About the Amiga Unix Distro

[AmigaFreak]

I've never really had a chance to mess around with the UNIX distro on the Amiga.

Also... I never really heard much about it. What is it like compared to UNIX System V, or any other UNIX-like OSes like Linux?

How does it compare, if at all really to AmigaOS? - Who here uses it often?... and I'm curious to know if you can dual boot Amiga UNIX and AmigaOS on the same system.

[Florida]

Try check out this web site: http://amigaunix.com/tiki-index.php
 
From Wikipdia:
Commodore-Amiga, Inc., in 1990, did a full port of AT&T Unix System V Release 4 for the Amiga computer family (in addition to the proprietary AmigaOS shipping with these systems by default), informally known as Amix. Bundled with the Amiga 3000UX, Commodore's Unix was one of the first ports of SVR4 to the 68k architecture. The Amiga A3000UX model even got the attention of Sun Microsystems, though ultimately nothing came of it.[citation needed]
 
Unlike Apple's A/UX, Amiga Unix contained no compatibility layer to allow AmigaOS applications to run under Unix. With few native applications available to take advantage of the Amiga's significant multimedia capabilities, it failed to find a niche in the quite-competitive Unix workstation market of the early 1990s. The A3000UX's price tag of approximately $7,000[citation needed] was also not very attractive compared to other Unix workstations at the time, such as the NeXTstation ($5,000 for a base system, with many times the number of applications available), the SGI Indigo (starting at $8,000), or the Personal DECstation 5000 Model 25 (starting at $5,000). Sun, HP, and IBM had similarly priced systems. The A3000UX's 68030 was noticeably underpowered compared to most of its RISC-based competitors.[citation needed]
 
Unlike typical commercial Unix distributions of the time, Amiga Unix included the source code to the vendor-specific enhancements and platform-dependent device drivers (essentially any part that wasn't owned by AT&T), allowing interested users to study or enhance those parts of the system. However this source code was subject to the same license terms as the binary part of the system - it was open source but not free software. Amiga Unix also incorporated and depended upon many open source components, such as the GNU C Compiler and X Window System, and included their source code.
 
Like many other Unix variants with small market shares, Amiga Unix vanished into the mists of computer history when its vendor, Commodore, went out of business. Today, Unix-like operating systems such as Minix, NetBSD, and Linux are available for the Amiga platform, but the commercial and AT&T-licensed Amiga Unix has not been revived.

[mongo]

I've never really had a chance to mess around with the UNIX distro on the Amiga.

Also... I never really heard much about it. What is it like compared to UNIX System V, or any other UNIX-like OSes like Linux?

How does it compare, if at all really to AmigaOS? - Who here uses it often?... and I'm curious to know if you can dual boot Amiga UNIX and AmigaOS on the same system.

It is Unix System V Release 4.

[olsen]

I've never really had a chance to mess around with the UNIX distro on the Amiga.

Well, Amiga Unix was an expensive product. Few people managed to use it, especially since you needed custom hardware to install it (one or two different models of tape drives could be used for that purpose), and either an A3000 or A2000/A2630 to run it.

Incidentally, this was not a "distro", it was the official AT&T Unix System V Release 4, kernel and userland, when this sort of thing still meant something

Also... I never really heard much about it. What is it like compared to UNIX System V, or any other UNIX-like OSes like Linux?

As this was the actual AT&T Unix, it was pretty much what you could expect of the industry standard at the time. As the kernel was specifically tailored to match the hardware, it did take good advantage of what was available. It had multiple screen support (a feature Linux borrowed: if you are familiar with the Alt+Function key switching between TTYs and X11, now you know where it came from), support for different screen resolutions, even for the A2024 high resolution greyscale monitor and the A2410 framebuffer. The A2232 could be used to hook up serial terminals to it, and you'd use the A2065 Ethernet card to connect the machine to your LAN. For its day all this was fairly decent, and compared to what Sun's offerings could do at the time, Amiga Unix with all the hardware installed was significantly cheaper.

Commodore's Unix developers shipped the kernel in binary form, but provided the source code of the Amiga-specific bits. So you could hack your Amiga kernel, within limits, and extend it, too. Pretty great for its time, if I may say so. Did I mention that Amiga Unix could be installed in dual boot mode? You could run AmigaOS and Amiga Unix on the same machine, and you'd decide which operating system to use through the system boot menu. Pretty cool for its time.

