From Google Groups comp.sys.tandy
Nov 13, 2001
Frank Durda IV
Bruce Tomlin wrote:
: IIRC, Microsoft did sell Xenix to SCO, but that was (long?) after the Model
: 16/6000 version of Xenix came out.
The history is a tad more complicated than that. Let me start with a
section from the History of Tandy Computers I wrote some time back
(this is from a larger document, which is why there are references to
things not covered here):
Model 16 and UNOS (September 1982 - December 1982)
By the end of the
summer, it was obvious that TRSDOS-16 (now known internally
as "Bowling-ball DOS") was a complete turkey, and a replacement had to be
found. Machines were being returned and a well-publicized lawsuit appeared
in the press regarding non-performance. Since the person who had been
pushing for the real-time system had given-up and quit, it was decided to
adopt the UNIX-like plan proposed a year earlier. The choice was UNOS, a
UNIX-like system from Charles River Systems. Development systems were
obtained, and the programmers were busy writing drivers and doing the
At the time, AT&T was not allowed to sell computers by law.
They were allowed to license software, and that is what was done with
UNIX. However, AT&T did not allow potential competitors to call it UNIX,
as AT&T felt that one day they would be allowed to sell systems, so they
kept the name for themselves. So the various names like UNOS, BSD, ULTRIX
and XENIX appeared on the scene.
I mentioned XENIX. In the early fall, Microsoft asked Radio
Shack if they
would be interested in having XENIX for their Model 16, since it was
obvious to everybody that TRSDOS-16 was doomed, but the hardware had good
potential. Radio Shack was now intent on the UNOS course, basically told
Microsoft to drop dead. In theory, Microsoft didn't know about the UNOS
Some weeks later,
Microsoft asked for technical details for the Model 16 so
they could evaluate using it as a XENIX platform. Apparently they wanted to
look into marketing XENIX on this platform, even if Radio Shack didn't
sell the package. Once again, Radio Shack told the Bellevue boys to
Then, at a meeting on unrelated topics, Microsoft demonstrated a barely
functional port of XENIX running on a Model 16. Before Radio Shack could
get mad, Microsoft said: "We can have a shippable product in 12 weeks, and
we'll even throw in BASIC and MULTIPLAN for free on the deal". Radio Shack
wavered a bit, but than stuck to their guns, saying they were already
committed to the UNOS work and it was nearly complete, which was quite true.
Microsoft then used a tactic we would all see later as a Microsoft standard
business practice: Microsoft said that if Radio Shack didn't buy XENIX and
went with UNOS, Microsoft would not do a port of BASIC or Multiplan for UNOS,
and future releases of Multiplan and BASIC for all other Tandy platforms
might be "delayed". What just about anyone would call "a threat" caved-in
whatever resistance remained in Radio Shack computer merchandising, and
after a bit more wheeling and dealing, Radio Shack agreed to buy XENIX and
to abandon all work on UNOS.
pity, because the people working on UNOS didn't know anything about the
Microsoft maneuver, and were actually allowed to work for some additional
weeks before being told that UNOS was dead. In fact, UNOS was less than
a week from testing when it was killed-off. After all the programming
staff stopped banging their heads against the wall in frustration, they had
to assist Microsoft in writing XENIX drivers that would actually work on
more than the two machines they had at Microsoft. Finally, in early winter,
XENIX arrived in the testing labs. And that was just one thing to keep us
That's the preface on where Tandy was in all this, now on to the saga of
the bouncing XENIX.
In late 1982, once Microsoft got Charles Rivers UNOS kicked out of Tandy and
got XENIX put in its place there was a period of time where Tandy pretty-much
had to fix many many problems with z80ctl (the I/O driver layer of XENIX for
the 16/6000), which Bob Powell of Microsoft originally wrote based only on
the Model II Technical Reference. That means he had zero real-life
examples, didn't know about all the hardware flaws you had to dance around,
and had to go from scratch in 90 days, so the achievement was significant.
However, the result was simply too unstable for any use beyond the demo
presentation. The system could and would actually start fires in the video
section of the machines in certain crash conditions.
A couple of people at Tandy did surgery on Bobs work (KarlB, RonL, JEIV) in
those early days, eventually getting it so it didn't crash every time you
tried to do two things at once, but because the fundamental design was
rushed, it had many basic flaws. This was where XENIX stood around the
original release version 1.0.x in early 1983. It also helps explain why
there were at least ten releases in the first year of that product.
XENIX also shook this and all
subsequent computers much harder than any
other OS, so any timing problems or noise problems that had remained hidden
or intermittent for years suddenly were persistently fatal problems.
In the fall of 1983, Microsoft violently and abruptly
(probably right after Steve Jobs gave them a sneak peek at the Macintosh).
Microsoft abruptly turned control of 68000 XENIX development to SCO, who
was already a licensee unprotected memory version for the 808x platform,
but Microsoft now turned the 68xxx processor work over as well, including the
completion of a XENIX version based on AT&T UNIX System III. It isn't clear
how far Microsoft had gotten on this before it all went to SCO, but Microsoft
had been working on it for a while.
