Return to BSD News archive
Received: by minnie.vk1xwt.ampr.org with NNTP
id AA1675 ; Tue, 23 Feb 93 14:53:02 EST
Path: sserve!manuel.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!sun-barr!news2me.EBay.Sun.COM!jethro.Corp.Sun.COM!lamer!lewy
From: lewy@sun.com (Lew Yobs)
Newsgroups: comp.os.386bsd.questions
Subject: Re: WFJ's talk last night...
Date: 19 Feb 1993 00:35:36 GMT
Organization: Sun Microsystems, Inc.
Lines: 120
Distribution: world
Message-ID: <1m1a0oINN8ds@jethro.Corp.Sun.COM>
References: <C2nHuD.5EC@raistlin.udev.cdc.com>
Reply-To: lewy@sun.com
NNTP-Posting-Host: lamer.corp.sun.com
Here's my notes on the talk at SVnet meeting last night.
Notes are completely unofficial. flames > /dev/null
---------------------------------------------------------------------------
What: SVnet Meeting
Date: 93/02/17
Speaker: William Jolitz (BJ in this text)
Topics: 386BSD Present and Future, Law suit, etc.
1. Intro remarks
Release 0.0 (92/03/17) was binary distribution
Release 0.1 (92/??) freely redistributable and modifiable
Independent groups working on various extension efforts, e.g.
- group in Russia working on 16-bit ISOcode port
- group in former DDR working on ISOcode port too.
- guy at novell also working along same lines
WJ gets contributed s/w pieces from folks all around the world.
Estimates > 30K machines are now running 386BSD.
Widespread use as non-commercial system for academics and
experimental development has developed as intended.
2. Futures
386BSD has embedded in it some late 1960's concepts
but now older machines are on the way out: mainframes, mini's.
Emphasis on "going forward" for benefit of new research as main
theme for 386BSD.
A. Near term
More stable; more extensible system.
Too much is involved now with implementing changes in the unix
system: kernel + libraries + allications must all be changed
to accomidate / take advantage of changes to low levels.
=> Establish firm interfaces -- wait for 0.2 release!
B Far term
BJ will try to stay involved for 'the long term'
focus on design for networking
Opens for questions:
Q. Request to elaborate on methods for localizing change side-effects.
A. Began moving in this direction with 0.2 work:
- Moving in the direction of C++.
- classes idea for kernel objects
- loadable drivers
- subsystem encapsulation: drivers, filesystems, protocol code
- autoconfiguring kernel
- no cdevsw[]
- code is not PIC (position independent code) yet so addresses get wired in
- kernel is *much* smaller than in 0.1
- BIG 0.2 items:
rewritten VM system
"__" files in kernel were bandaids and are being ripped out
pagecache
"clustering" - ability to load much buffer space at once
See usenix paper by Kleiman and McVoy
vnode objects / vfs interactions cleaned up
(More information discussed at meeting here I did not get.)
stacking filesystems
e.g. stack a compression filesystem with other fs type
to achieve "both"
OVERLOAD INTERFACES are WJ's goal for system design
*** Somewhere during the meeting the topic moved to status of USL / UCB / BSDI
lawsuit. No notes were taken by me on these discussions. See other
postings in the newsgroup for claims / counter-claims, dirt, smut.
3. Release 0.2 status
Is still not ready.
- still are significant problems
- much of kernel has been rewritten
new vfs layer and fs code
- idea of portals
are files that are versioned with processes (??)
new inode type
new ipc type
- need cheaper way of making network daemons for future network focus
- ability to "export" user code to kernel
- shared libraries
is one of biggest current problem areas
best of OSF and Sun: Doesn't know if can ship OSF loader
- Los Alamos Nat'l Lab contributed code to allow kernel to operate
at any location and with kernel size > 640KB
Can he release this??
- tcp/ip enhanced stack: triple 0.1 speed
- snmp
- bpf (berkeley packet filter)
- tossed alot of interfaces; kernel much smaller
- sockets now in vnode layer => is a loadable fs
- can use up to 1GB of RAM
- libraries have had their namespaces cleaned up; garbage names removed
- any posix incompatibilities should be brought to BJ's atention
*** Only freely modifiable and redistributable code will be put into
kernel and libraries.
Q: What is used for testing?
A: Has own test suites and group of early release testers.
*** Requests that all code submitters provide own test suite to verify code
submitted.
---------------------------------------------------------------------------
---
______________________________________________________________________
Lew Yobs Sun Microsystems
______________________________________________________________________