Subject: Re: goals for NetBSD binary packages
To: Klaus Heinz <k.heinz.apr.fuenf@onlinehome.de>
From: Hubert Feyrer <hubert@feyrer.de>
List: tech-pkg
Date: 04/25/2005 10:21:57
On Sat, 23 Apr 2005, Klaus Heinz wrote:
> What do people expect from binary packages?

Availability & consistency.

Availability on the latest 1-2 official releases (2.0, 1.6.2), and
 	ideally all platforms that NetBSD runs on. My initial motivation
 	to start pkgsrc was to get binaries available for NetBSD/amiga.
 	Aparently that hasn't worked out. I guess some work on cross
 	compiling would be needed here. Having the fine infrastructure
 	for that in NetBSD should give a base to start.

Consistency should be given to make sure everything available as
 	binary pkg can actually be installed, i.e. all dependencies
 	are there as well. Esp. in (recent) light of security handling,
 	efforts need to be made to handle things properly. As far as I
 	know, all the bits are done by individual developers with not
 	enough communication. I think it would be good if pkgsrc's
 	management (to which I do _not_ count myself for quite some
 	time now!) could sort these things out - analyse current and
 	desired situation, define steps, document and perform them.

I think that pkgsrc is an excellent system for installing software from 
source. The "stable" branches are really helpful. There are still "work in 
progress" areas like what the right way to update systems are (under 
various constraints), but I have faith in pkgsrc developers that these 
issues will work out.

Areas that I see in pkgsrc that need attention are:

  * better integrate binary pkgs in "release" handling, not only from
    the pkgsrc team  but esp. with interfacing to operating system
    projects like NetBSD (also DFBSD, maybe OpenSolaris at some point).
  * Try to get crosscompiling going to some extent - even a 30%-solution
    means more pkgs for NetBSD/amiga than we have today!
  * Keep overhead low. Try bootstrapping pkgsrc on a slow machine, or
    build things there. There's a lot of overhead going into "portability"
    across various operating systems (which is nice to have, but not
    crucial for NetBSD), which costs performance on slow platforms.
  * Documentation. Existing one's scattered across too many places.
    Undocumented items include upgrading under various constraints.


  - Hubert