Hacker News Reader

Comment by rbanffy | Hacker News Reader

At the time Commodore made PCs, it could have been worthy to port AmigaOS to x86 (8088 was a bit too puny, but 286 and 386 could compare to 68K). An earlier lost opportunity was when Sun approached the offering to market the 3000/UX as a low-cost Unix workstation.

The world could be a lot different had Commodore management been so boneheaded.

OK. I'm being harsh. At the time, with the information they had, those decisions that now appear stupid could make sense.

View on HN

By the time x86 was really competitive architecturally with 68k -- by which I mean non-segmented, 32-bit, etc -- it was already too late for Commodore and Atari. They were already pretty much dead financially.

Both sold PC clones, and I suspect if those lines had been very successful $$ wise they may have attempted to do something like what you said. Or more likely some kind of Amiga-esque runtime environment overtop of (choke, gag) DOS. We would have all screamed bloody murder about it though.

In any case, AmigaOS and Atari's TOS were pretty heavily tied into the 68k ecosystem. Switching ISAs would have really been starting over.

Moving from 68k to PowerPC almost killed off Apple, and they were many multiples the size of Commodore & Atari.

Myself I made the switch from 68k Atari ST to 486+early-Linux in 1992. At that point we were already doing Unix-like things on the Atari (MiNT with unix ports, or, on the 030/020 systems, full NetBSD Unix), so it felt like a natural jump. SVGA + a 486 was a pretty damn competitive workstation-like system at that point.

I dont think the platform move was.. a cause of Apple's troubles, it was mostly poor product planing, marketing and OS related issues (which were not really made better or worse by the ISA move)

I don't think it was a unitary cause, sure. But I think in retrospect it's clear Apple could not afford this kind of ISA move in addition to all the crazy OS work they were doing (and never finishing). It was half-baked until OS 8.

In 1996 I helped administrated a group of MacOS boxen, all PowerPC. At that point in time only half the applications were running native PowerPC, and it was buggy as hell. If I recall, even parts of the OS were still running 68k emulated.

In an OS with no memory protection or even proper dynamic memory allocation, it was just a shitstorm. Could not run FileMaker and Netscape at the same time without dying all over the place. I was shocked, because a couple years earlier I remember finding the 68k Macs I used extremely stable, even with Multifinder, etc. Limited, but stable.

IMHO Apple was lost in the wilderness for most of the mid-90s and a big part of it was just having too many balls in the air.

But they had to do _something_, staying on the 68k wasn’t an option. It couldn’t compete with the Pentium and Motorola had effectively abandoned the series, focusing their efforts on the PowerPC.

Apple made a lot of bad decisions in the 90s, but I don’t see how the switch to PPC was one of them.

I don't think it was a bad decision, it was inevitable they had to switch as Motorola left 68k customers in the lurch. The actual bad decision was not investing in better OS development in the 80s.

Put it this way: the Lisa had an MMU, memory protection, and an OS with better multitasking. Yes it was expensive to make, but they could have scaled it down instead of building a whole other, competing system that was technically inferior from an operating system perspective.

The original Mac was built as if it was a one-off product, and they lived with the results of that for two decades. And these were Jobs' choices.

> the Lisa had an MMU, memory protection, and an OS with better multitasking. Yes it was expensive to make, but they could have scaled it down

Yes and no.

I mean, the Mac is a cut-down Lisa. But you're right, cut down too far.

(Aside: the Lisa was a 68000 as well, I am fairly sure. No MMU, but it did have multitasking.)

I think the problems were in scaling it down far enough to get the price down to vaguely reasonable.

That meant:

• 128kB RAM (same as the Sinclair QL, which launched 2 weeks earlier)

• No hard disk. Keeping a full multitasking OS when running from 400kB single-sided double-density floppies -- indeed, from one floppy -- would have been significantly hard.

Before an outraged Amiga fan jumps in: yes, I know Amiga did it, but that was 2 years later, with 4x as much RAM and floppies holding 2x as much data. Tech was moving fast in the 1980s.

