Re: LZ: Compression and patents
[Prev][Next][Index][Thread]
Re: LZ: Compression and patents
On Wed, 11 Sep 1996, Jon Niehof wrote:
> In regards to LZW compression vs. arithmetic:
>
> I believe that LZW is no longer patented. I think Microsoft actually uses
> it in the real-time compression interface in DOS 6 and later.
>
> Arithmetic, however, I *know* is still patented--which is why the MRCI
> doesn't use it, even though it is a better algorithm.
>
> BTW--no algorithm works very well for code. I'd say 20 percent compression
> would be very good. This means that this whole project might not have much
> in the way of usable results! However, you might prove me wrong--I'd love
> to see it. Plus, it would be a great learning experience!
I just went to my util directory, since it contains lots and lots
of small program files, and used pkzip to compress all the files
that began with F, then took an average on the compression. It
was 41%. That means they were now 41% their former size, or they
were reduced by 59%.
As we all know, pkzip is one of the champs at this, so this isn't
a fair test. Also, the average size of the files was about 20k,
which is much larger than they'll be on the 85. Larger files
compress better.
But I used to use one of those tiny huffman encoding programs
on my palmtop and I did some comparisons with pkzip then. An
older version of pkzip, of course. Probably the previous version.
Pkzip, as I remember it, was about 10% to 20% better on average.
Given all that, I think 20% reduction is probably on the low
side. Maybe not. I'm doing a lot of guessing here. I
think I'd bet lunch on an average of 25% to 30%. But I
wouldn't bet my car on it.
Barry
References: