Re: Call for standardised package installation procedure

On Tue, Apr 08, 2003 at 11:19:42PM +0900, Gavin Sinclair wrote:
> The OS package management system will never serve all needs, so if
> "having some installed this way and some installed that way" is your
> concern, then it will never be the solution.

I'd appreciate if you gave one need impossible to satisfy with the
scheme I proposed in my previous message and easily handled by the
"pure-Ruby solution". I believe the "Pure-Ruby" solution is not
necessarily more flexible or powerful, as we can always try to add
whatever is needed upstream (that is, integrated into the dists,
by going through the effort of becoming a member or contacting one
official developer).
 
> Packages that don't support any kind of automated install are
> helpless, and are probably minor/experimental anyway.  It should be
> conceptually easy for any package to support auto-installation.

Such packages would still be in RAA.succ but not in the pkg-mgt system.
 
> That leaves a language-specific pkg-mgt system.  The lowest common
> denominator (except for "unsupported packages" which I think are an
> unimportant minority), it can serve any RAA package, and is available
> to everybody.  Keeping a repository would please me very much.  I
> can never remember what I have installed, and would like to be able to
> ask.
> 
> Pure-Ruby solution with repository: lowest common denominator,
> greatest possible features.

The Pure-Ruby solution would be only one of the backends in the
architecture I proposed. It would essentially involve leveraging
raa-install and rpkg...
 
> > At least, until such time as the Ruby package management system was so
> > widely deployed that all RAA modules used it, *and* all the O/S package
> > management systems were just wrappers around it.
> 
> The first caveat is good, but we have to start somewhere.  The second?
> OS pkg-mgt systems are irrelevant in the scheme of things, even in the
> short term.  Apart from being conceptual misfits (as I argued
> elsewhere), and insufficient for a full solution (above), consider
> this.  For Ruby package management to truly work (irrespective of
> implementation), some standards need to evolve regarding metadata, to
                        ===========================================
Cannot overstate that! The metadata will be useful regardless of the
implementation, and could be transformed if needed in the future.

The first thing to do would be identifying what things are needed when
installing, and if possible defining a templating system (DESTDIR, ruby
executable, etc)

> allow versioning, dependencies, etc.  Also, such an important piece of
> software needs technical input from the entire Ruby community.  These
> conditions cannot be met if everyone is concerned with their own "OS
> package manager" (which not everybody even has).

It'd be one metadata model to bind them all :-)
If everyone cares about its own OS package manager, at the end of the
day we'd have something tailored for every OS. It's precisely because of
the desire to respect diversity that this should be customized for each
platform (i.e., not that I only care about APT, rather that whatever
comes should be able to work on any system, even if it's not the same
matz uses). 

> A solution that doesn't serve the entire Ruby community is not a good
> solution at all.

Right.

-- 
 _           _                             
| |__   __ _| |_ ___ _ __ ___   __ _ _ __  
| '_ \ / _` | __/ __| '_ ` _ \ / _` | '_ \ 
| |_) | (_| | |_\__ \ | | | | | (_| | | | |
|_.__/ \__,_|\__|___/_| |_| |_|\__,_|_| |_|
	Running Debian GNU/Linux Sid (unstable)
batsman dot geo at yahoo dot com

Never trust an operating system you don't have sources for. ;-)
	-- Unknown source