From vojta@math.berkeley.edu  Fri Mar  5 02:22:56 1999
From: vojta@math.berkeley.edu (Paul Vojta)
Date: Thu, 4 Mar 1999 18:22:56 -0800
Subject: TeX and software packaging
Message-ID: <199903050222.SAA00121@bosco.berkeley.edu.berkeley.edu>

On Wed, 24 Feb 1999 07:25:52 -0700 (MST), 
"Nelson H. F. Beebe" <beebe@math.utah.edu> wrote:

> It is a huge pain for me to complete an installation of the new
> LaTeX-2e release each half year, because I have to compare the old and
> new trees to figure out which additional packages have been unpacked
> (at user demand) during the previous six months, and then manually run
> LaTeX on each associated .ins file (and finding that file can require
> grepping the package directories to locate which one produced a
> particular installed file), and manually follow instructions like ...

I deal with something that may be very similar, whenever I upgrade our
site to a new version of teTeX.

Over the years, I have observed that computers are very good at
eliminating tedium (as well as creating it ...).  Therefore, I have
devised a solution that works at my site.

I wrote a program, called fpkg, that maintains a database of all the
files within a given directory.  When run without arguments, it checks
the database against the actual contents of the directory and prints
out the discrepancies.  The database consists of a list of files, plus
a package name associated to each file.  A directory may be treated as
a file if you don't want fpkg to descend that directory when comparing
against the database.

Then, to upgrade teTeX, I install teTeX in a new directory, run fpkg to
get a list of files, register those files under package tetex, and then
move over the additional packages from the old directory, keeping the
fpkg database up-to-date.

I feel that capabilities of this sort should be included in any TeX
packaging arrangement.

--Paul Vojta, vojta@math.berkeley.edu