Tandy was already mad because Microsofts working arrangement on XENIX was
pretty difficult, with no MS source code (other than the original UNIX code
that we obtained directly from the BSD 4.1 and AT&T V.32 releases we had to
license), apart from a one-time release of some of the source code and some
disassemblies generated by a fairly powerful disassembler that the late
Ron Light wrote to combat what Microsoft had been doing even further back
in the Model I era.
Only when Microsoft started losing interest in XENIX and lobbed the ball to
SCO did Tandy finally obtain up to date source code to all parts of XENIX.
Before that we had code but each new release from Microsoft contained only
executables, leaving our old copies of the source code somewhat useless,
and causing older fixes to habitually disappear if Microsoft didn't remember
to roll them into their source tree. Now that Tandy had control of the Z80
side of things as well as all of the 68000 code, Tandy was finally in a
position to fix long-standing bugs and flaws in the earlier work, mainly by
throwing-away most or all of the Microsoft code and writing new stuff.
When Microsoft was getting ready to lob the ball to
SCO, Tandy suddenly found
themselves unable to get help from Microsoft for their V7-based XENIX
because Microsoft had moved to another project (Windows), and SCO wasn't
that interested in doing V7 anymore with III in the works. Tandy did get
full source code at this point for the V7 version (now that Microsoft didn't
care about it anymore, and we set about fixing the bugs Microsoft ignored.)
In working with SCO, we found that SCO had a version of SCO XENIX that worked
on the Apple Lisa, but who SCO did that work for (if anybody) is not clear.
(I saw a couple of those systems running when I visited SCO in August of
1984 to write the initial XENIX bootstrap work for the Model 2000 80186 port.)
So while Tandy/SCO was
getting XENIX 3.0.0 ready, IBM introduced the IBM AT
with its 80286, which meant protected code was finally possible on the PC
platform. Yeah you couldn't do paging, but neither could the 16/6000.
SCO started work on a port to the 80286 AT
platform, and suddenly Microsoft
decided XENIX-286 (as they would call it) was going to be an OS that IBM
would buy a lot of copies of. Microsoft was very scared of any measurable
part of their OS sales ending up at SCO (even though Microsoft still got a
royalty), so Microsoft initially started contributing more heavily to SCO
for the 286 platform (like providing a beat-up version of Microsoft C 4.0
that ran on XENIX), and then Microsoft abruptly grabbed the controls and took
over all porting work and OEM distribution management for the 80286 platform.
SCO was somewhat left out in the cold.
At Tandy that
was very annoying, because it meant that IBM would get the
code first if Microsoft was involved, and SCO was a LOT easier to work with
than Microsoft. SCO would let us have source code that we had a right to
see anyway! Despite this, a significant "race" developed, with the goal of
getting XENIX 286 out for our platform before IBM had theirs ready, even
though we were getting our stuff from the same place (Microsoft). (Whether
IBM was aware of this race or not is unlikely.) The only things different
between the Tandy and IBM verions would pretty much be the point at which
each of the two companies stopped accepting new code fixes from Microsoft,
the locally-produced drivers and any applications, and the documentation,
which each organization got from Microsoft and pretty much had to clean-up
because it was ugly, wrong, etc.
Then one day, we got word that Microsoft was going to bring out a AT&T UNIX
System V port for the 80286 that Microsoft was working on, and it was decided
at Tandy to shelve the System III work done so far (probably 90% finished)
and focus on getting the System V version before IBM. Microsofts System V
"port" barely ran when it first arrived (of course, it barely ran on the PC/AT
too), and as before we didn't get source code to the kernel and other
components, so we had to rely on Microsoft to find and fix bugs that we
uncovered. In general, it was an ugly nine+ months. As it turned out,
IBM beat us to the stores, but when we got a copy we found that they simply
stopped accepting fixes from Microsoft earlier, and they barely touched the
Microsoft documentation, leaving Tandy with a far more stable version with
a lot better documentation. Still, we got punished for not beating IBM.
By the time the 80386 came along, Microsoft had
changed their minds again
and decided that IBM wasn't going to buy that many copies of XENIX, so they
dumped the entire XENIX world (now just 80286 and 80386) back to SCO,
and the XENIX 386 project was pretty much an all-SCO thing, much like the
XENIX System III 68000 port.
Microsoft never stepped-back into the XENIX world after 1987 in any
significant way, which is probably why from 1987 through 1991 or so, SCO did
pretty well and XENIX saw some pretty decent growth and support.
Microsoft probably could have yanked XENIX back from SCO at any point again
if they thought it was worth it, but they never came back.
As a point of trivia, the 1982 royalty Tandy paid to Microsoft for XENIX was
$85 per copy, and that included everything, along with BASIC and Multiplan.
Tandy broke that into several chunks, so you would pay several hundred
dollars if you wanted the CP, CT, S, and other manuals and software that
made up the complete XENIX system, plus the BASIC and Multiplan packages,
even though Microsoft got all their money up front. Later XENIX versions had
similar per-copy royalties for the complete system but were sold by the
retailers in pieces.
Frank Durda IV - only this address works:|"Skip
Stritter sucks irregular,
<uhclem.dec01%nemesis.lonestar.org> | deformed, banana skins". - An
| actual comment in the Microsoft
This Anti-spam address expires Dec. 31st | XENIX port of the PCC compiler.
Copr. 2001, ask before reprinting.