A 512kB original Mac with 800kB floppies was technologically doable in 1984, I suspect, but instead of costing $2500 when it launched (~1/4 of a Lisa), it would have been $5000 or something, and I strongly suspect it would have flopped too, as a result.

The Lisa had a custom MMU, if I recall. There were other 68000 based systems that were similar.

My key point is around the operating system. They had a system in LisaOS that had multitasking, memory protection, proper dynamic allocation, etc. It was a better OS than what was shipped on the Mac. It was unfortunate they threw it away. Likely due to personal & organizational politics more than technical and cost reasons (Jobs hated the Lisa). A few years later they were running Mac OS on old Lisa hardware.

> even parts of the OS were still running 68k emulated.

Up until MacOS 8.5 you could move some system extensions to 68K Macs and they'd work.

I had my Color Classic have MacOS 8 looks (even running on 7.5) that way.

Apple couldn't stay on the 68K though - Motorola was mothballing the 68K line and investing on their own 88000 and, being faced with an OS rewrite, Apple did the rational thing - find out who has the best chance of making something better.

> Apple was lost in the wilderness for most of the mid-90s and a big part of it was just having too many balls in the air.

I remember that, when MacOS 9x seemed like a dead end, Apple published MkLinux. It was my first Linux desktop.

> But I think in retrospect it's clear Apple could not afford this kind of ISA move in addition to all the crazy OS work they were doing

You might be surprised. Apple did an x86 port of Classic MacOS, based on DR-DOS.

https://lowendmac.com/2014/star-trek-apples-first-mac-os-on-...

That is not the problem. Even emulating the apps isn't the core problem.

Drivers, and support for the vast range of x86 hardware, was the problem, and secondarily, using the underlying hardware and supporting an MMU, memory protection, and preemptive multitasking... those were the big problems, and in the end, all Apple's attemps to solve them failed, and it moved to UNIX.

That worked, but it needed to happen when it became viable to run the entire "legacy" OS in a VM, and that didn't come until around the turn of the century. That was not feasible in 1992.

The 268 had a segmented memory model which effectively made it possible to only access RAM in 64K pages. The 68000 had a flat memory mode and other differences that would have made a port of AmigaOS to a 286 very difficult because of the downgrade in capabilities (video controller architecture is another huge issue, too). In many ways the 286 should have been awesome, but wasn't. The 386 was more of a 68000 series competitor, and I believe it came out about the same time as the 68030 - so by the time Intel chips could compete with Motorola, four years had passed (and in that era, four years was like two decades).

Indeed, a 286 would cause some breakage and some OS-level APIs would need more complicated code, and some programs would be difficult to port due to HGA, VGA, and SVGA graphics being very different from Amiga modes (although somewhat superior).

Still, would be fun to have a 386 version of the Amiga OS.

ISTM that if Microsoft had managed to add preemptive multitasking to Windows 3.x in a backward-compatible way, this would've given us as close to an Amiga-like OS as one could ever hope to run on 8086 and 286 hardware. Perhaps even more so, if hardware could be expected to have multimedia capabilities as a standard, without relying on 3rd-party drivers - PC had this on the PCjr, and later Tandy.

> ISTM that if Microsoft had managed to add preemptive multitasking to Windows 3.x in a backward-compatible way, this would've given us as close to an Amiga-like OS as one could ever hope to run on 8086 and 286 hardware.

8086 didn't have the instruction set to do preemptive, and the 286's memory model was 64kb pages... and barely good enough. The enabler for MPC (Multimedia PC) was the combination of 386 (and 386sx) + 16 bit sound card + CDROM + PCI VGA+.

>8086 didn't have the instruction set to do preemptive

Can you elaborate of that?

AIUI all you need is interrupts (CPU has them) and a timer (the PC platform has them).

