Why not A/UX on BasiliskII, really?

[andreadandrea]

Hi everybody,

I read the many posts on why A/UX does not run on Mac emulation, so I know the old story (no MMU , no special video, etc).
But it happens that lately somebody has managed very proficently to run a 20+ years old Uniplus on Lisa emulation (after a very interesting restoration of the original disks of that A/UX father, check lisalist if interested) and that made me think, without being neither a coder nor an unix guru, it must be possible to adapt BasiliskII to run A/UX somehow!
Maybe bypassing the standard installation, maybe working directly with a dump of an installed hd, maybe running videoless, maybe leveraging the 0.7 kernel sources, maybe who knows, etc.
Was the effort completely abandoned?
The original hardware will soon disappear (and is damn slow when it still works, btw), so it would be great to keep at least the original software running on emulation.

My 2 eurocents,
Andrea

[Gorgonops]

Honestly it would be "relatively trivial" for someone to write an emulator capable of running A/UX if someone with sufficient technical ability felt like spending the time to do it. What's needed is an emulator that can simulate the "Bare Metal" of an A/UX-capable machine, and bare-metal emulators exist for far more complicated machines than that.

(The Amiga, for instance. Macintosh hardware tends to be rather "arcane" in the sense that Apple was prone to using hackish electrical shortcuts to save parts, but it also was fairly "simple" compared to the custom chips in other 68k-based home computers. Or video game consoles, for that matter.)

So really, the problem is a lack of interest rather than any technical issue, other than the fact that an emulator like BasiliskII that can "cheat" by patching the Mac ROM can run the very device-independent "classic" MacOS considerably faster than a bare-metal emulator could. "No one" seems to care about running alternative Macintosh OSes under emulation so BasiliskII's emulation style is the optimum one under those circumstances.

Support for Macintosh II-level hardware is starting to make it into MESS. MESS may well be the first emulator to run A/UX, as emulating the bare metal is a MESS priority.

[johnklos]

Much of the work into the various m68k emulators (ARAnyM, UAE and friends) is on versions which don't emulate the MMU, and people who work on MMU versions (for GNU/Linux and NetBSD, mostly) usually work from older versions of the emulators with MMU patches. The speed difference between non-MMU and MMU versions is significant, so there hasn't been much effort to keep both up to date in the same place.

Maybe that's something which needs to be encouraged...

[Unknown_K]

If it is so simple then why don't you code it?

There are plenty of cheap 68K macs around if you need to run A/UX. I would rather see coders expanding A/UX 3.11 with useful utilities then bothering to write an emulator for it. There just isn't that much software out there that is A/UX native.

[johnklos]

If it is so simple then why don't you code it?

Not sure who the you is here, but many people don't know how to code.

There are plenty of cheap 68K macs around if you need to run A/UX. I would rather see coders expanding A/UX 3.11 with useful utilities then bothering to write an emulator for it. There just isn't that much software out there that is A/UX native.

There are many, many times more contemporary computers which are in the hands of interested people who might not have the time, space, power (electricity), or money to buy an old Mac. I used to live in NYC, so I had to make sure all of my machines were as small as physically possible, so having another machine around to JUST run A/UX wouldn't have been very practical.

Make it into a downloadable program, though, and lots more people would work on it, I'm sure. Plus, the emulators would generally run faster than the m68040 Quadras, so doing an A/UX pkgsrc bulk build would be a lot faster than running on a real Quadra.

While it'd be nice to keep around older machines, that's not a good reason for not working on an emulator.

[Unknown_K]

The YOU would be the person who said something is trivial to do (asuming they know enough about the task to come up with that opinion) yet ask somebody else to spend the time to do it.

Generally the systems that get EMU's tend to be game oriented. Few people will bother coding an EMU for something obscure unless they want to have it for themselves or there is a huge community that would use it (and help report bugs or offer to code). I am happy a few people will bother to port some apps over to A/UX, doesn't seem like the diehards are running out of hardware to spend the time to make an EMU.

