Date: Wed, 5 Aug 92 14:19:57 EDT
From: DJ Delorie <dj AT ctron DOT com>
To: ericb AT lsid DOT hp DOT com
Cc: lwj AT cs DOT kun DOT nl, djgpp AT sun DOT soe DOT clarkson DOT edu
Subject: compressors

>Unless you distribute unzip with djgpp, you are not using unzip in a
>commercial product.  So I don't believe the above copyright should be
>a problem.

The problem is that I ship djgpp on floppies to some folks, and must
include the decompressor.  Other companies do the same.  I currently
do not ship a decompressor.

>As an alternative to zip or zoo, you could use something like tar or
>cpio to make archives, and use compress or freeze to compress the
>archives.  You'll probably end up with better compression in the
>process, since you're compressing the whole thing at once.

Without a special program (djtarx?) you can't extract a single file
from a .tar.Z file, since two programs are required.  Maybe adding compression
to djtarx is a better solution?

>1. I agree that a decent archiver should handle empty directories, and
>   zip and zoo currently do not.  I guess that's one reason to switch
>   to cpio or tar.  Are there any empty directories in the djgpp
>   distribution?

The objs and p_objs directories in the libsrc directory.  The
makefiles currently build them, but they should be produced at the
user's site on installation.

>2. I think an archiver should store exactly what you give it.  If you
>   want to convert the format of a text file, use a program that is
>   made to do that.  Thus, I don't think text vs. binary should be an
>   issue for an archiver.

Converting MS-DOS text files to Unix format saves space.

>3. Arbitrary comments and descriptions can be placed in a file, and
>   the file placed in the archive.  Why should an archiver have to
>   do any more than this?

The comments tell you *how* to unpack, or what version is in there, or
any information you might need *before* you unpack, like which
directory to be in when you do so.

>4. Rename as files are added.  I guess I don't see why you'd want to
>   do this, but I agree that existing archivers can't do it.  Most
>   will allow an individual file to be extracted to standard output,
>   so you can send it to whatever file you want, but that would be
>   tedious to do for many files.

The point was to rename the *directories* so that you can build the
archive from one structure, and have it unpack in another.

>5. Sounds like you want to split an archive up on multiple floppies,
>   without having to put it together on a hard disk first.  This would
>   be nice, and is not available from most archivers.  But I do okay
>   with split to split up a big archive, and cat to put it together.

This only works if you have enough hard drive, which some users may
not.

DJ
dj AT ctron DOT com
Life is a banana.