I have to agree with Ken on all of that. We built tools for all the Unix
platforms out there at the time. A/UX was the most painful.
On Jul 11, 2014 1:55 AM, "Ken Seefried" <seefriek at gmail.com> wrote:
I wrote this a while back, but found it in the drafts
bucket so I
apparently never sent it...
A/UX has (had?) its warts...
Warts like a toad with skin cancer.
It was based on SVR2 (another quick-n-dirty port from Unisoft, the goto
guys for "I got an mc68k, I need to ship Unix next week and the budget is
tight"), which was already 4 years old when A/UX was released (SVR4 shipped
a couple of months after A/UX). Various bits of SVR3 and BSD were bolted
on with different degrees of frankensteinisms. Some apps were newish
versions, some were ancient R2 versions (like UUCP, important at the time),
so moving between Unixes was frustrating. The native dev tools were pretty
awful, networking was flaky (NFS was particularly fun in a dev environment)
and the filesystem seemed to be moderately unstable for most of it's run.
It didn't actually implement all of POSIX (or a bunch of other things) so
porting software was always an adventure. And whenever I tried to use some
neeto-keeno Apple-ism slapped on the top, I was always tripping over some
new-and-different showstopping bug. That's just what I remember 20+ years
down the road.
So, yeah...pretty warty.
The company I worked for at the time hacked our TCSEC B-level technology
into that mess so that Apple could extend the "see...we got POSIX" line to
"see, we got Orange Book evaluated" so they could sell to military &
intelligence customers. So lot's of crawling around in the kernel.
Not pretty. But enough sold to justify the contract.
Not the best UNIX implementation compared to what
we
have nowadays but in light of its contemporaries it was
probably no worse than average.
When A/UX came out in 1988, it went up against the likes of SunOS 3.5, Sony
NEWS-OS 3, SGI IRIX 3 and Ultrix 4. I would argue every one of them is a
much better Unix implementation than A/UX. So much better it's really not
even a contest. My daily-driver-at-the-time Sun 3/160 w/ SunOS 3.5 was so
much faster, functional and stable it was physically painful to think "time
to muck with A/UX". Even SVR3.2 was more consistent and stable, if not as
feature-filled (and I really thought layers+dmd5620 was wonderfully clever,
but that's a wild tangent).
To really get to contemporary levels of crappy, you've got to go down to
Xenix. And unlike A/UX, Xenix could run Microsoft Word well.
When A/UX was finally put out of it's misery in 1995, it still ran only on
mc68k, still based on SVR2.2, and was up against Solaris 2.4, OSF/1, Irix
6, UnixWare 2, AIX 4, Ultrix 4.5 and eminently usable versions of Linux and
*BSD, all running on platforms a order of magnitude or two faster than what
A/UX ran on.
The Mac OS GUI was still single-threaded,
cooperative-
multitasking, no protected-memory System 7.0 with some
application compatibility issues... so the benefits of UNIX
weren't really extended throughout.
The much vaunted MacOS emulation layer didn't really show up till version
3, and didn't work particularly well if you didn't have the latest versions
of your applications from mainstream vendors. More than "some" application
issues (as in anything that wasn't written 32-bit clean was a coin toss you
usually lost).
That all said, I don't see why A/UX would
have been any less
suitable a foundation for Apple's next-generation operating
system than NeXTstep?
Mercifully no one at Apple even considered going down that road.
A/UX existed so Apple could sell hardware to the Federal sector. Even
though the A/UX team looked around for other markets, it was a checkbox
product and it never had any more mindshare with Apple senior management
than that. And it shows.