Bug#408529: ucf: Better documentation: Not a replacement for debconf handling
On Sat, 4 Aug 2007 22:57:37 +0200, Raphael Hertzog <hertzog@debian.org>
said:
Tsk, top posting.
> Hi Manoj, you never replied to Frank's question below. I would like to
> know what needs to be fixed/improved in the developers-reference. The
> issue of handling configuration file is one of policy and until we
> write a proper "Best packaging practice" paragraph on the topic, I
> don't think there much to be done in the devel-ref.
Well, sure, and policy is pretty clear on what is _required_,
think. Like, all configuration files live in /etc, preserve user
changes to configuration files, preserve configuration files on delete,
and remove them on purge, and so on.
> Furthermore, I would expect the ucf documentation to contain warnings
> much like the debconf man page warns you about improper usage of the
> tool in debconf-devel(7).
debconf is a complex framework, and ucf is a simple tool to use
to prompt users on options for a narrow, well defined situation. It
does only two things, really, provide users with information required
to make their decision, and on a decision, either cp a file over to a
known destination or not.
The bug in question was not improper use of ucf; ucf was quite
correctly asking the question the packager wanted. Asking the question
itself was wrong; but that has nothing to do with ucf (had they asked
the question using debconf directly it would still be equally right or
wrong.
I do not think these things belong in an ucf man page, as would
warning in rm's man page about not removing /var/lib/dpkg/.
> Cheers,
> On Sun, 25 Feb 2007, Frank Küster wrote:
>> Manoj Srivastava <srivasta@acm.org> wrote:
>>
>> > severity 408529 wishlist
>>
>> Severity set to `wishlist' from `wishlist'
>>
>> Am I correct in reading into this some criticism of the way I
>> reported this bug? Can you be more explicit, please?
No, I was hurried, and opted to make sure that the resulting bug
was wishlist, rather than checking to see current state. Consider it
defensive bts usage.
>> > reassign 408529 developers-reference
>> And is there a specific reason why you did not remove the "patch" tag
>> upon reassigning?
At this point, no specific reason comes to mind.
>> > Don't we already have a policy directive to not overwrite
>> > user changes? And does not the dev ref say packages should not ask
>> > any more questions than needed?
>>
>> Of course, but that's not what this bug report is about.
>>
>> > Creating a new file using debconf, and using ucf to ask the
>> > user whether or not to use that, is poor development practice, and
>> > should be so mentioned in the dev ref about proper handling of
>> > configuration files. One should not clutter up ucf documentation
>> > with a treatise on proper practice.
>> I still think that ucf's manpage is a proper place to document this,
>> in order to try to prevent ucf's frequent misuse.
There are many ways to misuse things. Just like we do not say
that rm can be misused in gazillions of ways, ucf does not either.
>> If you think this should rather be done in the developers reference
>> (which currently does not mention ucf, let alone advertise its use),
>> could you go a bit more into detail where you think the right place
>> would be, and what the important bits of information would be?
I am not a dev-ref author, nor do i read it all that often, so I
have no particular opinion on this.
manoj
--
You'll be called to a post requiring ability in handling groups of
people.
Manoj Srivastava <srivasta@debian.org> <http://www.debian.org/~srivasta/>
1024D/BF24424C print 4966 F272 D093 B493 410B 924B 21BA DABB BF24 424C
Reply to: