[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: texmf.cnf again

Manoj Srivastava <srivasta@debian.org> writes:

> On Tue, 10 Jun 2003 09:42:32 +0200, Emile van Bergen <emile-deb@evbergen.xs4all.nl> said: 
>> Hi,
>> On Tue, Jun 10, 2003 at 08:55:07AM +0200, Andreas Metzler wrote:
>>> which leaves two solutions
>>> * Try to minimize breakage by choosing a default. Results may vary
>>> depending on whether you think breaking one unix guru's sytem is
>>> more important than ten newbies. ;-)
>>> * Don't ask the question. Proposed solutions are ucf and
>>> <[🔎] E19OvQK-0004MW-Sg@mid.downhill.at.eu.org>. Another possibility
>>> might be to simply move /etc/texmf/texmf.cnf to /var if it is not
>>> supposed to be edited.
>> I've suggested it before: why not maintain the auto-generated
>> texmf.cnf in /var, and offer the administrator to symlink
>> /etc/texmf/texmf.cnf to it?
> 	Because that leaves users out in the ghetto. If a TeX package
>  needs some changes to go into the configuration (as happened with
>  Adam Di Carlo's package), the user is never asked. 

Sure, he can be: if the user has the symlink, he'll get the easy
autoupdate.  If he's removed the symlink and has his own texmf.cnf,
there should be a note in the package's README.Debian that this
package requires configuration to be useful.

This seems entirely reasonable, not "poor packaging quality" at all.

> 	It is still the either you give up control of you
>  configuration file to me, and only edit configuration files in a
>  manner I prescribe, or else you are unsupported. This may adhere to
>  the letter of the policy, but is poor packing quality, and we can do
>  better. 
>> No merging, no parsing, no complexity, no loss of information, and
>> you can choose between power and user friendliness on a per machine
>> basis.
> 	And no support.

"Here are the program files, documentation on how to configure them is
in /usr/share/doc/pgkname/README.Debian" certainly doesn't sound like
quality below Debian's standard -- is there some obviously better way
to support people with weird texmf.cnf that hasn't been mentioned yet?


Brian T. Sniffen                                        bts@alum.mit.edu

Reply to: