				10 Jul 1994

DISCLAIMER:
----------

   Your results could be quite different from mine due to factors
such as disk speed, disk fragmentation, disk caching, RAM size and
TSRs loaded.



COMMENT 1:
---------

   The sets of data used in these tests were designed to simulate "real"
situations although YOUR data is the only true test.  Since many archives
contain only certain types of files (e.g., audio, binary data, executables,
ASCII text), most of the sets contain a collection of a single type of file.

   Due to numerous requests, the Calgary Corpus is included (with addtional
"per file" compression data) although it is not considered a good test for
archivers.



RESULTS:
-------

   The following nine sets of tests were run on a DFI 486/33 EISA with a
Fujitsu 500M drive (disk caching disabled except as noted) with 8 Meg of RAM
running DOS 6.2.  No TSRs were loaded.  Times are rounded to the nearest
second.

   The total number of individual tests performed exceeded 2000, consuming
slightly under 14 days (~1,200,000 seconds) of execution time.  Any test whose
execution time was under 30 seconds was run multiple times to try to increase
the accuracy of the report.

   Each test set contained approximately 230 instances of archivers, file
compressors and storage tools.  Most of the programs had all of their options
tested (except where the number of permutations would have been excessive or
where different command options caused small differences).

   Set 1:     42 files, ~2.28 Mbytes, mixture of text and binary files

   Set 2:     88 files, ~0.71 Mbytes, small ".bas" and ".pas" files

   Set 3:     11 files, ~1.05 Mbytes, ASCII word processing output

   Set 4:     17 files, ~1.19 Mbytes, various graphics formats

   Set 5:     37 files, ~4.12 Mbytes, sound files (.voc and .wav)

   Set 6:     18 files, ~3.25 Mbytes, Calgary corpus

   Set 7a: 1,639 files, ~4.36 Mbytes, batch files used in benchmark testing
       7b: 1,639 files, ~4.36 Mbytes, as above (SMARTDRV caching enabled)

   Set 8:     10 files, ~1,87 Mbytes, FORTRAN unformatted binary files


   Tests 1-6 and 8 included both compression and extraction tests.  Due to
excessive time consumption, tests 7a and 7b included only compression
tests.



Archivers tested (CAPITALS INDICATE NEW SINCE LAST SUMMARY):
-----------------------------------------------------------

	AMGC 2.2          arc 6.02        arj 2.30        arj 2.39a
	ARJ 2.41          dwc a5.01       ha 0.98         HA 0.982
        hap_pah 3.0       hpack 0.75      hpack 0.78      HPACK 0.79
        hyper 2.5         lha 2.13        lha 2.55b       limit 0.4
	LIMIT 1.0         pak 2.51        pkpak 3.61      pkzip 0.92 *
	pkzip 1.1 *       pkzip 2.04g **  RAR 1.40.2b     RAR 1.51
	sqz 1.08.2        sqz 1.08.3      UC2 Rel 1       UC2 Rel 2
	ZIP 2.0 ***       ZIP386 2.0 ***  zoo 2.00        zoo 2.10

	* = unzipped with pkzip 1.1's pkunzip
	** = unzipped with pkzip 2.04g's pkunzip
	*** = unzipped with Info-Zip's unzip



Compressors tested (CAPITALS INDICATE NEW SINCE LAST SUMMARY):
-------------------------------------------------------------

	compress          diet 1.44       FREZ 2.50       gzip 1.0.7
	GZIP386 1.0.7     GZIP 1.2.4      GZIP386 1.2.4   huf
	NEW-LZW *         scrnch          splint          sqpc 1.31
        vocpack 1.0

	* = New compressor based on LZW from recent ACM publication.



File storage tools tested (also archivers w/"store" option):
-----------------------------------------------------------

	arj 2.30          arj 2.39a       ARJ 2.41        cpio
	ha 0.98           HA 0.982        HPACK 0.79      LIMIT 0.4
	LIMIT 1.0         pkzip 2.04g     RAR 1.40.2b     RAR 1.51
	tar               ZIP 2.0         ZIP386



COMMENT 2:
---------

   These are tests of archivers, file compressors and file storage utilities
that *usually* work with any type of file.  I did not attempt to examine the
various executable-only file compressors (e.g., LZEXE, PKLITE), but I did
include DIET since it can be used as a generic file compressor.



REQUEST:
-------

  If you know of any archivers/compressors that I can readily add to my
test suite, please contact me at one the following address.  I would also
like suggestions on ways in which my tests could be improved and on file
types that should be included in a test set.

===============================================================================
Greg Flint    | Purdue University Computing Center | afc@klaatu.cc.purdue.edu
Systems Pgmr, | 1408 Mathematical Science Building | flint@purccvm.bitnet
Supercomputer | West Lafayette, IN 47907-1408      | purdue!klaatu.cc!afc
  Group       | PHONE: 317-494-1787                | FAX: 317-494-0566
