On Fri, 20 Jun 1997, Perry E. Metzger wrote:

> 
> Ted Lemon writes:
> > Hm.   Having read this paragraph, it occurred to me that it might be a
> > mistake to try to solve the DOS partition problem with a single
> > partition table - it might make more sense to have two levels of
> > partitioning that are recognized by the i386 kernel.   If you have a
> > drive that's got a DOS partition and a NetBSD partition, perhaps it
> > would be addressable as /dev/sd0* and /dev/sd1*, where sd0* would be
> > the first partition in the DOS table, and sd1 would be the second.
> > I'm not saying this would be the precise implementation - I can see a
> > lot of problems with what I just described - but something along these
> > lines might be a good solution.
> 
> The FreeBSD "slice" notation seems to be the current consensus. In it,
> my DOS partition in MBR slot 0 would be sd0s0c -- my boot partition on
> the NetBSD slice would be sd0s1a.

What about the idea (which someone broached) of making the default MBR
(sp?) behavior ot make each partition an a, b, e, and f partition on wdX,
and then mount a vnd on the whole partition, and do the netbsd
partitioning in the vnd? The node names could be the same as FreeBSD's
slice names, and then all we'd have to do to make it work is teach vnd how
to do disklabels (which it should learn anyway), and do something so
rootdevice works (like maybe have mountroot fire up the vnd on the root
DOS partition).

> Its not perfect, but its better than what we have now, and its
> reasonably compatible with FreeBSD's notation which isn't any worse
> than any other notation we could use.

I thik staying reasonably compatable is GOOD. :-)

> We need to move to 32 bit device numbers first, though -- the current
> 16 bit ones aren't going to be sufficient.

What's involved, other than the headache of actually switching (that
agonizing period where /netbsd and /dev might disagree)?

Take care,

Bill