[dyson@iquest.net: Here is what I promised :-)]

----- Forwarded message from "John S. Dyson" <dyson(at)iquest.net> -----
Message-Id: <199806202223.RAA05437(at)dyson.iquest.net>
Subject: Here is what I promised :-)
To: current(at)FreeBSD.ORG, hackers(at)FreeBSD.ORG
Date: Sat, 20 Jun 1998 17:23:15 -0500 (EST)
From: "John S. Dyson" <dyson(at)iquest.net>
Reply-To: dyson(at)iquest.net
Sender: owner-freebsd-current(at)FreeBSD.ORG

Gang:

This is to follow up on my promise for a more formal goodbye. I
have worked on FreeBSD for quite a while, trying to push the
envelope of the existing kernel structure. While there is still
*some* room for growth in the current structure, there are also
problems that exist in *every* conventional kernel out there:
(Linux, *BSD, etc...) I didn't want to say my formal goodbye
until I was sure that I had something good to say about the
new works that I and certain other individuals have been
developing.

With all of the other stuff out of the way, I am definitely
still in the OS world. Think of what I am doing now as the time
that we will be "off in the corner" thinking about solving the
problems that we have learned about when working on FreeBSD,
and reviewing the competition. In the same way there were long
periods when FreeBSD was being worked on, but changes and modifications
weren't ready for public consumption (in fact, alot of the work
that I was doing when I left, is frozen due to it not being
ready yet), this new project is in that mode today. There
is alot of technology out there that solves the existant
problems in a way different than conventional techniques.
The different technology sometimes has its own new set of
problems, and the challenge that we have taken on is to solve
the existant conventional kernel problems, without making the
solution impractical due to its own failings.

It appears that we have found a direction that will apparently
bear very sweet fruit, with few bitter seeds. The technology is
significantly different from existing kernel structures, and has
fault tolerant aspects that are almost impossible to imagine in
a conventional setting. The design and concept phase is
progressing rapidly, and an initial structure will likely be
complete soon. It should be bootable, but not complete (from
an API point of view), in the next month or so.

The API and emulation support should be showing signs of life,
soon after the initial booting. Entire subsystems should port
into the then-existing structure, with some robustness against
subsystem failure, and suprisingly little change. The kernel
itself provides an amount of support that simplifies the number
and type of abstractions, while being able to support all of
the traditional abstractions to userland. The spectre of
total transparency of CPU and machine boundaries with robustness
is becoming very clear with this new structure. There is even
the goal (and practicality) of *some* transparency in a
heterogeneous network also. In no way is efficiency going to
be significantly compromised, and in fact, there will likely
be areas where efficiency is suprisingly good.

So, that is also kind-of an early status report of the progress
on the new project. I believe that this will become a significant
forward force in free kernel design and flexibility. That
certainly doesn't take away from what the current kernels are
today, but after spending enough time trying to answer problems,
it became clear to me that the answers that we were getting
may have been correct, but the wrong questions were being asked.

I hope to have contributed to something soon that will not compete
with any OSes, but will be a useful adjunct and represent a
capability that until now, hasn't been very practical or
efficient.

John

----- End forwarded message -----

-- 
Wolfram Schneider <wosch@freebsd.org> http://freebsd.org/~w/