[Gorgonops]

The YOU would be the person who said something is trivial to do (assuming they know enough about the task to come up with that opinion) yet ask somebody else to spend the time to do it.

Getting nasty, aren't we? If you're referring to me in particular, well:

A: I never asked anyone to do it.
B: I know that I'm not capable of doing it, at least not without dedicating way more time than I have, and
C: If I had that time I'd spend it doing almost *anything* else. I just don't care.

However, I am intelligent enough to know what does already exist out there in the way of emulators, and I'm also capable of *reading documentation*. (Hit Google, there's plenty of documentation out there. Not all of it complete, but there are lots of hints.) A Macintosh II is no more complicated a piece of machinery then, say, an Atari Falcon or TT, Sun3 Workstation, or Amiga, and complete MMU-equipped emulators *capable of running UNIX-esque operating systems* exist for those machines.

(And yes, at least the SUN emulator will run the UNIX supplied by the vendor in addition to NetBSD or Linux. Heck, lets expand our scope beyond 68k machines. QEMU, Hercules, and SIMH all emulate machines with MMUs and run big complicated UNIX-y OSes. Why are we pretending that a bare-metal MacII emulator is some sort of Holy Grail? Oh, yeah. No one cares enough about A/UX to have written it yet.)

Perhaps "trivial" is the wrong word. A/UX undoubtedly exercises the hardware *extensively* and getting the emulator running in a bullet-proof manner would undoubtedly involve hour-upon-hour of painful debugging. There is certainly the issue that while *most* of the MacII hardware has been reverse-engineered, as witnessed by the fact that Linux, NetBSD, and OpenBSD can all run on it Apple has never *really* come clean with the gross details of how those machines work. Someone doing this work would have to pick a model, make a best-effort stab of emulating it based on the third-party documentation, and then start grinding. However, all those issues apply to other platforms that *have* been emulated successfully already. (How forthcoming do you think Sega, Atari, Konami, etc. have been with supplying official documentation of their custom Arcade cabinet boards to the MAME developers? Not very, you think?) "Trivial" in context merely meant that it should be no harder than any other emulator of similar scope, and the fact that such emulators already exist in open-source form means that large parts of it don't need to be written from scratch. If history is any guide a competent, knowledgeable hacker can usually get an alpha-quality emulator off the ground in a few months to a few years of "spare time" work even if they *are* starting from close to scratch.

The real point of my statement is that there apparently *isn't* someone with the proper talents to do this work who also has sufficient motivation and/or time to do it. The butch coders that could slap a bare-metal Mac emulator together in a few man-months of work (and yes, they exist) are apparently too busy working on MAME, UAE, or any number of other things they find more interesting or lucrative than making it possible to emulate A/UX.

I seriously have no idea what point you're trying to make by lashing out. "Ooooh, mean guy is saying something is easy when he can't do it himself! See, see!". Whatever. Technically, based on your statements we essentially agree, IE, "It's technically very doable, but no one cares", so I don't get the point.

[Unknown_K]

Well you made it sound like anyone could do it in a few weeks, and that is not the case for any of the EMUs made.

Anything can be coded if people want it badly enough. I can't see A/UX getting very popular even if it was emulated perfectly. Mostly A/UX is boring server software, with little native apps.

[Anonymous Freak]

Gorgonops, I think the main gist of Unknown_K's offense is that you made it sound like it's something that someone with knowledge should be able to just throw together. Yet as someone that, by your own admission, has neither the knowledge, nor the time, in addition, you don't even have the desire even if you HAD the knowledge and the time, your statement comes off as (wether you meant it or not,) entitled.

Stating "Honestly it would be "relatively trivial" for someone to..." while not having any of the stated qualities yourself, comes across badly, regardless of what it is you are stating the triviality of. Now, if you do have the required talents, feel free to make the statement.

For example, I have read about recapping motherboards, and recelling battery packs. Many people make it sound "trivial". Yet whenever I mention these tasks, I do *NOT* call it "trivial", because I have not tried them, so I do not know anything of the possible triviality. However, I am more than willing to say that replacing the hard drive in an iBook is "trivial to someone with technical skill", because I have done that. I can also state that the skill necessary to pilot an airplane from takeoff to cruise is "trivial", while not making any such statement about landing. But I wouldn't dare state anything of the like about piloting a spacecraft, regardless of how simple professionals make it look.

