>> So, on the PC ports, sdxa is always the first partition, root or
>> otherwise?
> Actually, on the i386 port, I consider a is the root partition, not
> necessary the first partition.

It depends on what you mean by "first".

Traditionally, BSD has eight partitions per pack, indexed from 0 to 7
(via trailing letters a through h on the /dev entry names), which used
to be compiled into the driver and nowadays are taken from the label on
the pack.  How these partitions are laid out on the pack varies.  The c
partition (index 2, the third in the table) is (almost?) invariably the
whole pack.  On a boot disk, the a partition is usually the root
filesystem and the b partition is swap space; these are usually but not
necessarily the lowest-addressed and next-lowest-addressed partitions
on the pack.  On a non-boot disk, anything goes, though if a partition
is used for swap, it's often the b partition, for the sake of some sort
of consistency.

Thus, a traditional Berkeley `a' partition is always the first
partition in terms of that table of eight partitions.  It is not and
(at least as far as I know) never has been necessarily the partition at
the beginning of the pack.  Nor has the boot partition necessarily been
an a partition, nor swap a b partition; "config vmunix root on ra0g
swap on hp0d and hp1f" would have been unusual (and a minor maintenance
headache) but workable.

Now, when people have been talking about the SV way, with something
like /dev/disk/scsi/c0t0u0p0 or something (that's a fictitious example:
controller 0, target id 0, unit 0, partition 0), I have been assuming
that the "partition number" refers to the slot number in a similar
(eight-element?) table, which (like the Berkeley partition table) may
or may not be sorted with respect to location on the disk.  Thus,
whether a SV partition-0 corresponds to a Berkeley `a' partition I
don't know; it depends on whether this assumption is true.

Based on what I read here, it sounds as though the Amiga port, unique
among the NetBSD ports and mildly contrary to BSD tradition, juggles
the partition numbering at boot time so that (possibly among other
things) the boot partition is always an `a' partition.  This seems to
be what started this whole flap off.

					der Mouse

			    mouse@collatz.mcrcim.mcgill.edu