Subject: Re: Installer and root partition roulette
To: None <jope@n2h2.com>
From: Ken Nakata <kenn@synap.ne.jp>
List: port-mac68k
Date: 07/25/1998 11:02:56
Fri, 24 Jul 1998 16:57:56 -0700 (PDT), El JoPe Magnifico wrote:
> 
> However, I just tried to get our hardware guy to give me a nice, clean
> explanation, and he went off onto a tangent about primary versus logical
[...]
> because I haven't run across the logical/primary question _anywhere_ in
> either the MacLinux or MacBSD materials or discussions. (consensus was
> that it's probably due to the controller - does anybody out there know?)

AFAICT, he was talking about the IBM PC partition table format.  It's
a BIOS thing, and has nothing to do with either drive hardware or
controller chip.

> Okay, hold on.  The Booter takes a drive and partition (by name, letter,
> whatever) as an explicit parameter.  Maybe it's default lookup scheme is
> the same as the Installer, but _not_ if the Booter is passed an explicit
> option.  Is this not how it works???

Yes, it is.

> Because the Installer lacks a similar
> such option, it's choking on Unix root partitions that may or may not be 
> BSD paritions, because it only has this default scheme, which is currently
> flawed and assumes (erroneously) any Unix root partition must be a BSD root
> partitions _and_ then assumes (erroneously again) that it must have an ffs,

Because no MacLinux was around (not even near existance) when Alice
(the original MacBSD development gang) decided the way partitions are
handled.  They should probably have used a unique partition type
string for BSD partitions, but it's too late to change it now.  And
A/UX was already nearly dead at the time, too, so to them, there
probably seemed little problem in recycling the A/UX partition type.
Don't forget Mkfs lacked the ability to change partition type until
recently, so there was a need to use existing partition software to
make partitions for NetBSD.  Using A/UX partitions fit the bill
perfectly at the time.

And it is apparent that Linux people in general don't care about
compatibility with other operating systems as much as we do.  We have
a number of kernel compatibility options such as COMPAT_SUNOS,
COMPAT_LINUX, and EXT2FS, none of which or counterpart exists on
MacLinux AFAIK.  Coming out of such a tradition, it is only
understandable that they chose the partitioning convention that was
known at the time of development to crash with NetBSD's.  And NetBSD
had already been around for quite a while at that time.  My point is,
deliberately or not, they, the MacLinux people chose not to coexist
with NetBSD on a single drive, whereas NetBSD has never been in the
position to make that decision.

No, I'm not blaming the MacLinux developers.  I'm just saying that,
from historical reason, our excuse has more validity than theirs 8-)
So, if you blame NetBSD for not being able to coexist with MacLinux on
a single drive, you should do the same or more to MacLinux for
choosing the partition convention which crashes with NetBSD's B-}

> All I'm suggesting is a dialogue before the Installer starts making this
> chain of incorrect assumptions.

Installer does not make a "chain of incorrect assumptions".  Installer
only mounts the root device.  Other partitions, you have to mount them
by hand in MiniShell.

Ok, that you should blame Linux was meant to be a joke (perhaps a bad
one), but seriously, I think there should be some kind of coordination
between the two camps.  It may be good enough for you if you can tell
Installer which is the root partition, but it would be much better if
there were a definitive way to tell NetBSD and Linux partitions apart.

Ken