You mean OS/2? :-)

> add preemptive multitasking to Windows 3.x

Just for accuracy: Win3.x had pre-emptive multitasking. However, it could only preempt DOS apps. Win16 apps could not be pre-empted.

This compares quite closely to the FOSS RISC OS for Arm machines, which can preempt CLI apps in a "task window", but GUI apps only multitask cooperatively.

Co-op multitasking is faster and more memory efficient. That's why Acorn and MS chose it. But it's very vulnerable to a single app failing to relinquish control, locking up the OS.

I say this with respect, but I don't think it would. The x86 architecture had an absolute paucity of registers, the mixture of MMIO and PMIO was insane, and it just didn't have any PC-relative addressing modes. Relocation for a flat address space would be an annoyance. Prior to PCs becoming about a billion times faster and the compact encoding coming into its own, x86 code is not fun to write.

Saying "some programs would be difficult to port" is putting it mildly; a huge part of the Amiga's software catalogue was written in assembler, and pretty much every single game wrote directly to the hardware, in a way that can't be abstracted without a complete emulation of that hardware. The Commodore graphics APIs were absolutely too slow to do anything except productivity software. There were retargetable graphics APIs (e.g. CyberGraphX, Picasso 96), introduced to allow well-behaved software to access expensive graphics cards you could buy for big-box Amigas. But generally you wouldn't use them for game graphics (unless it's a very simple game).

The Amiga really held well together; the custom chips were well designed to fit with the 68000. Perhaps too well, and that coupling became the downfall when the computer hardware moved faster than Amiga software could be brought along. It was destroyed once games became more about per-pixel operations (1990s 3D games) than blitting large areas (1980s platform games); the existence of VGA mode 13h made the former so much easier to write. Commodore even added a chip to provide a form of hardware-accelerated chunky-to-planar conversion, but it was too late, only in the CD32 and just couldn't compete with a native chunky framebuffer.

How did you grab that amiga386 username 32 days before writing this comment? Do you have a TARDIS in your backyard or something? Spooky!

It was the most incongruous noun-number combo I could think of, and a month later someone's genuinely proposing it...

> that coupling became the downfall when the computer hardware moved faster than Amiga software could be brought along.

Both hardware and software moved ahead while Commodore invested in the wrong things. Amiga was late to adopt VGA monitors and PCs quickly surpassed its capabilities. When I mentioned porting AmigaDOS and Intuition to 2/386's, it would be to keep the OS compatible at the C source level as much as possible, with workarounds for hardware when it's different, but always at OS API level so that non-game software (or non-Amiga-like games) could be ported from source.

I don't think it'd have saved Commodore or the Amiga (I think the Sun deal, where Sun would sell 3000/UX boxes was a better bet for that, or maybe not ditching the CBM 900 but, instead, building Amiga with Coherent instead of AmigaDOS).

Anyway, it'd have been fun.

> compatible at the C source level

I don't think most of it was originally implemented in C.

> at OS API level

That's the problem. Keep the original non-protected-memory APIs and you can't have memory protection; add memory protection, and it can't be API-compatible.

It's catch 22.

> I don't think most of it was originally implemented in C.

IIRC, AmigaDOS was written in BCPL and later moved to C.

> Keep the original non-protected-memory APIs and you can't have memory protection;

Kind of - you can always give the process the illusion it's running alone (or alongside the processes it started).

Could have packaged up the Amiga chipset on an ISA card, an all-in-one video/audio/io gizmo. Sell that to 386 owners and give the OS away. Bonus points for a ROM socket to insta-boot AmigaOS with no disk.

By the time they would've decided to do this, the Amiga chipset was looking pretty dated. 386's with SuperVGA and SoundBlaster cards were becoming common place in the early 90's. Once the 486 came out (1989), the price of 386 hardware dropped fast. Commodore seemed more interested in going the other direction: putting an x86 (BridgeBoard) into an Amiga.

I was an Amiga fan from roughly 1988 through 1994, started with an A500, expanded it, then moved on to an A3000. The platform was incredible. I learned a ton from it and taught myself C on that system. But by 1994, I wanted a Linux system, so a 486 it was...

