From owner-freebsd-hackers  Mon Oct 23 03:52:41 1995
Return-Path: owner-hackers
Received: (from root@localhost)
          by freefall.freebsd.org (8.6.12/8.6.6) id DAA02825
          for hackers-outgoing; Mon, 23 Oct 1995 03:52:41 -0700
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id DAA02819
          for <hackers@freebsd.org>; Mon, 23 Oct 1995 03:52:35 -0700
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id UAA24992; Mon, 23 Oct 1995 20:37:17 +1000
Date: Mon, 23 Oct 1995 20:37:17 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199510231037.UAA24992@godzilla.zeta.org.au>
To: hackers@freebsd.org, msmith@atrad.adelaide.edu.au
Subject: Re: New userconfig
Sender: owner-hackers@freebsd.org
Precedence: bulk

>In order to meaninfully interpret the values assigned the basic ISA 
>device parameters, I currently make the following assumptions :

>If a parameter is -1, the driver does not use the parameter.  It is never
>relevant, and is never consulted.

>If a parameter is -2, the driver has a compiled-in default for this parameter,
>but the default can be overridden by setting the value to another value.

Note that userconfig can't use special values to decide which values can be
changed, since it needs to be able to change to and from all values.
Currently unadvertised driver capablities determine which values actually
make sense.

The -1 and -2 above are back to front.  -1 has always meant `?'
(autoconfig).  There is a special rule for this in config/lang.l.
Drivers should never use compiled in defaults for this, but they may use
values read from firmware.

>Joerg has kindly provided details for modifying config(8) to recognise
>'auto' and 'none' as alternatives to '?'.  I'd like to ask further for
>comments (and advice) on setting the _default_ for nonspecified parameters
>to 'none' (-1), rather than the currently ambiguous state (some -1, some 0).

`none' is a verbose alternative to `'.  I think nonspecified parameters
always have a value of -1 except in cases where 0 is more natural (for
pointers and bitmaps).

>Julian, specifically to you, and generally to anyone else working on device 
>driver registration issues - is there a "better way", that takes advantage
>of the devfs concept/support/etc?

I think you'll eventually have to modify device tables in 1001 drivers.

Bruce