Thor Lancelot Simon wrote:
> BSD/OS isn't a descendent of 386BSD, though many of the machine-dependent
> parts had the same original author.

   Hrm.  That gets into the fuzzy parts of my "too long ago to
remember" memory.  But, it sounds familiar.  Sorry for the
tangent...

> FreeBSD retooled their entire disksubr some time ago; until then it worked
> basically just like ours.

   Ahh.  Okay.

> No, I assure you, I am painfully aware of the relevant history, as
> odd as that history may seem to you.

   Oh, the history doesn't seem odd (well, tossing out another
partition for something when there were only 8 seems a little
odd, but), it's just complicated.  As histories tend to be. :-)

> That's wrong, too.
> 
> What you're seeing is a dubious attempt to account for the presence of
> both the BSD disklabel and the BIOS MBR: in theory, the BSD system just
> isn't ever supposed to use 'd', because the entire portion of the disk
> that "belongs to it" is described by the 'c' partition, as always.

   Right.  I understand the reasons for doing it.

> Unfortunately, it's not that simple.  We have native tools for installing
> bootblocks, adjusting the partition table, and so forth, and those really
> _do_ need to deal with areas of the disk outside the part "owned" by our
> kernel; thus the necessity to use a partition to actually name the whole
> logical disk presented by the controller.

   Right.  Which is really about the same as the physical disk in
most cases, and the driver doesn't need to know when it's not.
Until there're RAID controller config things that do, but they'r
get that from the controller, not the "device", so...

> And that partition is 'd', and has been for, as I said, more than 10
> years now.  Is it strange?  Yes.  Bogus?  Quite possibly.  But it's
> also how we've always done it, not some recent change, as you posit.

   I apologize.  I never intended to posit that it was a recent
change.  Just that my experience with NetBSD, specifically, is
quite short, so I was wondering where it came from.

   Perhaps my complaint should be that disklabel doesn't warn
you about this.  It didn't care at all.  Of course, IMHO, it
shouldn't care, and should let you do whatever, which gets
back to my belief that it's a bug that something is treating
'd' as if it's the whole disk (as most NetBSD people would
expect), even when it's not (according to the disklabel, which
should be the primary information source).  (*shrug*)

   Anyway.  Thanks!

                             - Chris