Subject: Re: A new partition handling scheme: wedges
To: Giles Lean <giles@nemeton.com.au>
From: Ty Sarna <tsarna@endicor.com>
List: tech-kern
Date: 02/05/1998 14:59:11
Giles Lean wrote:
> 
> Not allowing at least the transitory existence of duplicate partition

I think I've been misunderstood, then.  Read what I first wrote on the
subject -- the partition should be availible, just under a different
(uniqified) name, and preferably *both* partitions with the same name
should be renamed.  In other words, if there are two "foo" partitions,
provide access to them as "foo.0" and "foo.1".  Do *not* provide access
as "foo", since you can't be sure which was desired.  Anything (such as
fstab) relying on access under the "foo" name *should* break, rather
than get the wrong partition with potentially disastrous results.  This
forces the admin to figure out which is which and (hopefully) tread
carefully.

My point was, duplicate partition names should *not* be considered a
normal and acceptable way of running a system.  It is an error
condition.  It should be an error condition that the system can handle
gracefully (by complaining and assigning new names), but it is still an
error condition.  If you're moving disks between machines and didn't
name them carefully, it may be unavoidable -- the system should provide
transitory access, as you say (and it does, as I've described it), but
should also should encourage the admin to fix the problem.  If you have
multiple OSes on the same machine, you definately don't want their
partitions to have the same names. That's begging for trouble.

(and IMHO, you still should name partitions uniquely within
organizations for optimum sanity, where possible.)