The offense derives from the implied sense of "well, you people are smart, why haven't you done this yet?" entitlement. I would argue that the fact that we haven't seen such an emulator is de facto evidence that it is not trivial. There are hundreds of obsolete-architecture emulators out there, including a good number of m68k emulators, and quite a few for even more obscure systems than m68k+A/UX. The fact that none can run A/UX is evidence against triviality of the task.

[Gorgonops]

Apparently once everyone saw the word "trivial" they completely glossed over the qualifier immediately afterwards that said "... if someone with sufficient technical ability felt like spending the time to do it". The fact that I personally have not the time, skills, nor inclination to build a Mac emulator myself does not in and of itself invalidate my statement. By your logic it should be completely unacceptable for anyone to comment upon *anything* that they themselves could not single-handedly accomplish, and that's ridiculous.

Anyone can look at a river and think to themselves "You know, if they felt like it they should be able to build a bridge there". Bridges are a well-understood technology and plenty of examples exist. Does that mean very many people have the skills and resources to actually build the things? No. However, even a *slightly* interested layman (let's say someone who's read a book or two and perhaps took a civil engineering class in school many moons ago) looking at a given river should be able to make a fair guess about how "easy" it would be to bridge it based on the observable conditions and their knowledge of similar bridges. Yes, in real life it takes a complete site survey and engineering analysis to nail down the actual costs and challenges for any particular project. But I scarcely see how that fact gives you the opening to jump all over someone when all they do is point to a couple of moderate-length highway bridges and say "well, those two bridges upstream cost about $45 million dollars each. If they wanted to build another one a mile downriver it should be pretty trivial, assuming they have another $50 million to spend."

It's not as if I waved a hand dismissively at the mouth of the Amazon or the English Channel and said "it should be trivial to build a bridge there". What I said, based on what I've read about the Mac hardware (which is a fair bit) and also knowing that stubs for MacII emulation already exist in MESS is essentially: "I'd think it *should* be easier than an emulator for (a machine which has a working emulator), and someone doing it should be able to reuse a lot of code from that emulator and others like it... so sure, shouldn't be *that* hard." Can you prove to me that it will be *massively* more difficult than any other emulator? (Of course if you *could* prove how difficult a Mac emulator would be then you should be the one working on it, since clearly you have precisely the encyclopedic knowledge of the Mac's hardware quirks that would be needed to write it in the first place.)

All I did was make a ballpark estimate and I stand by it. In sheer technical terms it should be easier, or at least no worse, than an Amiga emulator, and UAE went from completely unusable to "it actually sort of works" in about a year, starting *completely from scratch*. I suppose worse case it'll be about as hard as that Lisa emulator. That took a single person the better part of a decade of free time grinding away to make happen, but they were also starting very much from scratch and working on a nearly *completely* undocumented platform. If someone actually embarks on an A/UX emulator project, actually *works* on it and documents their challenges I'll be happy to eat crow and take back my "Trivial" description of their personal odyssey. But until that happens I can shamelessly snark all I want.

Ad Hominem arguments are emotionally satisfying but they rarely advance the cause. My stated position is that an A/UX emulator doesn't exist because the effort hasn't been put into it, not because it's impossible. Has anything you've said materially discounted that position? If I sounded a bit jaded when I stated said position it's basically because I've seen a lot of "wouldn't it be cool if...?" threads on this board which peter out simply because of a complete failure on the part of the 68k Mac community to recruit a skilled coder to champion the cause. (For the record I greatly respect the few productive coders there *are* still working in the field, such as the author of Classilla.) As stated, emulators for more obscure platforms do exist. In my humble opinion the reason that's so is because the devotees of those other platforms count among their ranks a far greater percentage of highly skilled technical users than the "Antique" Macintosh community does, not because their hardware is easier to emulate. How is this surprising, honestly? I'm sorry if the statement rankles, but... it is what it is. The Macintosh was designed for users, not hackers, and since the Mac "evolved" rather than died the hackers it had have moved on and are hacking out iApps now. (The few that *might* be nostalgic for the old days already have BasiliskII and Sheepshaver, so why bother?)

