[TUHS] BSD 4.1, 4.1x, Quasijarus, and 4.3x

Clem Cole clemc at ccc.com
Wed Feb 2 10:22:07 AEST 2022

Dan/Will - I think we need the political way back machine ...  below...

On Tue, Feb 1, 2022 at 5:19 PM Dan Cross <crossd at gmail.com> wrote:

> On Tue, Feb 1, 2022 at 4:35 PM Will Senn <will.senn at gmail.com> wrote:
>> All,
>> I did my research on this, but it's still a bit fuzzy (why is it that
>> people's memories from 40 years ago are so malleable?).
>> 1. What are y'all's recollections regarding BSD 4.1's releases, vis a vis
>> the VAX. In McKusick's piece, Twenty Years of Berkeley Unix, I get one
>> perspective, and from Sokolov's Quasijarus project, I get quite another. In
>> terms of popularity and in terms of stable performance, what say you? Was
>> 4.1 that much better than 4BSD?
> Yes but you need to understand the politics.  ARPA had been supplying
16/36 PDP-6/10s to the research community with BBN, MIT, Stanford and CMU
sort of being the prime places.   The compilers mostly were the same (close
enough) although the OS's (user interfaces) were all different.  Tenex vs.
ITS vs. WAITS vs. TOPS (sort of).

People forget that the real reason it happened was that UCB had a contract
to get Maxima running on a Vax for DARPA [this work would begat 3BSD].
Only after that, did DARPA switch to the 32 bit vax as the system of
choice.   Stanford wanted VMS and mostly everyone else wanted UNIX.   But
DEC did not support UNIX.    So ... ARPA funded CSRG to support UNIX for
the ARPA contracts.    But remember CSRG did not have everything - just a
support contact [which raised a lot of hackels both inside and outside of
UCB].  For instance networking for UNIX was at BBN not UCB.  And all of
CMU, Stanford and MIT had ARPA $s for basic OS research and other ideas.

The Stanford crew tried to show that VMS was faster than 4BSD.   Joy
famously did the 'FASTVAX' work over a period oa 6-8 weeks and got UNIX
within epsilon of VMS on almost everything and actually faster in a number
of places.  This would become 4.1 - add it was the version that most people
that had vaxen started to use.

> Was 4.1as obsolete immediately as McKusick says?
> Well yes in that very few systems ran it. It was Joy's first attempt at
creating sockets.  I'm probably not being fair, but it was the BBN stack
hacked to have a new interface.   I ran it on UCBCAD and Sam probably had
it running on 5-10 750s in Evans.  I thought it got sent to a few sites
outside of UCB, but 4.1B was pretty close behind.

> 4.1b sounds good with FFS, was it?
> Hmm, you'd have to check the sources.   I thought Kirk had FFS running
with 4.1a --- but maybe not.

> 4.1c's the last pre 4.2 release, but it sounds like it was nearly a beta
>> version of 4.2...
> Hmm, not so much a release candidate as a test vehicle.  Beta is probably
a good term for it.   4.2 was not ready when I finished so I brought 4.1c
with me to Masscomp - which is what we started with for the networking
code.  It was what we added MP support to and then added 4.2 fixes when it
came out. But 4.1c was breaking a lot of user mode code and so 4.2 really
did not go out until they had caught most of the issues.

>> 2. Sokolov implies that the CSRG mission started going off the rails with
>> the 4.3/4.3BSD-Tahoe
> Not really.  A few things happened.  First manufacturers (including DEC)
were now supporting UNIX so DARPA no longer felt they needed to fund
support.  You could buy a computer and system vendor made work.   Second as
I said, there were people inside of the EECS Dept at UCB that felt that
CSRG was turning the department away from research and they needed to
reclaim that.  And third as others have suggested, Joy went to Sun, and a
number of the others formed BSDi. Kirk was still there and there was still
some work going on, but it was nothing like it was.    The net result was
the desire to keep going was lost, less than going off the rails.  It just
did not have the same people, the same money and as such, the same effort.

> and it all went pear shaped with the 4.3-Reno release, and that Quasijarus
>> puts the mission back on track, is that so?
> Bluntly, no. Sokolov fell deeply into the nostalgia trap, and combined it
> with a revolutionary zeal that was off-putting at best. As far as I know,
> it was never more than one individual's project, and boasts of reclaiming
> the mantle of CSRG and BSD are grossly exaggerated. The world changed, and
> Unix moved in a different direction to accommodate; there's really no going
> back.
Amen -- I can add little to that.

> 3. I've gotten BSD 4.2 and BSD 4.3 releases built from tape and working
>> very well. I just can't decide whether to go back to one of the 4.x
>> releases (hence question 1), or go get Quasijarus0c - thoughts on why one
>> might be more interesting than another?
> Quasijarus is like 4.3 with some bug fixes and enhancements. If you want
> to run something like 4.3 in an emulator, it's not bad; I'm running it for
> a ham radio project (just for fun) and it's Y2K compliant and seems
> reliable.
Then again, if the idea is running BSD on a Vax, Ultrix 4.5 is even better,
and has all the DEC languages and layered products.

> 4. Is Quasijarus0c end of the line for VAX 4.xBSD? Why does tuhs only have
>> Quasijarus0 and 0a, was there something wrong with 0b and 0c?
> Well, no. Both OpenBSD and NetBSD ported back to the VAX, but the OpenBSD
> effort has ended due to lack of hardware and interest. It appears that
> NetBSD is still being actively developed on the VAX, however, so it's
> possible to get a "modern" 4.4BSD derived system on that architecture.
Right - either current NetBSD or the Ultrix 4.5 is what I would do.  If
NetBSD will run the Ultrix layered products, that would be the system with
the most; but I'm not sure if that will work.   Ultrix for instance
supports DEC (VMS) FTN - which I know you (Will) have been messing with.
That is the best (most complete) FTN for Vaxen.  I believe there is Ada,
PL/1 and maybe even Basic2 and Cobol/RPG [I think most of not all of the
VMS languages for the VAX were released on Ultrix but frankly, I don't

> 5. Has anyone unearthed an original 4.1 tape, or is Haertel's
>> reconstruction of the 1981 tape 1 release as close as it gets?
> For the fifth time today this reminded me that I wanted to find my images
> from Kirk's CD collection and move them over to another machine. Sigh.'
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://minnie.tuhs.org/pipermail/tuhs/attachments/20220201/f2010e0e/attachment.htm>

More information about the TUHS mailing list