Replied: Wed, 23 Sep 1998 13:35:17 EDT
Replied: "Chris Torek <torek@BSDI.COM> "
Replied: Tue, 11 Aug 1998 13:27:30 EDT
Replied: "Chris Torek <torek@BSDI.COM> "
Received: from SOUTH-STATION-ANNEX.MIT.EDU by po7.MIT.EDU (5.61/4.7) id AA20504; Tue, 11 Aug 98 13:15:06 EDT
Received: from services.BSDI.COM by MIT.EDU with SMTP
	id AA28215; Tue, 11 Aug 98 13:17:04 EDT
Received: from forge.BSDI.COM (torek@forge.BSDI.COM [205.230.224.24])
	by services.BSDI.COM (8.9.0/8.9.0) with ESMTP id LAA01960
	for <nathanw@MIT.EDU>; Tue, 11 Aug 1998 11:17:17 -0600 (MDT)
Received: (from torek@localhost) by forge.BSDI.COM (8.8.8/8.7.3) id LAA18039 for nathanw@MIT.EDU; Tue, 11 Aug 1998 11:17:08 -0600 (MDT)
Date: Tue, 11 Aug 1998 11:17:08 -0600 (MDT)
From: Chris Torek <torek@BSDI.COM>
Message-Id: <199808111717.LAA18039@forge.BSDI.COM>
To: nathanw@MIT.EDU
Subject: Re: bsd config paper?
Xref: steve-dallas.mit.edu netbsd.other:1
Lines: 37

I have given out copies (mostly of the postscript from dvips, but
for your purposes the LaTeX source is probably better :-) -- there
are a few "xfig" figures as well, but nothing that should be
too much of a problem).

There are a couple of problem with the paper besides just being
"unfinished".  Config itself has two separate jobs:

	- read a "metaconfiguration" file that says how boxes
	  (vaxen, sparcs, pcs, amigas, ...) are put together: what
	  busses can connect to other busses, what drivers there
	  are, what information they need.

	- read a "configuration" file that says how some particular
	  box is put together: "the sparc on my desktop, which I call
	  purr, has a floppy and a keyboard and ...".

and these jobs both tie in with a third thing (or two additional
things, depending on how you count): how to write the code to
do autoconfiguration on a (sparc|pc|vax|...), and how to write
a driver for a (sparc|pc|vax|...).

The NetBSD folks have done a good (but never-ending) job on making
"how to write a driver" no longer be "one for the 68k and another
completely different one for the sparc and a third for the pc, etc."
This helps some, but there is still an element of "this ncr53c80 is
packaged as an `esp' on Sbus on sparc" that creeps in here and there.

Anyway, the paper tries to address some of all of these all at
once, which is itself impossible, or (I now think) at least
inadvisable.  It also describes the code that went into 4.4BSD,
which is not the same as the code in NetBSD, BSD/OS and (soon?)
FreeBSD.  So I think it needs some major reworking and rewriting.

If you reallllly want it, I can send you the source files... :-)

Chris
