Hacker News Reader

Comment by drooopy | Hacker News Reader

I've always wondered why Apple didn't kept on developing A/UX as their replacement for classic Mac OS but instead chose to focus on either Frankenstein-like experiments, such as Copland or purchasing technology stack from 3rd party companies.

View on HN

They did. The original plan was a PowerPC native version of A/UX "4.0." This was the PowerOpen idea, in which both A/UX and IBM AIX were to be personalities and value-adds on a central OS. This new A/UX would still have been able to run Macintosh applications natively. Eventually it faltered because of second-system effect from Pink/Taligent running at the same time, as well as Spindler's brief flirtation with NetWare. This is the reason the Apple Network Server ended up running vanilla AIX.

(disclosure: my work. https://oldvcr.blogspot.com/2023/11/the-apple-network-server... )

Yep... the move from m68k to PowerPC was pretty much the deathknell. To really be a successor to A/UX, the core code would have had to be ported to that chip, and the amount of work required was simply too expensive to be justified. I also had heard, unofficially at the time, that Apple's license for the UNIX parts of A/UX were specific to the m68k architecture, and that they would have had to renegotiate the license to even be allowed to port it to the PowerPC.

What is interesting is that after A/UX died, some alternatives did spring up, such as Yellowdog Linux (which was NOT a macOS/UNIX hybrid, but rather vanilla Linux).

It is a shame, because if Apple had continued in investing it it, we would have had MacOSX much, much earlier.

Apple wouldn't have had "MacOS X" (as we got it) earlier. Had they kept up some Unix system they still would have been saddled with Toolbox, just running on a Unix kernel. They likely still would have produced Carbon but they'd still have a very backwards facing application API. It might be more stable thanks to privileges and process isolation but there's not necessarily a future there.

With NeXT, Apple got OpenStep which (IMHO) offered a superior programming model than Toolbox/Carbon. They also got a superior display subsystem, networking, and better first party tooling. Those are the things that made what got called "MacOS X". Just a Unix dragging Toolbox around wouldn't have gotten Apple where they are today.

A modernized version of Toolbox in the form of Carbon running as native unix applications in a preemptive multitasking system on A/UX would have been pretty good - maybe not as good as Cocoa, but good enough probably.

Is win32 fundamentally worse than Carbon or Cocoa? if so it seems to have not meaningfully effected the success of Windows.

That might have been good enough for 1998 but I don't think it would have been as workable in 2008 and later in 2018. Remember without acquiring NeXT they wouldn't have also gotten NeXT's engineers and management. Apple obviously had management issues in the early 90s. The management that led to the failures of Pink/Taligent and Copland wouldn't have suddenly built a great OS just because there was a Unix kernel underneath. Hell, the best Mac development APIs came from third parties (Symantec then MetroWerks) rather than Apple themselves.

Win32 is a bit different than Toolbox/Carbon. For one Microsoft had superior first party development offerings. The best development tools on the Mac were from third parties that were increasingly disinterested in the platform. Windows was also an order of magnitude larger of a market. An ISV suffering through the pain of Win32 had a huge TAM in front of them. They could be profitable by only reaching a small fraction. An ISV targeting the Mac platform had a much harder time breaking even to say nothing of profitability just due to the size and nature of the Mac market.

Interestingly, a group reverse-engineered the Mac Toolbox and used it for "native" ports to Unix of, among other things, PhotoShop and Illustrator.

http://preserve.mactech.com/articles/mactech/Vol.13/13.06/Ju...

The thing with raw Win32 is that only hard-core C folks use it as is.

It is similar to using Carbon.

Already in the Win16 days, anyone that cared about productivity was using C++ frameworks like OWL, Turbo Pascal, later MFC/VCL, Delphi, VB.

To your point the “Hello World” Win32 C example in Charles Petzold’s earlier editions of Programming Windows is infamous.

http://www.charlespetzold.com/blog/2014/12/The-Infamous-Wind...

Yeah, one of the things Windows 3.1 brought was message crackers and a strict mode via windowsx.h.

https://learn.microsoft.com/en-us/windows/win32/api/windowsx...

However still painful versus the C++, Delphi, VB alternatives.

Well, that's assuming a very specific, and stale, development direction; I think we would have seen a more native and natural migration to something very macOSX like. Sure, not exactly macOSX as we know it today, but something that we would have used and appreciated as much as what we now have.

