Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!cs.utexas.edu!uunet!mcvax!hp4nl!botter!star.cs.vu.nl!ast@cs.vu.nl
From: ast@cs.vu.nl (Andy Tanenbaum)
Newsgroups: comp.os.minix
Subject: The future of MINIX
Message-ID: <2835@ast.cs.vu.nl>
Date: 7 Jul 89 11:25:20 GMT
Sender: ast@cs.vu.nl
Reply-To: ast@cs.vu.nl (Andy Tanenbaum)
Organization: VU Informatica, Amsterdam
Lines: 58

I have been fairly quiet lately, not out of lack of interest, but waiting
for the dust to settle.  It seems that the overwhelming majority opinion
is that Bruce Evan's protected mode code is a big step forward.  I tried it
and, of course, it didn't work on my machine.  However, after a round of 
12,000 mile remote debugging with Bruce, we found the problem (2 lines)
and it works fine now.

I have also looked at his code more closely, and I am impressed. It is not
only very well programmed, but he resisted the temptation that most other
people have succumbed to, namely, changing my layout.  If I am to use it in
the book, I certainly don't want two different styles in there, and my 
changing everything by hand would undoubtedly lead to errors.  There is no
MINIX pretty-printer at present (hint).  The UNIX pretty-printer, cb, doesn't
produce my layout style, which I am very much attached to.

Another plus is that Bruce has been very helpful in answering questions etc.
My major fear, that the code would be too complicated for students, is
probably unjustified.  It is certainly longer now, but not that much hairier.
The 80286 stuff is fairly well isolated.  I took a look at how long the
listing in the book will be if I include only the same files as last time,
plus a few new ones that are essential.  It comes to something like 380 pages,
vs. 250 last time.  Together with, say, 30 pages of descriptive text about
how the new code works, and 20 pages of man pages in Appendix C, the book
will exceed 900 pages.  This is getting kinda large for a 1 semester textbook.
Printing it in 2 volumes makes it more expensive.  I might be able to
help a little here and there by not including things like klib88.x and
protect.c, which are very technical and not very conceptual, and perhaps
eliminating the cross reference listing.

Anyway, I have decided to use Bruce's code as the basis for V2.0 (subject to
Prentice-Hall not vetoing a 900 page book).

There are still a number of things that need to be done before I even start
with POSIX.  Bruce and I have discussed this by email.  The game plan is that
he is going to make some changes during the summer, and then post the new
version of the kernel.  I am going to look at fixes etc from the net 
concerning utilities etc, and post those cdifs.  All this should happen
around September.  Let us call that version 1.4b.  This will be my base for
starting to work on POSIX.  I'll do the kernel/FS/MM myself (looking at
what Simon Poole has already done), but help writing/converting libraries,
utilities etc is very welcome.

When all that is done, we will have a protected mode system that is based
on POSIX and reasonably fast for small machines.  MINIX seems to be
evolving in the direction of something more than an academic toy and towards
a more serious system.  If I run into Niklaus Wirth at a conference 
sometime, I'll have to ask him about that :-).

An additional note.  Amiga MINIX is actually up and running.  It is now
being tested.

Furthermore, MINIX 1.3 is being ported to the SPARC.

Getting all these versions back together, and then POSIX-based is going to be
a job on the order of cleaning out the Augean Stables, and I don't have a
Hercules card.

Andy Tanenbaum (ast@cs.vu.nl)