Well look no further than AROS: https://aros.sourceforge.io/

It’s a reimplementation of AmigaOS 3.1 that runs on x86, ppc, and m68k.

It blew my mind when I was able to set the viewport pointer to where my program code was, and SEE THE CODE AS IT RAN. Like, watch the bits flip as counters incremented, etc etc.

It was a wonderful, impactful lesson in foundations of software engineering, before I even knew what any of that meant.

A lot of that "before I knew what that meant" going on for me looking back at that time. It seems like even though we have better technology now we somehow back slid into the worst uses for it.

Even enjoying a Commodore Amiga as a teenager we cannot forget Atari 1040ST competitor which included MIDI and was also innovative. Just read that was popular in some locations for CAD.

It is difficult to talk about failed business, business many many times are successful for reasons beyond the technology involved. Just look nowadays how organizations are using Microsoft Teams, they could be locked, have a high cost for migration, bad management, and the core business working seamlessly beyond this.

Built-in MIDI and the excellent 100Hz monochrome screen were very strong selling points for the ST back then, for different niches of the market. Especially given the strong audio capabilities of the Amiga, including a MIDI interface should have been a no-brainer for Commodore to position it in the music production industry. On the other end of the market, the always slightly flickering 1081/1084 color monitor made the Amiga not very suitable for an 8 hour office job.

Commodore had an extremely expensive, weird 15Hz flicker free and very high resolution monitor. If I recall correctly, i̶t̶ ̶d̶i̶v̶i̶d̶e̶d̶ ̶t̶h̶e̶ ̶s̶c̶r̶e̶e̶n̶ ̶i̶n̶ ̶4̶ ̶r̶e̶c̶t̶a̶n̶g̶l̶e̶s̶ ̶a̶n̶d̶ ̶p̶a̶i̶n̶t̶e̶d̶ ̶e̶a̶c̶h̶ ̶r̶e̶c̶t̶a̶n̶g̶l̶e̶ ̶a̶s̶ ̶a̶ ̶m̶i̶n̶i̶ ̶s̶c̶r̶e̶e̶n̶,̶ ̶o̶n̶e̶ ̶a̶f̶t̶e̶r̶ ̶a̶n̶o̶t̶h̶e̶r̶.̶ ̶T̶h̶a̶t̶ ̶w̶o̶u̶l̶d̶ ̶h̶a̶v̶e̶ ̶b̶e̶e̶n̶ ̶a̶ ̶l̶o̶t̶ ̶o̶f̶ ̶f̶l̶i̶c̶k̶e̶r̶i̶n̶g̶,̶ ̶e̶x̶c̶e̶p̶t̶ ̶t̶h̶a̶t̶ ̶t̶h̶e̶ ̶p̶h̶o̶s̶p̶h̶o̶r̶ ̶w̶a̶s̶ ̶v̶e̶r̶y̶ ̶l̶o̶n̶g̶ ̶g̶l̶o̶w̶i̶n̶g̶.̶ ̶ Edit: wrong, see comment below.

That would be the A2024 which is emulated by WinUAE. It did not flicker or have long persistence phosphor. What it did have was frame buffer RAM on-board. The video signal output by the Amiga determined which 1/4 or 1/6 of the screen was refreshed. The driver code had a feature where the pane containing the mouse pointer could be refreshed more frequently.

Whoa! That's a rudimentary compression scheme to decrease latency.

If you want to know the details of how it works, see US patent 4851826: https://patents.google.com/patent/US4851826A/en

BBoAH has a pic of an actual A2024: https://bigbookofamigahardware.com/bboah/product.aspx?id=863

There was also a separate video card containing the framebuffer hardware, that could be connected to a "normal" high-res mono monitor: http://amiga.resource.cx/exp/moniterm

There are brilliant hacks and terrible kludges. This is the latter.

Yikes!

