SCO, Linux, and Microsoft in the History of OS: 1990s
August 18th, 2007
Daniel Eran Dilger
[continued from part two: SCO, Linux, and Microsoft in the History of OS: 1980s]
Intel 80×86 vs Motorola 68000.
In the 80s, a new generation of graphical computers from Apple, Atari, Commodore, and NeXT–all based on the Motorola 68000 family of processors–leapt past the previous generation of 8-bit computers. That new hardware enabled more powerful software using a fully graphical user interface.
IBM’s 1981 8-bit PC had originally used Intel’s low price 8088; by the mid 80s, PCs started using the compatible, 16-bit relative: the 80286. Over the next few years, Intel’s x86 line of processors used in DOS PCs lagged behind Motorola’s 68000 architecture in performance, sophistication, and design elegance.
Early PCs couldn’t approach the software capabilities of the Mac, Amiga, ST, and NeXT. However, the economies of scale related to the PC industry enabled Intel to eventually catch up, match, and then rival the performance of Motorola’s 680×0 family.
Encroachment by Intel PCs was disastrous for the hardware business of Atari, Commodore, and NeXT; Atari left the computer business and went back to video games in 1993. NeXT also dropped its hardware in 1993, and Commodore went bankrupt in 1994.
That left Apple as the only significant alternative computing platform apart from the British Acorn. Acorn’s largely independent history was documented in “Origins: Why the iPhone is ARM, and isn’t Symbian.”
RISC: Reduced Instruction Set Computing.
As Intel’s 486 began to offer strong competition to Motorola’s 68040 in the early 90s, Apple began searching for a new generation of processors to power future systems, involving both internal work on custom processors and work with Motorola to use its 88000 RISC processor.
New hardware wasn’t just related to bigger and faster computers; Apple also worked to develop a new RISC processor architecture appropriate for handheld mobile devices with Acorn. That resulted in the ARM architecture used to power the Newton; today ARM is inside the iPod and iPhone and the vast majority of similar devices.
ARM was among a new generation of RISC processors which used a simplified hardware design paired with more sophisticated code compilers to advance the state of the art in powerful and efficient hardware. Among the RISC architectures already in use in workstations were DEC’s Alpha, SGI’s MIPS, HP’s PA-RISC, and Sun’s SPARC.
The AIM Alliance: 1991 – 1996.
Meanwhile, IBM had lost control of the Intel-based desktop PC it had created in 1981, and failed to regain it with its PS/2 computers in 1987. After being pushed out of the PC hardware market by cloners, IBM was then abandoned by Microsoft in their joint venture to deliver OS/2.
IBM’s original 1981 design for the PC had considered using its own 801 processor, a far more powerful RISC design. Throughout the 80s, the IBM’s 801 developed into the 1990 POWER architecture.
Seeking to develop a new desktop processor architecture based on POWER, IBM formed an alliance with Apple and Motorola in 1991. Their AIM Alliance paired POWER with existing work done by Apple and Motorola to deliver PowerPC. The AIM partners intended to entirely replace Intel’s x86 PC platform, which appeared to be running out of steam, and put PowerPC systems running an advanced new OS in its place.
The Next Generation of Software: 1990 – 1995.
IBM planned to use PowerPC as a clean new design powering desktops and workstations running either OS/2 or AIX, its brand of Unix. Apple and NeXT began making plans to migrate their own 68×00 hardware to use PowerPC, and Motorola worked to port Microsoft’s Windows NT.
NeXT had already delivered its advanced operating system, starting with an initial beta release in 1988. By 1992, it had delivered its third release and was considered a mature operating system.
Like Apple, NeXT needed to move to new and more powerful hardware, and PowerPC seemed like the obvious solution. In addition to new hardware however, Apple also needed a new OS to replace its aging Mac System 7.
Starting under the name Pink, Apple’s new OS turned into a joint project with IBM called Taligent. It intended to replicate and build upon the work already done by NeXT to deliver object oriented frameworks for rapid application development.
The Age of Microkernels: 1990 – 1995.
