On Jun 11, 1:55 pm, Ian Hilliard <nos...@xxxxxxxxxxxxxxxx> wrote:
> RexBallardwrote:
> > There is a distinct possibility that when IBM had the discussion about
> > an Operating System, and Bill Gates said "We have an operating
> > system", IBM may have thought that Bill was talking about Xenix.
> > Microsoft had introduced Xenix for the Radio Shack Tandy 6000
> > computer. The only problem was that the Tandy computer was based on
> > the 68000 processor, not the 8086 or 8088 processor.
>
> Actually Xexix originally ran on the Tandy Model IIb. Unfortunately there
> were a lot of hardware and software problems with Tandy's Xenix offerings.
>
> The Model 6000 was quite an improvement. It still used a 68000 main
> processor and Z80 I/O controller, but the main I/O was on the main board
> and the BUS was physically more stable. The Model 6000 also had room to
> mount the hard drive internally (originally 5MB). Unfortunately, Xenix was
> still very unstable even with matched memory cards and all the recommended
> mods to the hardware, of which there were many.
The biggest problem was that there was no MMU. Xenix and OS/9, at the
time, both ran on unprotected memory. Eventually, the 68010 would
have the "redo" functionality required to recover from a memory fault,
and even the 80286 mask would be fixed. The Zilog Z8000 chip set
included an MMU, and ran UNIX very well. Jerry Pournell (sic) had
Zeke, his Cromemco computer that ran ZEOS, a Unix variant designed for
the Z-80 with special MMU chip.
Tandy did offer OS/9 for their 6809 based Color Computer. It was
pretty impressive, given that the hardware was so limited. It was
quite a contrast to BASIC in ROM.
> It was probably for good reason that Microsoft didn't offer Xenix for the
> original PC. Xenix had a lot of issues and the clunky 8088 processor, with
> its stupid memory model to keep it compatible with 8085 code,
There were many people who felt that the dumbest thing Bill Gates had
done up to that time, was to opt for a 4 bit offset and accept a 1
megabyte memory map. The 80286 was shifted to 8 bits (when not
running in "Real Mode". And the 80386 came with a built in MMU.
Ironically, the MMU wasn't that hard to build in hardware. Many
companies did build very effective memory managers for either
segmented or demand paged memory. Remember, AT&T Version 6 and
Version 7, as well as BSD 2.x used segmentation for memory
management. Demand Paged virtual memory was introduced to UNIX by the
BSD team.
> would have
> made Xenix even more unstable. Xenix didn't become all that stable until
> the original SCO got it running on the 386 using a flat memory model. As
> far as I can remember, the first version of SCO Xenix ran on the 286.
Xenix was never all that stable - period. Later more stable versions
were based on UnixWare and other variants of the original AT&T System
V and BSD kernels. There is also the possibility that the SCO team
may have peeked at the Linux play-book even before the Caldera take-
over.
SCO had some great technology, but after the IBM debacle, nobody wants
to go near it.
Linux does have a SysV API compatibility library, but it simply
imlements the public standards set by X/Open. The implementation is
radically different from AT&T's solution.
> Ian
|
|
