> Linux. After all there comes the file contents. BZip2 compresses the
> whole archive. At the beginning I thought to reach a much better
> compression rate than before but gzip is not so bad as I thought.
Bzip2 is indeed good, but being based on the Burrows-Wheeler transform it
probably performs worse than current 'state of the art' PPMD compressors. It
would
be really interesting to see the effect of your patch with a PPM compressor. A
while
ago I did a test on several different compressors on the source archive:
Packer total size compression improvement from gz loss from
best time (s)
tar 25764.5Mb 0 -2.42284
0.786226 0
gzip 7527.25Mb 0.707845 0
0.268287 41010.3
zzip 5987.13Mb 0.767621 0.204605
0.0800635 441934
szip 6501.28Mb 0.747666 0.1363
0.152816 37523.5
bzip2 6479.37Mb 0.748516 0.139211
0.149951 43998.6
PPMd2 8631.81Mb 0.664973 -0.146742 0.36192
36657.1
PPMd3 7199.65Mb 0.72056 0.0435215 0.234993
39515.3
PPMd4 6549.61Mb 0.74579 0.129879 0.159067
41225.7
PPMd5 6223.04Mb 0.758465 0.173265 0.114937
43005
PPMd6 6027.49Mb 0.766055 0.199245 0.0862218
44711.3
PPMd7 5892.44Mb 0.771297 0.217185 0.0652795
46276.5
PPMd8 5796.15Mb 0.775034 0.229978 0.0497504
47621.4
PPMd9 5731.01Mb 0.777562 0.238631 0.0389513
48880.5
PPMd10 5688.98Mb 0.779193 0.244215 0.0318506
50777.1
PPMd11 5661.65Mb 0.780254 0.247846 0.0271773
50990.6
PPMd12 5638.57Mb 0.78115 0.250912 0.0231944
51868.9
PPMd13 5625.2Mb 0.781669 0.252688 0.0208738
52697.2
PPMd14 5613.63Mb 0.782118 0.254225 0.0188553
53441
PPMd15 5609.34Mb 0.782285 0.254795 0.0181046
54141.2
PPMd16 5605.05Mb 0.782451 0.255366 0.0173525
54776.9
bzip2 = bzip2 -9
PPMdx = ppmd -o x -m 220
compression = relative compression
improvement from gz = relative improvement from gzip compression
loss from best = relative size difference compared to using different
compressors for each
package and using the individually best
compressor for each package.
time(s) = total compression time
PPM compressors have been to slow to use until recently[0]. PPMd is a
demonstration
program based on the article[0], I even got as far as creating a demo
package[1]. However, a more usable
program that contains PPMd code is 7zip[2], but being written originally for
windows I'm not sure
about the current state of the port, if there even is any now (comments Radim?).
0. http://DataCompression.info/Miscellaneous/PPMII_DCC02.pdf
1. oxtan.campus.luth.se/debian/ppmd
2. http://www.7-zip.org/
--
Magnus Ekdahl 0739-287181 [EMAIL PROTECTED] [EMAIL PROTECTED]
public key available at http://oxtan.campus.luth.se/magnus.public
ftp://ftp.se.debian.org/debian-non-US/pool/non-US/main/d/debian-keyring/
Key fingerprint = 18DE CB62 8A86 374E 824E 09ED 1987 4B18 1213 79F6
