Bug#162120: Support #162120
On Tue, 2003-07-08 at 17:47, Manoj Srivastava wrote:
> This, then, has all the negatives that ajt seems to think that
> not regenerating configuration files has. And would not work in the
> corner case where mere presence of a file changes program behaviour
> (though I do not know, off hand, if there are any instances of this
> behaviour out there -- /etc/nologin does not have the problem, since
> no one would rationally want to recreate that)
>
> So now we have an inconsistent way in which the project
> handles user changes to conffiles -- dpkg and ucf try to ensure that
> user5 deletions are honoured, and otrher packages explicitly do not
> honour these changes, by recreating the files -- even though these
> packages can't ensure that the configuration file present on the
> system always has relevant data, since since people can just
> echo '' > /rtc/file
>
> So, instead of creating useful defaults, these programs would
> rather that we change policy, even though changing policy does not
> actually obviate the need for the defaults -- since one can just
> empty the file rather than removing it.
True.
> > A counter-proposal from Manoj Srivastava was (explicitly?) to forbid
> > the regeneration of configuration files that have been deleted, but
> > this was shot down by AJT in no uncertain terms.
>
> It was? I only saw ajt coming in foaming at the mouth and
> spewing expletives, and have some of his ranting countered by me, but
> obviously your mileage varied.
I suggest that we proceed in two stages. First, fill the
lacuna in policy wording and describe what actually happens.
This was the original proposal. It should not be too controversial
because it can only help users.
Second, propose a change to policy such that it explicitly
forbid the recreation of configuration files that the admin has
deleted. This idea was rejected by AJT before and presumably
will be again. However, you might be able to get the proposal
accepted.
--
Thomas Hood
Reply to: