Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83 (MC840302); site haring.UUCP
Path: utzoo!linus!decvax!mcvax!turing!haring!jim
From: jim@haring.UUCP
Newsgroups: net.unix-wizards
Subject: Re:  Text-compression utilities
Message-ID: <268@haring.UUCP>
Date: Sun, 15-Jul-84 08:26:17 EDT
Article-I.D.: haring.268
Posted: Sun Jul 15 08:26:17 1984
Date-Received: Sun, 15-Jul-84 10:11:03 EDT
References: <1599@sri-arpa.UUCP> <691@hou3c.UUCP>
Organization: CWI, Amsterdam
Lines: 22
Apparently-To: rnews@turing.LOCAL

I have been fiddling with the compress/uncompress programs
posted by Spencer Thomas. I found a couple of bugs and also
removed the Vax `asm's, with no appreciable performance loss.
(I've tested it on a SUN). The problem is that it is a memory
hog, and I'm still wondering how to make it run with less.
Our interest is that we use compaction on all our news batching
since we use international rates a lot.

The observed compression we get with compact/uncompact is around
33% on news traffic. The new compress program gives 57% on the
same 600K byte sample of net.unix-wizards. But the nice thing is
that it is also much faster, >5 times, and I haven't profiled it
yet.

Since we are all comparing pack/compact/compress type programs,
perhaps we should all be running tests on (approximately) the
same data, say some large manual section, chunk of source, etc?

Time, compression-ratio and memory demand can all be traded off,
but (unmodified) compact/uncompact looks a loser in any case.

Jim McKie  Centrum voor Wiskunde en Informatica, Amsterdam  ..mcvax!jim