So enjoy *using* A/UX as long as you can, I guess. Feel free to shower derision an anyone else who dares suggest that someone *could* do more than that.

[andreadandrea]

Dear all,

when I wrote the post I suspected that a somehow harsh discussion could/would come out, I read many previous posts in this forum and understand that the A/UX community is somehow special.
As Gorgonops (many of whose views I share) said, mac users are mainly users, maybe. Maybe there are no mac gurus out there wanting to spend time in a difficult endeavour.
I will not discuss the etymology of "triviality" and the nuances of the word; I prefer to keep harshness out and just concentrate on facts.
I would say that this forum is a rather good representation of the vintage mac and A/UX die hard community.
I will distinguish that community in three classes: A/UX fans, A/UX interested and A/X curious. That is about the interest level, not the skill level; you can change the names if you want, I think the meaning is clear.
From the posts, I would say that in the first class there could be 10 people, in the second 100 people, in the third 1000 people. Of course I am speaking of magnitude orders.
I started myself in the third class, where I lived enough time to see I wanted to try A/UX, so I passed in the second class.
It took to me about two years (not to speak about money) in the second class to: take my old quadra from a country to another, find parts, update it, blow it, find another old quadra, update it, install A/UX on it.
So, the fans obviously have enough spare macs and spare parts to be safe to count on some machine to work with A/UX until pension and beyond.
The interested use to have a hard time to get a real machine to just try, a machine that could work 15 minutes and literally blow, so many years after it was produced. I know many users of this forum are from USA and look to the world from they point of view; let me tell you that outside USA it is much more difficult to find working parts.
Finally the curious; of course they will never try A/UX, mostly because curiosity is not enough to get an actual machine to try A/UX.
So the A/UX users community is 110 people in the best case, and going down because of people pensioning and hardware failure.
The people that are porting apps to A/UX are working for themselves and for that community.
But if some fan joins forces with some interested and with some curious and make an effort to produce a way to run A/UX on emulation, the A/UX user community will count also with the curious. And above all, it would be stable or maybe going up, because there will always be the curious.
In that case it would really make sense to me to make further efforts to port apps, because there would a live user community to use them and maybe contribute more.
I have a vintage mac and a BasiliskII mac and look at them with objectivity: the second works with most apps, uses modern peripherals (USB sticks, large LED screen, huge hd), runs fast, networks easily and with time will be better, the first runs all the apps, runs slow, has vintage peripherals, kind of networks and with time will blow up. I will keep its chassis as a souvenir and keep using the second at will.
It is clear that it is difficult to make an obscure and undocumented OS run on an emulator, but it must be done now.
In 5-10 years a fraction of the old macs will still work, so it will not possible to compare the emulator with the real one, which is so important during the development. In the same time a fraction of the fans and interested will be still there, so there will be no more users and developers.
This answers to the question on why and when doing the emulator. I leave the answer to the question on who to the next posts.

Regards,
Andrea

[techfury90]

I'm going to have to agree with Gorgonops here. Given sufficient chipset documentation, writing an emulator that could emulate, say, a IIci would be relatively easy. Hell, I had my Apollo emulator booting into the ROM monitor after just a week (I recycled the CPU emulation core from another project, as suggested)... (then again, the DN3000/3500 is just a 68020/030 bolted to an ISA bus with some special cards)

[Bunsen]

"Trivial" has a specific meaning in CS that is different from its ordinary usage. It doesn't mean simple, or easy, just that it is a solvable problem.

[Cory5412]

It's definitely doable, but in my eight years and one day on this forum specifically, I've never seen anybody post anything other than "if you had the time and skills and willingness it would be possible" along with "this is old stuff on old, relatively simple hardware, it should be emulated by now" on one hand, and on the other hand "but who is going to write it... you?" kinds of responses.

