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

[steve jenkin via TUHS]

“Imitation is the sincerest form of Flattery”...

I’ve seen mentions on-list of clean reimplementations of Unix, not ‘forks’ - as interesting & important as they have been.

We know that implementing a V7 kernel API was a masters project for Linus,
taking ~3 years to crank it out, presumably his sole work as a thesis requirement.

For me, the importance of V6 was two-fold:

	- ken & dmr removed the unnecessary excess, over multiple iterations, 
		creating the software equivalent of Gordon Bell’s “minimum computer” (eg PDP-11).
		“just enough, no more” for a timeshare O/S and software development platform.

	- they created a “just enough, no more" implementation language in C, 
		flexible enough to concisely express both kernel & user-land code,
		expressive & complete enough to write any systems code,
		and produce performant code, able to be instrumented & improved.

Both inventions satisfy the “nothing left to remove” criteria of great Engineering.

That individuals could reimplement the C language or kernel API in a modest time, not just read & understand it,
shows the level of that original achievement - a definitive O/S for the time [ before networking & graphics terminals ].

The only time (I know of) that the Unix API has been improved on - reduced, simplified & refined:
	is "Plan 9” from the same group, which also addressed Security, Networking & Grpahics.

Is the reimplementation question covered by the classic Levenez “Unix Timeline”? 
Not to my eye.
	<https://levenez.com/unix/>

	"This is a simplified diagram of unix history. 
	There are numerous derivative systems not listed in this chart, 
		maybe 10 times more! In the recent past, many electronic companies had their own unix releases. 
	This diagram is only the tip of an iceberg, with a penguin on it ;-)."

Index - single page - lists all covered O/S.
	<https://levenez.com/unix/indexunix_a4.pdf>


Warren’s page, Timeline of Unix development 1969-1989, is a very clear single page tree of BTL descendants.
	<https://www.tuhs.org/unixhist.html>

plus the Graphing Project, BTL descendants
	<https://minnie.tuhs.org/Unix_History/index.html>


Without Linux, we don’t have a good history of parallel & derivative implementations,
to show the critical importance of Unix V6 / V7 to the world of computing,
without which hardware is useless.


‘History’ is infinite, so I’d suggest a time constraint:

	Either 2000 or 2002, when “Plan 9” was Open Sourced and Inferno was being licensed.


I guess I’m thinking of small teams (& single person) and not large commercial efforts, typically forked from BTL,
which emphasised difference & incompatibility.

There’s been the odd mention of early reimplementations on-list, but I don’t have the Search Skills to extract them.

BSD began as BTL / USG Unix, then diverged with DARPA funding, following a winding path to become free of AT&T copyright material.
BSD spawned SunOS / Solaris.

AT&T, as well as V7, did PWB/Unix, plus many early variants for different research and production needs, especially real-time for switching.

Research Unix diverged from USL/USG after V7, eventually into Plan 9 and Inferno. It’s well covered in existing Timelines.

Bill Plauger did Idris, standalone and hosted..

Andrew Tanenbaum did Minix, inspiring Linus.

The Mach micro-kernel created its own descendants as a platform to host other O/S, including Darwin / MacOS.

The MKS Toolkit - Mortice Kern Systems of Canada - Unix tools for other platforms, but no kernel…
I bought it, found it wonderfully helpful, but is it “Unix”?
 - for users, yes
 - but not the kernel, it’s not an O/S.

Was “Wind River” and its VxWorks “Unix-like” or something else?
What about other Real-time O/S and embedded systems?

Does GNU count, as their HURD kernel has never attained the attributes necessary for wide use?

==============

Then again, what’s even “Unix” in my question?
There needs to be a selection constraint beyond time.

Even in 1974, every installation modified their kernel, creating many forks and original work,.

Sometimes the work was folded back in, mostly it seems to have been lost, or showed up in other products.

The V7 kernel API, the basis for "Spec 1170 “, then the POSIX standard & now the Single Unix Specification.
	<https://opengroupblog.wordpress.com/wp-content/uploads/2016/03/short-version-unix-web.pdf>

MINIX reimplemented the V7 API, creating the model for Linus.
As the first portable Unix, anything that reimplemented the V7 API is a candidate for my list.

Where does that leave the x86 reimplementation of V6 Unix [ MIT? ], used for teaching?
That’s definitely a “Heritage Unix” related system.

As would be an ARM V6 variant for teaching. R-Pi’s and knock-offs are prolific and cheap.

What of Unix-like systems who don’t get POSIX Certification?

Home projects, like Linux 0.1, are where interesting systems began, 
by definition, not highly resourced and more interested in the code than ancillary issues.

==============
--
Steve Jenkin, IT Systems and Design 
0412 786 915 (+61 412 786 915)
PO Box 38, Kippax ACT 2615, AUSTRALIA

mailto:sjenkin at canb.auug.org.au http://members.tip.net.au/~sjenkin