In article <Pine.LNX.4.21.0201210140221.17365-100000 / bartok>, Mathieu
 Bouchard wrote:
> On Mon, 21 Jan 2002, Eric Lee Green wrote:
>> Mathieu Bouchard wrote:
>> > The problem with .zip is that it was Not Invented Here (NIH).
>> Err, no. The problem with .zip is that it was invented over five years
>> after .tar.Z became standard on Unix.
> 
> Well that maybe one problem. But there's still a significant advantage at
> using .zip, or any other similar format.

What advantages? People talk about the ability to retrieve a random file
from a zip archive. Guess what -- I have *NEVER* done that. It's always
been a case of when I unzip an archive, I unzip it, I don't pull just one
file from it. 

>> Richard Stallman threw a fit and decided to re-write 'compress' in a
>> non-patent-infringing manner.
> 
> Well, I don't know how much he was involved with it, but he didn't
> "write" any code. JL Gailly did.

Okay, so be legalistic :-). Richard Stallman certainly did throw a fit, 
and gzip's copyright is owned by his Free Software Foundation, whoever
wrote the code I wasn't interested enough to care about at the time. 

>> This was somewhere around 1991, if I recall correctly. Note that pkzip
>> arguably violates the Unisys patent since it uses basically the same
>> compression algorithm as 'arc', which is basically the same
>> compression algorithm as 'compress',
> 
> well, there is pkzip 1 ('89) and pkzip 2 ('93). Pkzip 1 did compress
> significantly better than arc; 

Any references to that? I was rummaging around and it looked like 
pkzip 1 was arc with the serial numbers filed off. It compresses
significantly FASTER than arc, but I'm not sure about BETTER. Alas, as
I no longer have a copy of arc and pkzip 1 hanging about, I can't go back
to see. 

> Pkzip 2 is significantly better than Pkzip
> 1, and extremely similar to GZip in performance (normally +/- 1%); which

gzip is a very slow compression program, BTW. It is slower than hoary
old "compress", though it compresses better. One of my co-workers
benchmarked pretty much every compression program and subroutine out
there, and found that the fastest was one called 'lzop' (though it did
not compress as well as gzip). The patent situation with that one,
however, is unknown.

> A lot of compression algorithms are regularly invented that beat gzip and
> even beat bzip2; the latter only got more attention because it was
> implemented with a gzip-like interface.

Hmm, care to mention one that compresses tighter than bzip2? When that
co-worker benchmarked compression routines he could not find one, but he
did note that bzip2 was extremely slow. 

>> and many Unix programs are now packaged as both .tar.gz and .tar.bz2
>> files, just as, during the transition from .Z to .gz in the early
>> 90's, many Unix programs were packaged as both .Z and .gz files.
> 
> I don't see bz2 replacing gzip, because bz2 compression is not streamable.

? Most people use gzip or bzip2 in the context of compressing tarballs.
Granted, bzip2 is not going to replace gzip for anything requiring speed,
but nobody really cares. It's like with pkzip. Sure, you can access random
files in a pkzip archive. But who cares? Every .zip file I've ever downloaded
for Windows, I did one thing to it -- I unzipped the whole thing and 
clicked on the "setup.exe" file. 

-- 
Eric Lee Green       eric / badtux.org    http://www.badtux.org
    GnuPG public key at http://badtux.org/eric/eric.gpg
            BadTux: Linux Penguin Gone Bad