Wow, I had no idea that the ST had 100hz modes. That's so cool.

Folks who grew up in the post-CRT era might not realize what a difference 100hz vs. 60hz could make with CRT monitors when you were staring at a screen for 8+ hours a day.

Not sure what the graphics controller was capable of, but the 'paperwhite' monochrome monitor which was offered then supported 72Hz refresh rate. That was a lot then and most people won't notice the remaining flicker, but it's not 100Hz. Op might be misremembering.

(I'm a bit sensitive -- around that time, I removed a connection on the MB of my Schneider, nee Amstrad, CPC 464 so that it would think it's the NTSC version and would output the screen's content @60Hz and adjusted my monitor accordingly. 60Hz was just about bearable, 50Hz (PAL) only with dark background)

This is true, the monochrome monitor only did a bit over 70Hz. I was corrected in another comment in this thread.

The ST in mono was alas only around 71hz. Still very nice. Sadly my family could not afford a mono monitor and and I used my ST through the family TV but I used a friend's mono monitor a few times and it was very nice.

Something I always thought odd; why would just _having_ MIDI ports make the ST popular with musicians? Getting an Amiga MIDI adapter was super cheap and easy... although most early Amiga MIDI and stave-notation software (Sonix, Music-X, DMCS) fell out of favour as trackers stormed on the scene; most Amiga musicians preferred to use the Amiga as a sample sequencer than as the controller of other MIDI devices.

Is it really just default ports? If the Amiga had had a line in port rather than required an external sampler, would it have been known as "the sampling machine"?

The Amiga did famously form the backbone of 1990s NTSC broadcast television production as the Video Toaster, but that was because the Amiga chipset was able to genlock with an external video signal, i.e. it was an inherent feature of the machine rather than the default ports and peripherals it came with.

Put yourself in the shoes of the developer of a midi sequencer in the 80s. You could target the Atari, where midi ports are standard and everyone with an interest in music is a potential customer. Or you can target the Amiga where there’s no standard, effectively forcing you to source or develop a midi dongle yourself to sell with the software, increasing the price by at least $20 or $30, and adding complexity and development costs up front.

That's a good point. Something being a noted feature of the device would encourage people to develop software for it. The DSP in the Atari Falcon was a big draw as well.

Commodore did eventually promote a standard library for MIDI (https://en.wikipedia.org/wiki/Commodore_Amiga_MIDI_Driver) in 1990, perhaps too late to take anything from the lead the 520ST had gained

640x400@70Hz monitor (not 100Hz as OP claimed) was the second part of the puzzle. You want as much readable non headache inducing (interlaced Amiga) text on the screen for music sequencing. Not to mention Atari ST _with_ that monitor was still substantially cheaper than naked Amiga 1000.

Whoa, I was very sure that the SM124 did 100Hz, but you're right. I stand corrected! And I agree that the combination of the great monochrome monitor and the built-in MIDI made the ST much better suited as a music production machine.

286 and 386 were NOT comparable to 68k. Much Amiga development was done in 68k assembly which was an absolute joy compared to x86, and mapped almost directly to C. Plus the PC was starved for IRQ lines, and the Amiga environment lives and dies on interrupts. Because PCs lacked the Agnus and Paula chips that made the Amiga so compelling, AmigaOS on the PC would have struggled, especially during the 80s, had it been written. Most likely the developers of the day would have declared the task impossible.

IBM licensed some Amiga technology for OS/2 in the 90s, which is probably about as good a "PC AmigaOS" as we'd ever get (absent recent efforts like AROS).

  > OK. I'm being harsh. At the time, with the information they had, those decisions that now appear stupid could make sense.

I don't think you are. At the time, we in the Amiga community were wondering wtf was going on back then, when it was happening.

It was pretty clear the leadership were making bad decisions, when they weren't absent.

They were still investing in 8-bit computers because they thought they'd recoup the investment. They did the 128, the 128D, and almost did a laptop and the C-65.