Sadly, all of this came to an end very quickly when Commodore realized that such a great product must be killed at all cost.

How does it compare, if at all really to AmigaOS?

It compared to AmigaOS like any Unix or POSIX system did and does. At the time this was something very special. A real Unix system, not like for example Apple's A/UX offering. Amiga Unix was on par with, if not better than what Sun or AT&T themselves had to offer at the time, at the same price.

Who here uses it often?... and I'm curious to know if you can dual boot Amiga UNIX and AmigaOS on the same system.

Truth be told, Amiga Unix is hard to use today. There is no DHCP support, for example and no ssh. And because the kernel was last updated around 1991, you're stuck with plenty of buggy and functionally restricted code, the X11R4 is both old and buggy (it leaks memory), the 'C' compiler is positively ancient and you probably won't find a single configure script that still works. In a way, Amiga Unix is frozen in time like some fly stuck in amber

[Pentad]

Sadly, all of this came to an end very quickly when Commodore realized that such a great product must be killed at all cost....
...Truth be told, Amiga Unix is hard to use today. There is no DHCP support, for example and no ssh. And because the kernel was last updated around 1991, you're stuck with plenty of buggy and functionally restricted code, the X11R4 is both old and buggy (it leaks memory), the 'C' compiler is positively ancient and you probably won't find a single configure script that still works. In a way, Amiga Unix is frozen in time like some fly stuck in amber

Olsen,

That was a great overview!   I could not agree more with your statement "must be killed at all cost".  It was a great product at a great price so naturally Commodore rushed to destroy it.

I didn't realize that it didn't support DHCP but I assumed SSH was not there.  It is a real shame since it was such a great product.

I would like add to Olsen's great overview:

The documentation was just grand.  It had lots of books, a poster, and was written entirely in AMIX.  They even put that in the beginning of the book to show how powerful it was.

For those that don't remember, getting a real UNIX based workstation was very expensive.  The Amiga 3000 UX was used and sold to many CS students at the time because it was so good for such a low price.

I installed AMIX on my 3000 when I was in college and people were shocked that you could dual boot so easily with such a machine.  It really was quite amazing.  I did not have a network connection so I couldn't put it on the network.

SUN wanted to sell Amiga 3000s as their low-end entry into SUN Workstations which would have really boosted Commodore and the name.  Of course, Commodore made sure the deal fell through.

Atari also put out a UNIX based ST (TT030 Unix) because of the 3000UX.  Though I've never seen the TT030 Unix.

It's hard to think of what could have been...

-P

[nicholas]

I wonder how painful it would be to get AROS-hosted running on AMIX?

[Matt_H]

I wonder how painful it would be to get AROS-hosted running on AMIX?

Given the hardware involved, you might as well run the 68K build.

Of course, if you're trying just for the adventure, I'd rate the challenge as beyond extreme

[B00tDisk]

Given the hardware involved, you might as well run the 68K build.

Of course, if you're trying just for the adventure, I'd rate the challenge as beyond extreme

Toni Wilen was working on something equally crazy - adding FPU support to WinUAE to permit m68k Debian to install on WinUAE.  I'm unsure of the status of that side project.

[nicholas]

AROS-hosted on 68k Linux would be a hack to behold for it's absolute genius though!

[Failure]

olsen's description is excellent.  My two cents: I have an AMIX 2.1 installation that I used to try and get to run modern software for kicks.  I'm more or less a total hack when it comes to getting things to build...but I did have some success including running ssh and building newer C compilers which enabled yet more software to compile.  There were also more weird bugs than I have ever seen.  Half the challenge was virtually *nothing* expects to encounter sysv4 these days -- with good reason -- so all kinds of things bomb unexpectedly and require gentle nudges to direct them at sysv4-style workarounds.  An old O'Reilly book on System V programming was really helpful.  And lots of patience...25MHz and 16MB means that many builds basically thrash around in swap space.  A configure script might run ten minutes or more before finally bombing with some error.  And when that managed to finish, no guarantee it's actually going to build...what, and I suppose you want a pony too?

