Jules Gilbert compression demo

[Mike Attili]

On Friday June 7, 1996, I participated in a demonstration of
Jules Gilbert's compression technology (code named Bronze).
We initially made plans to conduct compression and
decompression on isolated systems using files that I had
arbitrarily selected.

I selected several sets of data, including the FSF distribtion
of dejagnu-1.2 and a library of C object files used internally.
After the OS and the compressor/decompressor were installed on a
removable hard drive, I took physical possession of the drive and
provided the demo data on floppies. From this point on, I personally
sat at the keyboard and ran the compression tests.

I used the FSF gzip as a reasonable (and widely available)
benchmark for comparison with Bronze. With the exception of files
smaller than ~100 bytes, Bronze provided, roughly, 10-20% better
compression ratios than gzip [i.e. (Orig/Bronze) ~= 1.15 * (Orig/GZip)].
For some files, the compression improvement was negligable,
for others it was much greater; on average, roughly 10-20%.
The results for text and for object files were similar.

The only characteristic under test was compression relative to
gzip. Compression time and memory requirements were not
evaluated with this version of the software. I expect substantial
improvement in both time and memory usage in a production version
of the algorithm.

Since these initial results did not meet my compression
requirements, formal testing was discontinued. I did sample
decompression and binary-differences as a quick check to confirm
that the software was working correctly. The compressed and
decompressed files were identical to the originals, for those
files tested.

In several weeks when Jules returns from other business,
I plan to repeat this test scenario with a revised version
of the compression algorithm that I expect will meet my
requirements.

Michael Attili
Head of Product Development
Software Partners/32, Inc.
Topsfield, MA 01983