> True.  OTOH, I'm not sure I buy your argument.  ISTM that you are
> basically arguing that if the FS which LOCALBASE resides on blows up
> duing the build, you can't get at your backups.  Well, if the FS which
> LOCALBASE resides on blows up at any other time, you can't get at your
> backups.  I don't see the risk increase significantly.

No, I'm not concerned about the the FS blowing up, I'm concerned about
the PKG blowing up.    Ie. a bug in the pkg preventing it from building
after its just forced you to toss your existing, known, working backup
system.  Planning for builds to fail is simply prudent when tracking
-current.

Fortunately, the fact that I was able to build ok on the 1.4.1/sparc
system satisfied me enough that a build on the i386 box (after
pkg_deleting the old amanda pkg) was not a major risk.  Not everyone
is in that situation though, hence my argument that the current amanda
pkg split is not a good idea.  

Sure, have amanda-{client,server,...}  for those that want it, but I'd
suggest that the amanda-all pkg should be exactly that, it should
build everything and be able to build _within_ the work dir, so that
sites that have only one NetBSD machine have a means of upgrading
without burning bridges.

Simon Burge's suggestion of doing a make package after the build is
another good insurance policy.

FWIW I got both the i386 and sparc backed up ok last night :-)

--sjg