
Marcello, Perhaps compressing the entire disk and using an on-the-fly decryption system (like Knoppix uses) would help. -brandon Marcello Perathoner wrote:
Michael Hart wrote:
I have some supercomputer time reserved for testing various compressions, so if anyone wants to send me any DVDs or their images, I can try to see how much we can pack onto a single $1 DVD.
That's nice but it won't help much.
The biggest savings we can get is by:
1. using bzip2 instead of zip
2. compressing more than one file at a time
Ad 1.
bzip2 is quite well known on linux systems. I don't know how well Windoze supports that. It may be a question of $$$ to Windoze users to get a decompressor.
Ad 2.
Compressing more than one file at a time makes for a smaller archives because a compressor always starts with a low compression rate and builds itself up along the way. The first KBs of a file have the worst compression rate.
Also, the compression rate will drop drastically every time the characteristic of the uncompressed data changes. If you tar a whole lot of text and image files together, you get a better compression rate if you put all txt files first and all image files last instead of interleaving them.
Arguably we can get the best compression rate if we tar all files of each file type together (all TXT files, all HTML files etc.) and bzip2 the tar files. But these huge files will be irksome to unpack.
Alternatively we could pack the files in chunks of 100 books, carefully ordering the files in the tar, so all TXT files come first, then all HTML files etc.
If we want to avoid sensible tools and formats for the sake of Windoze users we can get almost the same results by first zipping a lot of files with compression off and then zipping the zip file with compression on.