[geeks] Sun vs. SysVr4

[Greg A. Woods]

[[ since you seem to have a problem understanding the issues and facts
here I've moved this to over to geeks to hash it out....  or you can
just reply privately if you prefer!  ;-)  ]]

[ On Wednesday, August 22, 2001 at 01:27:59 (-0400), Dave McGuire wrote: ]
> Subject: Re: [rescue] FDDI questions.
>
>   Rather than spout uptime figures and compare the stability of, say,
> Solaris 2.3 (which was the first release of Solaris to even get
> noticed, and still not accepted, by the production SunOS4 crowd), I
> will ignore this painfully obvious spew of flame bait trolling.  I
> will give you the opportunity to drop this now...otherwise I will
> proceed to tear your point to shreds.  Nothing personal.

vs. say the uptime of some 3b2's I know about?  :-)

part of what I mean by "production quality" also has to do with the
ability to manage subsystems on the box properly and reliabily, as well
as the ability to add on third party software right from drivers down to
user utilities, all with reliability and predicatability in a
cross-platform environment

We're just now finally getting some of the features necessary to build
production quality systems in NetBSD -- it's taken decades to bend some
*BSD weenies into understanding what it takes operationally to run true
production systems.

>   What I will mention, though, is that Sun had NO part in the
> development of SVR4.  It was new around 1988/1989...before it was even
> an itch in Sun's pants.

Oh, how wrong you are!  ;-)

You appear to have fallen victim to the AT&T propoganda campaign to try
to distance themselves from their involvement with Sun in the
development of SysVr4.

As a matter of fact Sun did large amounts of the work on SysVr4 on
contract to, and in partnership with, AT&T....  I had many friends
working at AT&T (and some later at USL) and I have heard all the gorry
details first hand....

One of the most visible parts of SysVr4 that Sun did almost entirely on
their own was OpenWindows, which was the original official SysVr4
desktop interface, but their influence can also still be seen in many
other parts of the system, especially in the kernel, the sockets API for
TLI, NFS and other RPC goodies, etc., etc., etc., etc.  Sun may even
have had some fundamental influence on ELF and SysVr4 shared libraries,
though I don't remember their full origins.

I was porting some major applications and systems software to dozens and
dozens of different Unix platforms in the late 1980's and early 1990's
and I had early access to many versions of SysVr4 at that time.  Though
I didn't work directly at either Sun or AT&T, I had in some ways a
unique view of industry opinions and events since I was in many vendor's
porting labs during that time, and I had some pretty high-level access
to quite a few of each of their developers too (esp. at Pyramid and IBM,
and unofficially at AT&T, though also at Unisys, Fujitsu, Sequent, &
NEC).

The Sun/AT&T partnership, and the resulting view by other vendor that
they were going to be dragged down a path that they feared Sun and AT&T
would both have a massive head start on, was the final straw in the Unix
wars -- the thing that drove so many other vendors off the deep end and
caused many of them to scuttle Unix more surely than any of their
competitors ever could have.  The backlash can still be felt today,
though things have settled down considerably since those days.

AT&T's response was of course to run away from Sun like a scared puppy
and to eventually split off Unix into a "separate" entity (UNIX System
Labs, which of course more or less died not long after getting SysVr4
re-ported to a bunch of other hardware platforms and finally releasing
4.2).

Did you know, for instance, that the official sparc port of SysVr4 was
done for AT&T not by Sun, but rather by Fujitsu (who were back then a
big Sun partner too! ;-), because of their split?  IIRC the official
MIPS port was done by NEC, though the m68k and m88k ports were done by
Motorola.  AT&T of course did the 3b2 and i386 ports.

Since then the ownership of the official Unix source (and trademark) has
of course gone spinning in circles so far and so fast that the only
thing I've cared about it since is that SCO finally released the
original v5, v6, v7 and 32V sources.

In the mean time Sun's been slowly doing to SunOS-5 what IBM did with
AIX (especially in the 3.x.x releases, but again after taking over all
development from their contractors and their partner, Apple, since 4.1).

IMNSHO part of Sun's problem was that it didn't take the released
version of the SysVr4 sparc porting base and "start over" with it, but
instead they kept the source tree they had before AT&T cleaned it up and
passed it around to independent vendors to create the official
multi-platform porting bases.  The SunOS-5 kernel wasn't really up to
snuff again until, IMNSHO, 5.5.1.

As a result Sun even had to do their own MP support (SysVr4.2 MP was
superior in many folks eyes).  I've even talked to Sun developers who've
bitched at having to rewrite various bits of the AT&T code that they
didn't seem to realize their colleagues had probably written, and which
other AT&T partners had long ago re-written in the official SysVr4
releases too.

Sun is as much responsible for the medium term downfall of Unix as it
has been more recently for its rise again.  Certainly Sun is one of the
very few remaining "big iron" Unix vendors that still do their own OS
R&D, but ever since the AT&T split they've been pushing Solaris as
effectively a proprietary OS, at least in relation to other Unix
systems (but then they've always done that right from almost day one).

At least Apple finally got the right idea and have taken Mach and *BSD
and built what I think could be the next great workstation OS (though
they too have had to do a lot of work to bring it up to "production"
quality).  (Heaven forbid they'd stayed with IBM and continued
developing AIX though!)


All this said though, I still say SysVr4 was the best thing to happen to
SunOS.  Certainly SunOS-4 already had many of the early kernel features
that were to appear in SysVr4, and even some of the user-land stuff, as
well as stuff taken from AT&T, such as RFS.  SysVr4 was, after all, as
much Sun's brain-child as it was AT&T's....

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods at acm.org>     <woods at robohack.ca>
Planix, Inc. <woods at planix.com>;   Secrets of the Weird <woods at weird.com>