What made OpenStep so useful was that it was immediately _available_, but imagine if it was used as a source for tech to be added to the continuing A/UX development, instead of basically being Step One.

I am guessing here, but I worked at SCO (when it was a UNIX VAR, not three vexatious lawsuits in a trenchcoat) and the killer was always licensing fees, especially to AT&T. Sun bought a royalty-free license to SVR4 from AT&T which became Solaris, but it cost them on the order of $200M (if the figure I was told was accurate). SCO had a royalty-free license to SVR3.2 but had to pay out $250 to licensees for each copy of SCO Open Desktop 3 they sold (the precursor to SCO Open Server in 1995) for stuff like Motif, IXI Desktop, the Microsoft C compiler they still used ...

I suspect Apple went to AT&T for royalty-free terms and were quoted a price that, in the troubled Spindler/Amelio era, they simply couldn't afford for a niche product (the servers).

They could have ported their user space to Linux or BSD though?

I never used A/UX, so I can't comment on its relative quality.

But it took half a decade for Apple to ship OS/X after the palace coup that brought Jobs back and killed off all their indigenous OS development. And what they ended up shipping with the first revision of OS X didn't really feel very "Mac" like, and ended up being this weird fusion of NeXT tech with a sort of pseudo-Mac UX.

I guess I wonder what was wrong with the A/UX tech that stopped them from utilizing that as a foundation for something new.

EDIT, answering my own question: from WP "cooperatively multitasks all Macintosh apps in a single address space by using a token-passing system for their access to the Toolbox."

Not sure if this would apply to "hybrid" apps that used the Toolbox but with Unix syscalls as well, but I'm guessing yes, and that sounds like a serious limitation. Pure (non-GUI) unix applications would be fully multitasking, but GUI apps would have the same kinds of limitations as on System 7 -- poor/no memory protection, cooperative multitasking, bad memory management.

Apple actually did investigate Linux with MkLinux (a project to port Linux to PPC and run it on top of Mach). Neat, but arguably a symptom of their larger issues which were less technical and more leadership which was stifling whatever technical and design excellence there was left. Plus the Clones were eating them alive.

Any OS they released to replace the classic OS, which at that time would have been System 7, needed Macintosh App compatibility, and they wouldn’t have needed Linux or Unix or BSD or OPENSTEP or BeOS if they could have just shut up, clamp down on the mission creep and ship Copland. Apple’s Executive leadership in the mid-90s was unable to execute on that, and none of their side-projects were going anywhere in a sustainable (to Apple) fashion either.

I’m sure there were organizational, business, and technical failures, too, but it seems to me like Apple’s disastrous ~decade attempting to put the Macintosh on a modern foundation was primarily a product failure.

They wanted to do the right thing and synchronize the OS kernel switchover with the PowerPC transition, but chose to pursue a radical software compatibility break alongside the hardware swap.

When that failed for what should have been predictable reasons, they had no choice but to fall back to a pure emulation strategy with no OS architectural changes, and the opportunity to put the MacOS on a new path was lost.

At the time BSD was mired in a lawsuit from AT&T, and Linux was working but it wasn't clear it would become the most widely adopted *nix. This was Linux kernel 1.x days.

Yeah I was a Linux user in those days (1992 and onwards), and yes, 1.x was a bit rough around the edges but the momentum was pretty good. And 68k support was good fairly early on.

I can see that a large company like Apple wouldn't have gone near it at that point though.

Apple was transitioning away from 68k by the time Linux was released. So 68k support wasn't really that useful to them.

In your original question I think you're forgetting the context of the Unix market of the time. It was the high Unix Wars and A/UX was a minority position of a minority player.

The buzzword of the decade was Object Oriented and anyone not elbows deep in an existing Unix was trying to build an OS around objects. It was entirely unclear if jumping on the Unix bandwagon at that time would have a future. Even with Apple's acquisition of NeXT it was their high level OpenStep that was of most interest rather than the BSD base layer.

And with some success too: Cocoa/AppKit is still hanging around as an actual API that’s still being used by application programmers after 30 years, unlike winapi.

Yeah, after some quick looking into this, cost seems to be the reason why. And ultimately their decision to purchase an external solution seemed to have paid off handsomely.

Yeah, after AT&T was allowed to gain commercial value of UNIX they were clearly minded in recovering all the lost revenue from the free beer days of UNIX early adoption.

They licensed Sys V to build A/UX. That must have been quite expensive.

Reply on news.ycombinator.com