In message <200503090948.19761@gryphon.flame.org>, Michael Graff writes:

>
>My major gripes are that building on one machine and rsync'ing the binaries=
>out is becoming a major pain recently.  I guess it's time to start some sort
>of cvs pool for my /etc config files, and some smart script to detect changes
>to /etc and to (perhaps) propagate them into other machine's configs
>during /usr rsync.
>
>pkgsrc is really making this harder every day, but I'm just told I'm doing
>something odd and it would be easy if I just followed these simple steps
>provided by lots of helpful yet not-doing-what-I'm-doing people.  (insert
>about 20 annoyingly tedious steps here)
>

We really need to think seriously about this; it's my biggest single 
gripe about running NetBSD.  I run two different clusters, one tracking 
-current and one on 2.0; they all tend to have slightly different 
configurations and needs for packages.  It's mostly harmless to have 
extra packages installed, of course (I overprovision disk space when I 
can), but trying to build and move appropriate change sets -- patches, 
bug fixes, new builds of -current, etc. -- is a royal pain.

A lot of the evil lives in /etc.  I'm contemplating making /var/etc a 
union mount over /etc; all local changes would be in /var/etc, of 
course.  postinstall or etcupdate should warn if there's a change to a 
file in /etc that also has a mirror in /var/etc.  But this is by no 
means a complete solution to the maintentance problem.

		--Prof. Steven M. Bellovin, http://www.cs.columbia.edu/~smb