As IBM planned to move its various hardware platforms to PowerPC, it also developed plans for an operating system kernel and core platform called the Workplace OS that could be reused across its product line to ease the porting of existing software to run on new PowerPC hardware.
Workplace OS was built upon a Mach 3.0 microkernel. It intended to serve as a new PowerPC-ready core for running AIX and OS/2, and eventually Taligent. Apple had also developed its own kernel for Taligent similarly based upon Mach 3.0 microkernel designs; it continued using this NuKernel as the core OS for Copland, its independent internal project to replace Mac System 7.
An exploding trend towards running various systems on top of a microkernel originated with the Mach project at Carnegie Mellon University. Its early versions of Mach retrofit microkernel ideas into the BSD Unix kernel. This was the approach used by NeXT as the core OS for NeXTSTEP.
Mach 3.0 attempted to go even further to isolate the microkernel from the “OS personality” running on top. IBM’s Workplace OS planned to run DOS, OS/2, Unix, OS/400, and potentially any environment on top of Mach 3.0.
Other groups also began similar work on Mach 3.0, including GNU’s Hurd and the OSF’s version of Unix. Microsoft also initially referred to its own efforts to replace DOS with NT as a microkernel design.
Microsoft Promises to Catch Up: 1990 – 1995.
in 1991, Microsoft turned its work with IBM on “OS/2 NT” into the solo Windows NT project. NT promised to deliver software API compatibility with Unix’ POSIX and IBM’s OS/2, as well as a new 32 bit version of its own Windows APIs.
NT was also intended to run across a variety of hardware architectures, including Intel’s x86 PCs, the new PowerPC, and existing workstation hardware including MIPS and Alpha.
Microsoft also promised in 1991 to deliver an object oriented development framework called Cairo to rival NeXT and Taligent’s work, which it said would appear immediately after it delivered the first version of NT.
Wrong on All Counts.
PC magazines convinced the world that Microsoft–a company that had never developed an operating system of its own, had never delivered reliable software, and which had never shipped anything on time–would suddenly deliver a mature OS and set of high quality development frameworks within a few months.
The minority who knew better logically assumed that Apple and IBM’s Taligent would soon be able to match the software already delivered by NeXT. HP, after initially backing NeXT, moved to align with the Taligent project.
It turned out everyone was wrong. As IBM and Apple labored away at developing a perfect OS running on the perfect kernel running on the perfect hardware, problems began to surface:
- Interest in OS/2 rapidly flat-lined after the release of Windows 3.0 in 1990.
- Low end Windows PCs rapidly displaced the potential for growth of Apple’s Macintosh market.
- Corporate culture conflicts between Apple and IBM stalled the progress of Taligent.
- Difficult performance issues related to Mach 3.0 helped derail everything.
The Resurrection of the PC: 1990 – 1995.
The ideal designs related to OS/2, Taligent, and PowerPC initially seemed very promising to observers across the industry, as the Intel based DOS PC clearly needed a replacement. However, the scale of money behind the existing PC resulted in Microsoft and Intel being able to successfully delay their competitors’ progress with plausible notices of intent to deliver their own competing new technology.
Intel invested vast sums of money into building around the limitations of the x86 architecture, adding features commonly associated with RISC architectures. This huge investment allowed the PC to at least maintain its place in the running through the early 90s, using increasingly inefficient designs.
Microsoft had the money to buy an operating system team from DEC, which took over Microsoft’s end of OS/2. Led by Dave Cutler, that work eventually resulted in a design pattered much closer to DEC’s VMS, while also reflecting the current trend toward the microkernel design principles in IBM’s Workplace OS.
In 1993, Microsoft released its new OS as Windows NT 3.1. It attempted to replace the MS-DOS underpinnings of Windows 3.0 with a serious operating system design, but like the Workplace OS, it delivered very poor performance.
Windows NT wouldn’t be ready for consumption for several more years, but it demonstrated how Microsoft could wield its market power to prevent adoption of technically superior software from rivals simply by making announcements.
