James Fenn <jfenn / uklinux.net> wrote:
>
>On Wed, 24 Jan 2001 14:23:19 -0500 (EST)
>Mathieu Bouchard <matju / cam.org> wrote:
[...]
> > 	(b) tar archive, not gzipped, but containing only .gz files
>
>isn't that what a tar.gz is anyway?

No.

These translations are non-commutative.  If you put on
your pants before your underwear you get a different
outcome than if you put on your underwear before your
pants... :-)

To get a tar.gz file you tar, and then gzip.  What is
being talked about here is gzip and then tar.  If you
have many small similar files then taring and gzipping
will give substantially better compression.  But if
you gzip and then tar, then the outside world has a
tar file and you can extract any individual file by
itself.

Think of gzip as a type of encryption for a second (*).
If you tar and then encrypt, you canot even tell that
you have a tar file until you decrypt.  Whereas if you
encrypt then tar you can see that you have a tar file,
but you cannot read the files in it without decrypting.

[...]

Cheers,
Ben

PS Though encryption and compression are very dissimilar
   operations, there is a connection.  Perceived
   randomness of data is connected to information density.
   White noise has a density of 1.  Compression carries the
   same information in less data, so the density of
   compressed data moves towards 1.  Encryption scrambles
   data so that without all of it and a secret key you
   cannot tell that its density is not 1.

   This leads to several basic facts of information theory:
   1. White noise, compressed data, and encrypted data
      look pretty much the same.  Therefore if you need
      a large sample of fairly good random data, grab a
      large volatile data file, compress, encrypt, and
      then sample from there.  (Avoid headers.)
   2. White noise and encrypted data are not compressible.
      Don't even try.
   3. While there is no point in encrypting and then
      compressing data, if you compress and then encrypt
      you usually get a much better encryption.  For
      instance the classical Unix crypt utility can be
      trivially decrypted if you use it on straight
      text.  But if you compress and then encrypt it poses
      (somewhat) more of a challenge.  (Of course far, far
      better encryptions are widely available.  But even
      they will be better if you compress first.)
_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com