Path: utzoo!attcan!uunet!husc6!bbn!rochester!cornell!batcomputer!itsgw!mailrus!iuvax!pur-ee!uiucdcs!uiucdcsp!gillies
From: gillies@uiucdcsp.cs.uiuc.edu
Newsgroups: comp.sys.mac
Subject: Re: some (should-be) ground-rules f
Message-ID: <76000232@uiucdcsp>
Date: 16 Jun 88 18:46:00 GMT
References: <2758@utastro.UUCP>
Lines: 18
Nf-ID: #R:utastro.UUCP:2758:uiucdcsp:76000232:000:750
Nf-From: uiucdcsp.cs.uiuc.edu!gillies    Jun 16 13:46:00 1988


When you learn a little bit compression algorithms, you find that
almost ANY algorithm can be misimplemented to get poor performance.
And the decompressor will still work!

I can implement anti-Huffman coding to EXPAND the data AS MUCH as
possible.  The Huffman decompressor will still work.

I can implement LZW to do A RESET after every byte code.  The data
will always expand to 300% of its original size (assuming 12-bit
codes).  The LZW decompressor will still work.

Just because the data decompresses, doesn't mean it was compressed
well..  The LZW paper does not define a unique compression output.
Many possible outputs are possible, and some are VERY bad.

Don Gillies {ihnp4!uiucdcs!gillies} U of Illinois
            {gillies@cs.uiuc.edu}