At the time it seemed like a sane thing to do while now we know they probably should have bet the farm on the Amigas (and take that Sun deal for the 3000/UX)

Harsh? Truth... Commodore under Irving Gould and Mehdi Ali was a sinking ship due to severe mismanagement.

I only thing I remember about the name "Mehdi Ali" from various documentaries etc. is that he was a serial ship-sinker.

Irving saved Commodore several times, but he had no idea about computers and never used one. This also led to issues of understanding with CEO's, like Jack Tramiel. He left Commodore and went to Atari... and in a revenge move, worked on the Atari ST. While Jack didn't have deep knowledge on computers itself, he knew what businesses wanted and had an acumen for driving it, and especially on the idea of cost. This you can clearly see in how the ST got positioned; it was more of a business machine. The Amiga was more of a funtainment machine that was good at everything. This made it challenging to target a certain audience. It was great at video, music, and even did games, with a floppy drive, and a full-size keyboard, etc. MArketing had a difficult time to properly advertise it. Several CEOs followed, and Thomas Rattigan made the company profitable again, but drama followed very soon ... and only got worse with Ali; https://www.commodore.ca/commodore-history/mehdi-ali-the-end...

The best part of the whole "Jack Tramiel leaves Commodore, buys Atari" story is the part about Atari floating debt to Amiga. The debt was coming due around the time Tramiel bought Atari and, at the last minute, Commodore acquired Amiga and Atari was paid-off. There was litigation following-- Tramiel was never one to let something go quietly-- but in the end we got the Commodore Amiga.

In some alternate universe there were Atari Amigas and likely no Atari ST.

Probably a Commodore ST, though.

My understanding is that Commodore had a Z8000 machine in development. It would have been neat to see what they came up with, in light of the fact that Shiraz Shivji went to Tramel Tech / Atari. All silly speculation, but it is so neat that, effectively, two major players in a market “swapped” their engineering staff. Surely that’s an odd scenario.

I wouldn't be surprised if AmigaOS was tightly integrated with its custom chipset (denise, fat agnus comes to mind as a colorful name for some of these), making it a lot more difficult to port it to x86.

I had a Commodore branded Escom PC running OS/2 in early 1996 but I’m guessing they made PCs before that in some parts of the world? In Europe it was all Amiga in the 90s

They did in fact build PC clones from the 1980s onwards, starting with the XT compatible PC-10 and the AT compatible PC-20. These were made in Germany, IIRC. Apparently they weren't bad, but back then I didn't understand the point when they had the much more powerful Amiga platform.

Their point was that the Amiga was a gaming machine, and they had no intention to make a full personal computer out of it, at least that was their idea when they stubbornly refused to ship it without a hard disk. That would be understandable given the high prices of storage back then, but they also insisted with low res graphics and lack of a cheap RTC chip, which killed almost all chances to use it in offices. I consider myself brave for having drawn my own business cards using Pagestream on a A500 with one megabyte of RAM and a flickering interlaced TV screen.

I remember things differently. While the Amiga was a great gaming machine, I remember Commodore pushing it as a desktop productivity machine, more emphasizing stuff like paint and music apps.

IIRC, the lack of a hard drive issue came down to keeping costs down.

I don't remember a specific issue with "low res graphics". Of course the flicker issue was present, but that didn't really become an an issue until later in the Amiga's life, and only on the low end models.

Regardless, it was certainly a very well-rounded platform. Definitely not "just a game machine" by any stretch.

Caveat: I was in the Amiga world in the 1000/500/2000/3000 days. I've never used or seen the later generation stuff, such as the 4000 or A1200.

> I was in the Amiga world in the 1000/500/2000/3000 days. I've never used or seen the later generation stuff, such as the 4000 or A1200.

I had the 500, then the 2000 and finally 4000/040, with RAM expansion, hard disk and a decent monitor. It was a wonderful machine but suffered from the same limitation of the former models: AGA was better than AA, but too little too late.

