Subject: Re: Suggestions for easing 1.1 i386 installations
To: Simon J. Gerraty <>
From: Jordan K. Hubbard <>
List: current-users
Date: 10/09/1995 15:18:13
> Like I said, over-kill for a single OS single release install
> facility, but as you can see you can get very flexible...

Oh, I dunno..  Many people would say what I'm planning with FreeBSD's
even more general purpose `setup' utility is even more over-kill than
that! *grin(

But I don't think so..  Believe me, as the release engineer for the
last 3 FreeBSD releases (and author of the installation itself, which
is something you won't usually get me to admit to without a lot of
beer) I've seen that one needs to cover a lot more issues than just
simple installation.  What happens afterwards?  What happens if the
user wants to use it as a repair too, or to do an upgrade, or simply
to add more software or an extra drive?

I think that the ultimate installation tool also needs to be the
configuration management and general system setup tool, as SCO did
with `sysadmsh' (and as evil as that one is, it can't be faulted for
generally doing the job!) and for that, a fairly intelligent framework
needs to make up the underpinnings of the system.

I think that I'm on the way towards providing that fairly decently
with setup and as there are very few architecture specific aspects to
what I'm currently doing, and I could see it just as happily running
on SPARC as on the i386.  In fact, I see the "install scripts" people
write under such a system as being considerably more intelligent at
dealing with architecture differences since they're both:

	a) Architecture neutral themselves.

	b) Able to examine runtime variables like ${ARCH} in determining
	   whether or not to modify some aspect of their installation (or)
	   configuration behavior.

I don't think I'm doing justice here to the picture that's in my head
here anyway, so maybe you'll just have to take my word for it until I
find a better way of describing this.. :-)

Or you could just ignore me and we could continue to evolve our
installation and system management tools entirely separately.  There's
at least more than ample precedent for that! :-)

The "danger" with that, however, is that I'll go off and finish setup
(and I will, you can trust my motivation here for wanting a better
installation tool so that I don't go *completely* insane at FreeBSD
2.2 time!) on FreeBSD/i386 only, it being my only platform, and if you
guys decide later that maybe it really WOULD be a fine SPARC/mips
solution, well, then the architecture issues will have to be brought
in later and that may be messy.  I'll try to keep things as
non-Intel-centric as possible, naturally, but it'd help if I even knew
that I should be making any special effort!

I guess the bottom line is this: Are you guys firmly set in the
direction you want to go vis-a-vis installation tools?  I've done 3
different generations of the FreeBSD installer and I can tell you that
I haven't even scratched the surface of what needs to be/could be
done!  It would certainly be more than a little premature for me to
set FreeBSD install strategy firmly in stone at this point and I hope
that the NetBSD folks are at least keeping their options similarly