*** please followup to tech-userlevel ***
According to Michael C. Richardson:
> 
>   I prefer that you simply create /var/db on your writeable root either
> put a symlink there, and NFS mount whatever you need either on top of /var,
> or on its subdirectories. If you need to, you could also just restart
> dhclient after you have mount /var.
> 
In this context, would it not be advisable to fragment rc.* and other 
initial configuration procedures into segments appropriate to each 
filesystem mounted, so that it is easier to make sure that 
prerequisites are met before they execute?

It is tempting to suggest including such scripts in an FSTAB that in 
turn distinguishes between init-time mounts and other times.  I would 
suggest reversing the "noauto" option and make it something like:

	init:<script>
or	auto:<script>

where script can arbitrarily live in any currently mounted filesystem.

Also, remounting the filesystem at a later stage could safely ignore 
the "init" option.

Is this unreasonable?  Have I missed some important aspect?

++L