Message-Id: <199611140446.RAA03064@papaioea.manawatu.gen.nz>
Comments: Authenticated sender is <malcolm AT mail DOT manawatu DOT gen DOT nz>
From: "Malcolm Taylor" <malcolm AT manawatu DOT gen DOT nz>
Organization: Grafik Software
To: djgpp AT delorie DOT com
Date: Thu, 14 Nov 1996 17:45:08 +1200
Subject: Re: Smaller distribution archives
Reply-to: malcolm AT manawatu DOT gen DOT nz

> Michael F Brenner <mfb AT mbunix DOT mitre DOT org> wrote:
> : "Malcolm Taylor" <malcolm AT manawatu DOT gen DOT nz> wrote:
> : > What should I call it? [the new free archiver that is 2.5 times as
> : > compressed as PKZIP?
> 
> Well, odds are that someone has already patented bits that are
> necessary for basically any compressor. The ingenious thing behind
> Info-ZIP was that they wrote a compatible code that doesn't touch the
> patents PK had applied for and is very portable (and stable, I already
> had PKZIP files that pkunzip didn't decompress correctly but unzip
> did).

Patents are very specific. I'm certain that nothing in this 
compression algorithm is patented. Most of the patents refer to very 
fast implementations of the LZ77 algorithm like LZO (this ones free) 
or LZRW (there are patents on at least one of the LZRWs), and the 
very fast implementations of binary arithmetic coding.

> Any `new-generation' compression is likely to use arithmetic coding,
> which is covered by loads of patents (which is not a problem for
> commercial programs since they can just license, but a big problem for
> freeware (as the GIF example should show)). 

As I said above most of the arith coding patents are on the fast 
_binary_ coders. Mash uses a multi-symbol coder that is not as fast 
as the Q coder, but is suited to the algorithm (a binary coder would 
be useless for this algorithm). This arith coder is free and free of 
patents - it's the CACM coder.

> Just adding a solid
> archive option to a LZH type compression algorithm may improve the
> compression quite a bit, but also makes the archives more difficult to
> handle since they cannot be appended or updated without decompressing
> and compressing the complete archive 

A problem, but when you take note of the increase in compression 
(sometimes upto 1/2 the original compressed size) I think it's worth 
it.

> (and .tar.gz is already
> established, at least in the unix, so nobody will want to switch).

And this doesn't allow for updating without recompression. The 
greatest use for the archiver would be in distribution or storage 
archives, neither have a large call for updating.

Malcolm