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

Re: PROPOSAL: automatic installation and configuration

Adam Di Carlo <adam@onshore.com> writes:

> Massimo Dal Zotto <dz@cs.unitn.it> writes:
> > The advantage of my tool, or any similar solution, is that the changes
> > are very small and limited only to the installation scripts. The only
> > changes required are substituting code like this:
> > 
> >     print message
> >     print -n prompt; read answer
> > 
> > with:
> > 
> >     answer=$(dpkg-getconfig PKG_VARNAME --message message --prompt prompt)
> > 
> > The installation script doesn't need to be aware that the value is read from
> > a database, it just asks a question and gets an answer. But if the answer
> > is asked directly to the user we can't get a chance to store or read it
> > in a database, so the scripts must be changed.
> Also a disadvantage.  You don't really have any namespaces defined, or

My proposal/scripts say that PKG_VARNAME must start with the package
name and an underscore. Via that package name the configfile will be
selected and for shared configuration, you need a virtual package or a 
virtual config file.

> any heirarchy or method for shared configuration (i.e., ask for my
> news server, unless it's already be ask for before).

Its a bit tricky to automatically detect those, but I think that all
questions that look at variables outside their own configfile should
be considered shared and should only be asked when not marked answered 
already. Another option would be to introduce a commandline option to
tell the question script to ask only if not defined.

> However, I do prefer the small hack which actually implemented to the
> ideal hack which has been debated for years and never implemented.
> > If this change in policy would be approved it would take a few days or weeks
> > to modify all the installation scripts.
> [...]
> > I would really prefer to have the next freeze delayed of a few weeks rather
> > than having to make another totally manual mass installation.
> This is totally naieve. There is no way we could have a complete
> transition in 1999 -- period.  It's just not going to happen, mark my
> words.

Maybe, maybe not. But all Packages I install normally will get whishlist
bugreports with diffs. I'm realy behind the thing and will push it.

> I just hope we have some partial transition in 1999 (maybe "mostly
> complete" for post-potato).

Actually its up to the maintainers of all packages that ask questions
to adapt, once policy changes to one method or another. But they are
few questions that actually ask, so it should not take too much time
if one is determined.

May the Source be with you.

Reply to: