While composing an answer to Manoj, and before I had come to a
conclusion for myself, I wrote the parts below.  They're probably boring
now, but just in case they're going to be needed again, here they are in
the archive.


As a matter of fact, the common extension for "configuration files" in a
TeX system seems to be cfg, not cnf, but there are more, and many don't
have a special extension, but are simply in a directory config, or just
anywhere.  Map files are also a file that one might want to change.
This gives an estimate of about 300 configuration files in tetex-base
and tetex-extra, but there might be more:

[EMAIL PROTECTED]:~$ dlocate -L tetex-extra tetex-base | egrep 
texmf-tetex.*(cfg|config|map) | wc -l
296

On the other hand, there are files in config directories which on a
properly handled system should never ever be changed.  ini files are one
important example, so we're "down" to 260 files:

$ dlocate -L tetex-extra tetex-base | egrep texmf-tetex.*(cfg|config|map) | 
egrep -v ini$ | wc -l
259

So let us investigate whether upstream in fact intended these files to
be configuration files, be it on a per-document, per-user, or site-wide
basis.  The LPPL, under which many of these files are licensed, poses
some restrictions on modified/derived works, but has a special clause
that certain files can be made exempt from these restrictions, and may
be distributed in modfied form freely - e.g. to create customized
distributions.  I think if a file in a LPPL'ed package is exempt from
these restrictions, it means it is intended to be a configuration file
by the upstream author, otherwise it not.

Judging from this condition, many files with extension .cfg are not
configuration files,
e.g. /usr/share/texmf-tetex/tex/latex/microtype/*.cfg or seminar.con.
Now there are, of course, also (CTAN) packages under different licenses,
e.g. GPL, without such a distinction.  It's hard to judge which files
should be configuration files.  Look through documentation, newsgroups,
try to contact upstream,...

During sarge development, we found out that it makes some sense to
modify map files to get some particular pattern of font inclusion in PDF
documents.  Later we decided that this had bad side effects (I forgot
the details) and were glad to be able to move them back.  This might
require some complicated code changes to the debian-specific or upstream
font-handling scripts.

Furthermore, these files only change rarely.  If they do, it's often a
larger code rewrite, and cfg files must be rewritten, too - I expect
they often must be exchanged with the upstream package.

-- 
Frank Küster
Single Molecule Spectroscopy, Protein Folding @ Inst. f. Biochemie, Univ. Zürich
Debian Developer (teTeX/TeXLive)

Reply via email to