It's fun to play with but that's where it probably stops these days.  From a modern user's perspective most would find it formidable or, more likely, just unusable.  It's difficult to install and out of the box you get broken networking, DNS that only works with "some assembly required", a clock that can't figure out what year it is unless it's patched, a buggy, closed kernel, etc.  In a world where even Debian can be installed on x86 architecture by literally clicking "next" until it's complete, AMIX is a dinosaur.

Interesting side note, the system I run AMIX on is a superkickstart Amiga 3000, so I actually have it set up to quad boot if memory serves: WB 1.3, 3.9, AMIX and NetBSD 5.

[nicholas]

Is it compatible with 040/060 accelerators?

[Matt_H]

Is it compatible with 040/060 accelerators?

Nope. 020 and 030 only.

[nicholas]

Nope. 020 and 030 only.

That's a shame.

[olsen]

olsen's description is excellent.  My two cents: I have an AMIX 2.1 installation that I used to try and get to run modern software for kicks.  I'm more or less a total hack when it comes to getting things to build...but I did have some success including running ssh and building newer C compilers which enabled yet more software to compile.

You're a true hero, I'd like to let you know

I have a fully packed Amiga Unix system in storage which I used to toy and experiment with. Bootstrapping a 'C' compiler was always on my list of interesting things to try, but the effort never got off the ground. It wasn't just the 'C' compiler, but the entire toolchain that would have to be bootstrapped, and that's an awful lot of work...

I also own a NeXTstation which was built in the same year as my A3000UX. Same situation there: although it runs NeXTSTEP 3.3, which Apple used to support up until 1995, if memory serves, there is practically nothing you can install on the box in terms of modern software and be very disappointed. And this was a BSD-ish system, with a powerful CPU (back then NeXT was one of the few which used the 68040 in Unix workstations), plenty (32 MBytes) of RAM and a fast memory/SCSI interface.

There were also more weird bugs than I have ever seen.  Half the challenge was virtually *nothing* expects to encounter sysv4 these days -- with good reason -- so all kinds of things bomb unexpectedly and require gentle nudges to direct them at sysv4-style workarounds.

svr4 weirdly did survive in commercial systems made by SGI, Sun and others. But it became the odd one out as the BSDs evolved and of course Linux became what it is today.

The bugs in the Amiga Unix kernel were unavoidable. svr4 was so new back then that it hadn't been used much outside of AT&T's domain. Commodore managed to ship two updates to Amiga Unix before the product was killed and the Unix support group was disbanded. More work would have been necessary to keep the product alive and running.

An old O'Reilly book on System V programming was really helpful.  And lots of patience...25MHz and 16MB means that many builds basically thrash around in swap space.  A configure script might run ten minutes or more before finally bombing with some error.  And when that managed to finish, no guarantee it's actually going to build...what, and I suppose you want a pony too?

This sounds so familiar... But back in 1990 this wasn't such a bad system. It's the software which has outgrown its host.

It's fun to play with but that's where it probably stops these days.  From a modern user's perspective most would find it formidable or, more likely, just unusable.  It's difficult to install

I still think the installation script worked much better than what you could expect at the time. Remember the first few release versions of NetBSD and their installation procedures? Now that was difficult to cope with.

[olsen]

Is it compatible with 040/060 accelerators?

Funny that you'd ask, because that was a major bone of contention back then. Not only did Commodore manage to kill the product, they killed it at a time when the introduction of the '040 could have really made it into something worthwhile, had Amiga Unix supported this platform. The 68040 used to be quite expensive when it was introduced and few machines were built which used it. But at the time it was binary compatible with the entire 68k family, for which a lot of software (especially for Sun systems) still did exist, and it was faster than the SPARC CPUs used by Sun workstations in the same price range.

Now this could have really been something... But there's one problem: the 68851 PMMU, which was merged into the 68030, allowed for very compact MMU tables, but the integrated 68040 MMU did not. You'd burn an awful lot of RAM for setting up the 68040 MMU, and the MMU table format was not even compatible with the format used by the 68851/68030. This required major rewrites and redesign of the virtual memory subsystem in your Unix kernel. This was a major stumbling block, and back at the time few vendors rose to this occasion.