Part of it is that the Mac never attracted the kinds (and amounts) of technical computing people that I suspect Apple wanted it to when A/UX was first introduced. A few people used A/UX as a file server and I'm sure that there were a few straight-up technical and programming-able users, but not in the kind of scale that you'd really need to get a project like this going now that it's literally twenty years later.

This doesn't mean that I have no hope for it at all, but what I will say is that I'm working on keeping my physical Macs running until someone does bother with it.

It's not something to lose hope on, but it's not something that I suspect you'll be able to get anybody on the forum right now to do, or admit to doing at the moment. We'll see though I know I'd be interested in getting the result and trying it out, not only out of interested for A/UX, but out of interest for networking multiple A/UX systems, and emulating/virtualization being one of the best and easiest ways to do this..)

[Bunsen]

Speculation and argument about emulating an Amiga on an FPGA went on for years until that one guy went away and quietly built it. We just haven't had that one guy yet.

[Gorgonops]

"Trivial" has a specific meaning in CS that is different from its ordinary usage. It doesn't mean simple, or easy, just that it is a solvable problem.

And I apologize for flying off the handle over people nitpicking my use of it in that sense. (Which *was* annoying, particularly since I put a strong qualifier on that usage in the same sentence.)

Anyway, yeah. This is a completely solvable problem, it just requires someone with sufficient time and motivation to dedicate themselves to it. In the OP's last post they made a legitimate point about just how few people there are left in the world with a "strong" interest in A/UX; most Mac users have never heard of it, let alone run it. And despite protestations to the contrary it's starting to get "not *that* easy" to even assemble a machine capable of running it. The hardware is old, increasingly rare, and the software isn't that easy to find either. (Yes, Google-ing will turn up an illegal download. You still need to know what you're looking for and care enough in the first place.) There are a significant monetary and logistical barriers standing the way of anyone who might have a "casual" interest in running A/UX after learning that it exists. Thus if you're stuck waiting for someone with just a "casual" interest in Macintosh emulation, or even someone with a keen interest in Mac emulation but no knowledge of A/UX, to make an emulator capable of running it for you it's just not going to happen.

Out of curiosity I downloaded MESS 0.138 and compiled it last week. (Ironically only three days before 0.139 came out.) A notable milestone has actually been passed: as of that version the SE/30 is marked as a "working system". (It beat out the other MacII-family machines because its video support is more easily grokked than real NuBus cards.) In reality that doesn't mean what it sounds like; the SWIM chip emulation is broken, meaning there's no way to read floppy images, and there was still as of that time a generic problem in the "mac" umbrella driver that prevents running System 7 or higher successfully. Nonetheless after enough fiddling I was able to install System 6.0.8 on a hard disk image using the "macplus" driver and boot and run it as a "macse30". So, there you go, there *is* a "cycle-accurate" 68030 Mac emulator in existence that works to a significant degree without resorting to BasiliskII/vMac-style ROM hacks. It's far from complete but it's a start.

What could be done to improve this situation? Obviously, having a coder willing to dedicate him/herself to filling in the blanks with the Mac driver would be a huge boon. From casually looking at the changelogs it appears to me that much of the work on it is being done by one person, and that person contributes to MESS *as a whole*. Thus it's unlikely that the Mac driver will ever evolve any faster unless someone willing to spend "all" their spare time on it gets on board. However, even if you *don't* have coding skills you could get involved. If you have the ability to install A/UX on a physical machine you could offer the MESS developers a .chd disk image of a working install. And of course you could also file specific, detailed, and reproducible bug reports on your own attempts to run various software on MESS, help in unearthing technical documentation for the hardware, and possibly even offer to donate a complete working machine with A/UX installed should a developer show sufficient interest. If you can cultivate an existing developer's enthusiasm in the project that might *almost* be as good as contributing code yourself.

