Subj : Re: Synchronet vs Mystic vs ??
To   : Arelor
From : tenser
Date : Tue Apr 25 2023 12:34 am

On 23 Apr 2023 at 06:05p, Arelor pondered and said...
 
 Ar> Afaik FreeBSD is not the original.
 Ar> 
 Ar> NetBSD has an initial release date set at 19 April 1993, whereas Freebsd
 Ar> has it on 1 November 1993.
 Ar> 
 Ar> NetBSD was derived from 386BSD just the same as FreeBSD. OpenBSD is a
 Ar> fork of NetBSD. If anything, the o"original" the big *BSDs are branched
 Ar> from is 386BSD.

This is correct:  Bill and Lynne Jolitz ported the "Net/2"
distribution of BSD from UC Berkeley to the 386, famously
writing a series of articles about it in Dr Dobbs Journal;
the series was later collected into a book.  The result was
386BSD.

However, the Jolitz's were a bit, um, "challenging" to deal
with and progress on, say, necessary bug fixes, supporting
new hardware, etc, was all slow, and unofficial patchsets
were floating around on the net.  Eventually, these were
collected and distributed as part of NetBSD.  FreeBSD was
an independent effort along the same lines.

Of the two, FreeBSD concentrated on excellent support for
the PC platform, while NetBSD was more concerned about
portability; for a while in the 90s, if you had a reasonable
computer of almost any description, it was likely there was
a NetBSD port for it.


OpenBSD was a fork of NetBSD, created after Theo de Raadt
was kicked off the NetBSD core team and his access to the
NetBSD source repository suspended.  The details there
are not terribly interesting; suffice it to say that he
has a tendency to behave badly.  OpenBSD claims to favor
security and correctness over all else, but honestly, I'm
not really sure that's _actually_ true, but that's their
main selling point.

Dragonfly is the other main BSD fork; it derives from
FreeBSD.  Matt Dillon, the principle author, forked it from
the FreeBSD 4.x series, as he disagreed with the direction
they wanted to take for SMP implementation in FreeBSD 5 and
beyond: all of the BSDs were original single-processor only
(indeed, same with Unix; multiprocessor systems were not
terribly common in the era it was written and on the types
of computers it was written for: minicomputers).  As time
moved all, all of them realized that multicore was going to
be a critical, particularly with the end of Denard scaling,
so set out to modify their systems to support SMP (symmetric
multiprocessing).  Suffice it to say that retrofitting that
onto an existing operating system that wasn't designed to
support it from the start is challenging; lots of code was
written making lots of assumptions about things that it
presumes can't happen when there's only one CPU, but that
very much can, and do, happen when there are multiple.

Anyway, there are several ways to do this, and in particular
to expose it to userspace programs.  The most common is a
1:1 thread model, where a single user-space thread corresponds
to a single thread of execution in the operating system kernel.
This is far and away the most common.  The second most common
is to use what are called "Green Threads", or N:1, in which many
userspace threads are multiplexed onto a single kernel thread;
the problem with this is that, if a user-thread initiates some
blocking operation (say, opening a file), this blocks all
threads.  Finally, the least common is N:M, where N userspace
threads are muxed onto M kernel threads, and M != N, usually.
This was a hot area of research in the late 90s, and this is
what FreeBSD tried to do in FreeBSD 5.  Dillon thought this was
too complex and would yield poor performance, and so forked
DragonFly to explore a different model.  He has since pivoted
to working on advanced filesystems, with HAMMER and now HAMMER2.

Incidentally, he was right: after a long struggle, the FreeBSD
people _did_ implement their N:M threading model, but it wasn't
a good match for real-world applications, and was very complex.
Eventually, they deprecated it and went with the 1:1 model.  I
suspect some support is still in there, but practically no one
uses it.

As noted, all of these descend from the original BSD work
done at UC Berkeley, which had a long history of Unix
involvement, starting with the 5th Research Edition in
1974.  Ken Thompson had gone to Berkeley, and took a Unix
tape with him when he went on sabbatical there in '74.
This, all of the modern BSDs descend from Net/2.

Net/2 was interesting: it was an _almost_ complete Unix
distribution, just with some small proprietary parts removed.
This was the culmination of a large body of work spearheaded
by Keith Bostic, to create a freely available BSD under a
permissive license.  It was very close in interface and
usage to the final 4.4 releases, which in turn, were mostly
POSIX compatible; the various 4.3 and earlier releases were
not.  A company spun out of Berkeley, BSDi, around the time
of that release to market a commercial version of BSD called
"BSD/OS".  Famously, their phone number was "1-800-ITS-UNIX".
This led to a lawsuit where AT&T sued both BSDi and UC
Berkeley, and Berkeley counter-sued AT&T, over copyright and
trade secret infringement: AT&T had been legally barred from
competing in the computer industry as long as they held their
regulated monopoly status over the telephone system.  However,
shortly before all this went down they'd split themselves
up and thus been released from the consent decree that so
barred them: they intended to enter the industry marketing
Unix (they did, but were largely unsuccessful).  This lead to
a period of FUD over the future of BSD-derived systems, which
gave the new Linux upstart a chance to get a foothold in the
industry.

Linux, on the other hand, was a complete reimplementation
from scratch.  Linus Torvalds wanted something that took
better advantage of his hardware than Minix, the teaching
system by Andy Tannenbaum that he had been running.  Linux
took a far more traditional approach to building a Unix-like
system, in that it's a monolithic kernel (the kernel exists
in a single address space), while Minix is a microkernel
(services are logically distinct and isolated from one
another, and communicate via message-passing).  Famously,
Tannenbaum took Torvalds to task for this decision, declaring
Linux obsolete before it was finished.  Of course, it is now,
by far, the most popular and important operating system in
the world.

--- Mystic BBS v1.12 A48 (Linux/64)
 * Origin: Agency BBS | Dunedin, New Zealand | agency.bbs.nz (21:1/101)

.