (Toby Nixon) writes:
< In article <269DE3BD.28935@ics.uci.edu>, John Roy asks:
< 
< - So if I'm sending ASCII digits would I do better with MNP5 or V.42bis?
< - I would guess MNP5, but...
< 
< No!  Remember, MNP5 compresses individual characters into, at a 
< minimum, 4 bits.  [...]
<  V.42bis typically achieves an average of about 4-to-1 
< compression on English text files.  You will always (unless you 
< purposely construct a file that doesn't compress well with V.42bis) 
< get better compression with V.42bis than with MNP5.
< 
Grossly generalized: MNP5 relies on a limited character set, and V.42bis
relies on recurring sequences of characters.

Standard ASCII text has about 6.5 bits per character, so MNP5 gives you about
15% compression. On the other hand, any given character (in English text) can
only be followed by certain others; the cumulative effect of this is a rather
good compression. Program code is even better.

So, about anything compresses better with V.42bis, including executable code,
but excluding (this is from experience) digitized sound. Pictures are a border
case, depending on whether they're dithered.

-- 
Matthias Urlichs -- urlichs@smurf.sub.org -- urlichs@smurf.ira.uka.de
Humboldtstrasse 7 - 7500 Karlsruhe 1 - FRG -- +49+721+621127(Voice)/621227(PEP)