Date:        Sat, 28 Jun 2003 16:02:53 +0100
    From:        David Laight <david@l8s.co.uk>
    Message-ID:  <20030628160253.Z2608@snowdrop.l8s.co.uk>

  | Erm why?

Because almost no-one ever remembers (or wants) to race around fixing
all those /etc/hosts files when a system gets its address altered.

  | Also you need to put entries into /etc/hosts for any system you
  | NFS mount from during the boot process.
  | Then you can boot systems when the name server is down.

First, there shouldn't be any system which is "the name server",
there should be several systems which are "name servers", so the
chance of them all being down are slight.

Second, if you really need protection from that, then let the hosts
file be used second, not first, so it gets used only if the DNS isn't
functioning (which is what the original DNS resolver interfaces did,
and which worked well enough).   You still have problems if hosts get
renumbered, but less often.

  | (Or if you have nasty non-hierarchic NFS mounts that make it almost
  | impossible to bring all the systems up in a sane order after a
  | power outage.)

If you're stuck in an evnironment like that, you're almost certainly better
off using amd over fstab mounts - it much decreases the chances that you'll
get into a state where everything deadlocks (which is way to easy to do
when you start NFS mounting backwards and forwards without thought).

kre