No bashing it any way, I just wish things went differently. What I learned on the Amiga helped me immensely with my jobs years later; I only have to be thankful for having the opportunity to use it and learn from it: a multitasking OS on hardware without memory protection forces one to be careful when writing code, for a small memory leak or invalid pointer can either render the system unusable pretty quickly or crash it altogether.

I'm looking at an advert in Amiga Format 54 (the Christmas 1993 edition from 29 years ago)... probably the pinnacle of the Amiga's popularity. Here are some sample prices:

Amiga 600 standalone: £194.99

Amiga 1200 standalone: £289.99

20MB HDD £89.99

60MB HDD £179.99

80MB HDD £195.99

120MB HDD £219.99

210MB HDD £359.99

Amiga 600 with 20MB HDD £289.99

Amiga 1200 with 85MB HDD £499.99

Amiga 4000/030 with 120MB HDD £969.99

Amiga 4000/040 with 4MB Fast RAM and 120MB HDD £1699.99

RTC fix: £14.99

Commodore 10884ST monitor (requires interlace for 640x512) £199.99

Commodore 1940 monitor (640x512 without interlacing) £284.99

You can see how the prices add up very quickly and make things unaffordable for the typical family. A business could afford to pay £1300 per computer, but a family likely could not. Most people plugged their machine into their existing TVs, as it seems did you.

If anything, this gives a good example of how technology has beaten inflation. The 20MB HDD is £4.50 per megabyte, or £4.7 million per terabyte. Today you can get a 16TB drive for £215, or £13.44 per terabyte, which is 351,000 times cheaper

I agree pretty much on everything, adding a hard disk would have placed the smaller machines in a whole different segment. I however recall also the A2000 was sold without a hard disk, just to advertise it ad a lower price, which I find absurd since that way it was essentially an A500 plus expansion slots.

To be fair, the expansion slots were a pretty useful thing and allowed a lot of extra hardware (including the Video Toaster) that the A500 sidecar was incapable of. But yes, the original A2000 was a souped-up A1000, later upgraded to a souped-up A500.

I looked this up and it appears the A2000 launch price was $1495 USD in March 1987, I don't know the UK price but it would've been around £950 at the time. That's ridiculously cheap (until you understand it's an A500 with Zorro and ISA slots). The A3000, probably the best machine Commodore ever made, launched for $3379 USD in 1990.

I don't have any magazines from 1987, but Amiga Format issue 1 (August 1989) has an advertiser selling:

A2000 standalone £799

A2000 with 1084s monitor and XT bridgeboard £1099

A2000 with 1084s monitor and XT bridgeboard and A2090a (HDD controller card for the A2000 because IDE hasn't been invented yet) with 20MB HDD £1369

A2090a with 20MB HDD £479

A2090a with 40MB HDD £750

A2058 (up to 8MB RAM expansion) with 2MB RAM populated £449

I can see why you'd offer a working system, 1MB RAM, floppies only, if it more than halved the visible price.

Commodore was known for years for making bad business decisions. The bankruptcy surprised no one.

Yes, it would be hard to name a major computer industry company which wasn't boneheaded in those days. It wasn't so much that Commodore made bigger mistakes, more that the industry turned against them and preferred to bet on other horses.

Dell pretty much did everything right, there's really not any stories of anything that went wrong for that company. Although they weren't really innovating and developing a platform, just making (really good) clones of someone else's.

Workbench 1.2 (?) had an easter egg that said "We made it, Commodore fucked it up".

https://www.youtube.com/watch?v=omMOuyTLmyg

> it could have been worthy to port AmigaOS to x86

No, it could not, and this is why:

AmigaOS is sometimes called a microkernel, but it's not really, inasmuch as one of the biggest problems in microkernels is efficient communications between the kernel (in supervisor mode, i.e. x86 ring 0) and the servers running in user space (x86 ring 2 or 3 depending on OS).