And lest someone tell me to put up or shut up again, I don't know how to be more clear: *I no longer have any working 68k hardware, nor am I sufficiently interested in A/UX to buy any*. Even if I cared to try (and had the time) to learn enough about MESS to contribute to it I'd be completely shooting in the dark. (It's worth noting that being able to run the installer for an OS can be one of the toughest milestones to hit in writing an emulator. Starting from a known-good dumped hard disk image saves having to run down a bunch of ratholes prior to the main job of improving the core emulator.) This is a job for someone who really cares, and all I'm doing here in my backwards and unhelpful way is to try to cheerlead someone who cares into doing something concrete towards their goal rather than just cursing the darkness and accepting fate. Sometimes getting people mad is the only way to motivate them.

Anyway. Feh.

[Charlieman]

When I am bored at work, watching the progress bar going nowhere, I look on the bookshelves around me. This week I picked out a copy of Understanding Unix (Que, 1988).

The authors told me all about state of the art Unix text editors (ed and vi) and how to use rlogin securely. The authors speculated that Unix would become more popular thanks to Apple's A/UX (for beefy Macs) and IBM's AIX (for beefy PS/2 PCs). Or perhaps Unix users would buy Sun workstations.

To be fair to the authors, they did get one bit of speculation right: people bought Suns. And as much as I would like an A/UX emulator, I can understand why nobody has created one. Few people used A/UX and few people wrote Mac applications for A/UX. And if you want to run the Unix apps, they are ready to run on your preferred version of System V or BSD.

[johnklos]

As a slight tangent, if people want to develop for A/UX on an emulator now, you can come close by emulating an Amiga with MMU and installing Amix (Amiga Unix) which was based on AT&T System V Release 4, just like A/UX. Amix and A/UX are binary compatible.

[ChristTrekker]

Amix and A/UX are binary compatible.

There's a piece of trivia I'd never heard before.

[tlc630]

Here's another piece of trivia: Atari MiNT binaries are frequently runable on A/UX.

tfl

[Gorgonops]

Here's another piece of trivia: Atari MiNT binaries are frequently runable on A/UX.

I would love to see corroboration for that statement. MiNT adds "Unix-Like" extensions to the DOS-ish API of Atari TOS but I can't imagine compiled binaries being very much alike. I can see being able to compile at least trivial *sources* on the two platforms...

[tlc630]

Sorry...mixing up the antique OSes I fiddle with. Atari MiNT binaries will frequently run on MacMiNT.

[ClassicHasClass]

That I could see, given how jet (Just Enough TOS) works. I rather like MacMiNT despite its limitations.

[protocol7]

As a slight tangent, if people want to develop for A/UX on an emulator now, you can come close by emulating an Amiga with MMU and installing Amix (Amiga Unix) which was based on AT&T System V Release 4, just like A/UX. Amix and A/UX are binary compatible.

I haven't been following UAE for a long time so I didn't realise they'd added MMU support

Guess I can finally bust out those AMIX images I got years ago and see it in action.

[Gorgonops]

I'm curious if Amiga UNIX can be made to work on the MMU-patched UAE. I'm a little skeptical as the patch documentation (which is pretty sketchy) targets running Linux, and the one significant website I've found talking about AMIX assumes you need real hardware. But I don't know much about Amiga, so... might be worth a try.

(I've never futzed much with UAE simply because I don't have any nostalgia for the Amiga. Thus I don't know how exactly-specifically it can be configured to emulate a particular model and how the MMU patch interacts with that.)

The Atari emulator Aranym has the MMU support better integrated into it but it likewise targets Linux, most possibly because the Unix that was briefly bundled with a particular model of the TT030 is essentially vaporware. (If anyone on the planet has a working TT030 UNIX system it might be interesting to see if it has any binary compatibility with A/UX.) The only 680x0 emulator I *know*, as in having tried it, that runs a commercial MMU-dependent UNIX is TME, the Sun3 emulator. But of course that would be a given since a Sun emulator that didn't run SunOS would be by definition pointless.

(There may well be working emulators for even more obscure UNIX workstations that I don't know about. As an amusing side note M.E.S.S. has added a skeleton driver for NeXT systems. Don't hold your breath for that one...)