[TUHS] Is there a good, even definitive, list of reimplementations of the Unix kernel? What would good cut-off criteria be?

[Clem Cole via TUHS]

Below, I think I can add a little color to this [note Bill refers to me in
the DDJ articles].

On Fri, Apr 24, 2026 at 6:02 AM Steve Jenkin via TUHS <tuhs at tuhs.org> wrote:

>
> ....
>
>
>         386BSD was Originally intended to be "a university curiosity,”
> Jolitz said, a non-commercial, non-industrial strength way
>                 for students, facility and researchers to have access to
> Berkeley code on inexpensive machines.

This is 100% true.  You need to back up a bit.   Bill had previously worked
for National Semiconductor, where he did the in-house NS16032 port of a
flavor of 4.2BSD to something Nat Semi had made. I have little memory of
the bus, but I think it was Multibus-I, like the original Stanford
University Network (SUN) 68000 base board Andy Bechetalshim had designed.
I personally had no experience with it, although he brought one to USENIX
and showed it off.   I don't know if the compiler was GCC, PCC, or
something else.

At some point, he and his wife formed a company to try to sell a "luggable"
based on those boards.   I should ask Lynn what she remembers.  I don't
know what they were using for a resale license; it might have been an
arrangement with Nat Semi.   But that system was "encumbered" with AT&T IP
and would have needed a license. Also, while (on paper) the NS16032 was
considered architectural, one of the best of the Z8000/68000/386 time
frame, in practice, it was riddled with bugs, and when you wrapped a system
around it, it was a lot more expensive than a 286-based PC.

For whatever reason, Bill's luggage system did not sell very well, and he
eventually gave it up.

Soon after Bill had announced his NS16032 system, the first 386-based PCs
appeared.  Compaq had the first, but the Bay Area terminal company Wyse
released its 32:16, which sold extremely well as it offered the best
price/performance at the time.   I don't have the details, but Bill had
access to one.  And started moving 4.2 to that system (and I believe he was
using a GCC-based toolchain).

I should note that during the same timeframe, I was consulting for NCR's
Chief Architect, Lee Hovel, and had access to much of the Western Digital
documentation.  I don't remember how it came up, but Bill was having issues
with the WD1003-WA2, which IBM had used for the PC/AT and which most
manufacturers like Compaq and Wyse were using.  He reached out to me, and I
ended up writing a good bit of that driver for him [as discussed in the
article].

He would then assemble a system and create a tar image, which he called
386bsd.tar.gz, that resided in a hidden area on ucbvax [this was the first
bag of bits called 386bsd].  Note this system was encumbered, just like any
other 4.2BSD release. However, it included a copy of X10 (I think it was
X10, not yet X11).  But it was a complete system; the idea was to give
students, faculty, and researchers what they were currently getting from
Sun, but using "PC economics."

The key point here is that if you had a Berkeley license (which at that
point meant you needed an AT&T license too, and almost all Universitys had
such a license since it was $100), Bill would send you the directions on
how to download it.   There was no formal announcement, but people went to
places like USENIX, and it was discussed, and the "secret code" was
passed.  Many people/teams did, particularly University Researchers.  As I
have said before, I've often wondered what he would have done if Linus had
known about it.  At the time Bill made this available, Linus was a student,
and his University had both AT&T and Berkeley licenses.  If he had known,
he could have had it.


>
>
>         Increasingly through last year it became apparent that what CSRG
> wanted was “basically the same thing as BSDI":
>                 an unencumbered commercial system.
>
And this is also true.   By the time Bill created the first 386bsd kit, the
idea of a truly unencumbered BSD was gaining momentum.  Keith had done an
amazing job of getting the user space cloned, which Kirk and others had thought
was unlikely, if not impossible [Keith wrote Nvi himself, which was often
said to be the tool that would never be replaced].