AmigaOS did not have that. It ran on a 68000 and everything was in the same, flat, shared memory space. This makes inter-process comms easy, but it also has a cost: a rogue program can trample all over other processes' memory, including the kernel's memory.

Thus bringing down your system, the famous Amiga "guru meditation".

You can't do this on an 80286 because it doesn't have a flat memory space, even in Protect Mode. You need an 80386 running in 32-bit mode: what we now call x86-32.

But AmigaOS has that total absence of memory protection. You can implement that on a 386, no problem, but it makes your OS unstable. This is why a rogue NLM could bring down a Netware 3 or 4 server: a flat memory space.

This is why many alternative OSes of the late 1980s flopped: they couldn't use the memory-management hardware that appeared in the 386, and in Motorola's 68030 (in the 68020 it was an external MMU chip, IIRC.)

This is also why CBM couldn't take advantage of the MMU in the 68030 and 68040. If it rewrote AmigaOS to use the MMU, all the legacy code would have broken and failed to run, because the mechanism it used to communicate with the kernel would have gone away.

(This is aside from the issues of 68000 assembly code conversion and register starvation and all the other things others have brought up later in the comments, which are legitimate and valid.)

So, while it might have been technically feasible to port AmigaOS to the 386 -- I mean, today, AROS shows it's possible -- you'd inherit all the limitations of the original, but on much more expensive hardware.

This would make it un-competitive.

This is why OS/2 flailed and then failed. This is why Windows 3 evolved into Win9x but then had to die: it was replaced by a true 32-bit OS that used the MMU: Windows NT, derived from bits of Portable OS/2 and some of the design of DEC VAX VMX, because MS hired the lead programmer of VMS and his team.

It's why MacOS couldn't fully exploit the 68030 and 68040, and indeed, the PowerPC. It's also why Apple's Copland project failed. This stuff is really hard. Apple ended up moving to UNIX.

x86 would have put Amiga in the same boat as Microsoft: driver hell for eternity. Also why x86 macOS was a bad idea.

What they needed to do was invest in the successor Amiga chipset to the original one. Yanking the funding for Amiga R&D in the late 1980’s is what doomed Amiga and Commodore.

x86 has very little to do with driver hell in itself.

x86 macOS had a long and successful run.

Or perhaps you mean porting Classic Mac in the 90s would have been a bad idea?

The core of MacOS (NeXTStep) was already ported to x86 long before Apple was even involved. Remember NeXTStep ran on x86 in the early 90's, as they exited the hardware business. It was not compelling for server workloads. It was also very picky on drivers, as the other poster implies. Plus it was difficult to build open source software for (SunOS was the gold standard for commercial Unix at the time. Everything you downloaded ran on Sun!)

I knew guys who ran early ISPs, tried it for servers, then realized they'd be better off with a Sun box and then eventually early Linux (Slackware). They were trying to run a news server off of it, IIRC.

I know this history.

However, the point remains that Apple ran consumer products on Intel chips for 14 years or so and this was successful. I'm trying to parse out the claim being made. Eg. Is it an argument that the timing would have been bad in the 1990s? Notable that Apple also killed the experiment around PowerPC clones around the time Jobs came back.

Timing does seem bad, seeing they had transitioned to PowerPC only 4 or 5 years before NeXT was acquired. And Intel had not yet really taken a huge performance lead. PowerPC's were still pretty competitive. It wasn't until the mid-2000's that it was clear they were not going to get a G5 or whatever running in a laptop.

Apple also wasn't rolling in cash like the are today. They were struggling. They wouldn't have had the resources in the late 90's to support consumer-level x86 and the various hardware/driver configurations. NeXT was fine with supporting a limited set of configurations since they weren't targeting the "consumer" anyway. They were targeting high end commercial "workstation" customers. NeXTStep 486 cost something like $800 if I remember. The developer tools were more $$$. The people running that stuff could afford to spec out their machines properly.

The missing turn was rather not to open source the whole OS...

Reply on news.ycombinator.com