On 31 Dec 2000, Johan Ihren wrote:
> Richard Rauch <rauch@eecs.ukans.edu> writes:
> 
> > It's hard to say how many partitions one might actually want.  The primary
> > motivation for more here seems to be for handling shared access between
> > multiple operating systems.  How many operating systems might one install
> > on a single disk?
> 
> I fully agree that the number of needed partitions seem to be more
> dependent upon the need to access filesystems belonging to other OSes
> than a need to micromanage storage local to NetBSD. This is exactly
> why I think an easy way out where we go from 8 to 16 or thereabouts is
> sufficient.

For most purposes, probably.  However, the comments about partition-size
== sizeof(backup-media) make some sense.  Also, for comparative (or
other) purposes, one might easily install multiple OS's on a single
disk.


> I hope I will never fragment my NetBSD storage into more than, let's
> say 8 pieces (/, /var, /usr, /usr/src, /usr/{local,pkg}, /usr/home and
> two more) on one disk. Likewise, I sincerely hope I will never have to
> run more than the present 3 OSes on my laptop. And if there only was a
> VMware for NetBSD I'd be much closer to finally ditching Lunix... ;-)

(^&


> > At the same time, you don't really need all of those partitions to start
> > booting any single OS.  One is enough to start it; 2 is probably as many
 [...]
> > ...so, what if additional partitions were accessed by something like a
> > vnode?  (Or what if regular vnodes were extended to let you also mark raw
 [...]
> I cannot comment on whether this is technically feasible, but I
> certainly agree that the need for "partitions" for foreign filesystems
> belonging to other OSes is for exchanging data, not to boot from. So,

Even many of your estimated ~8 real NetBSD partitions could be done this
way.  If this idea is sound, then all that you'd need on a real partition
is enough material to get partway through the boot.  Once the system was
more or less available, the boot scripts could mount (via these vnode-like
objects) the remaining filesystems and finish booting.

In that case, 8 partitions in the disklabel would be quite enough.

Well, like I said, it's an idea.  Maybe not a terribly elegant one (at
least IMHO), but...at least no one has come down on it as technically
infeasible.  And it solves the problem in an open-ended manner, if it can
be done.  (^&


  "I probably don't know what I'm talking about." --rauch@eecs.ukans.edu