Future announcements related to the NT successor called Cairo–and really everything Microsoft and Intel promised–had a dampening effect across the industry, creating an illusion of invincibility that would later unravel in series of spectacular failures. Until then, the PC platform that the duo contributed toward came to be hailed as “WinTel,” a moniker neither company liked.
As WinTel PCs homogenized the computing landscape, alternatives from Commodore and Atari fell off into obscurity. Apple’s Mac remained as a viable platform largely because of its entrenched position in education and creative markets, but with the increasingly improved PC crowding into its core markets, Apple’s entire future looked increasingly bleak.
Competing with NeXT to Nothing: 1990 – 1995.
Meanwhile, NeXT was the only company to actually deliver the next generation of software, pairing object oriented development and a powerful core OS into a real desktop product that worked.
The company faced marketing its software technology against the promised features announced for Apple and IBM’s future Taligent and Microsoft’s future Cairo, two vaporware products that never ended up being released.
In 1993, NeXT abruptly abandoned the development of new PowerPC systems and dropped its hardware business entirely to focus on selling its software.
It ported NeXTSTEP to run on industry standard Intel PCs, followed by other ports to run on higher-end workstation hardware from Sun and HP.
While the industry had good things to say about NeXT’s pioneering technologies, it was repeatedly thwarted by its partners, who seemed eager to work with the company, but then repeatedly left it hanging at crucial points:
- IBM paid to license NeXTSTEP but never sold it.
- Sun invested in OpenStep to port it to Solaris, then dropped it for its own Java initiatives.
- HP got involved in OpenStep then dropped it for Taligent.
- Digital planned to use OpenStep on top of OSF/1 but didn’t.
Unix Wars: 1988 – 1995.
Outside of the lower end PC desktop arena, the computing world had largely come to use Unix. Most of the world’s workstation makers were shipping versions licensed from AT&T, including HP’s HP-UX, IBM’s AIX, SGI’s IRIX, SCO Unix for the PC, and Apple’s A/UX for the Mac.
A second branch of Unix resulted from work in the late 70s at UC Berkeley, which became known as the Berkeley Standard Distribution, or BSD. It originally required a license from AT&T for the portions that came from its commercial Unix code, but BSD had also developed its own improvements, the most significant of which was its networking code. This non-AT&T code was freely distributed under a permissive “BSD license.”
In the mid 80s, X/Open started efforts to standardize Unix. AT&T worked with Sun, then a leading distributor of BSD, to merge their software features into a new version called Unix SVR4. Sun adopted the new changes in SunOS 5, which shifted from being a flavor of BSD into Unix SVR4 and took the name Solaris 2.0.
A rival group of workstation vendors viewing Sun as a competitor–including DEC, HP, and IBM–started an independent group called the Open Software Foundation in 1988. The OSF released its own rival unified version of Unix in 1992 called OSF/1.
The fighting between commercial vendors of Unix helped to splinter standards and paved the way for commodity PCs to replace Unix workstations with the common–if greatly inferior–Windows platform. Once Unix vendors recognized that Microsoft was a greater threat than each other, the OSF merged with X/Open to create The Open Group, which now certifies the UNIX trademark and its POSIX compatibility specification.
The Unix/BSD Lawsuit and Linux: 1991 – 1995.
Another skirmish that helped to splinter Unix as a platform related to free distribution of Unix-like operating systems. In 1991, the remains of code proprietary to AT&T were removed from BSD for distribution as a fully free version of Unix. AT&T’s Unix System Labs sued in 1992 and won an injunction to stop distribution.
After being held up for almost two years, the case was settled after some slight changes were made, and in 1994, a legally unencumbered 4.4BSD-Lite version was released.
During that period, the legal uncertainty around BSD pushed momentum behind the new Linux kernel and the Free Software Foundation’s GNU software, which together served as a second free alternative to AT&T’s Unix.
Apple’s History with Unix: 1988 – 1998.
Apple’s connection to IBM and the OSF led it to begin porting OSF/1 to the PowerMac as a replacement for its 680×0-only version of A/UX that it had shipped for classic Macs since 1988. In 1996, Apple decided to ship IBM’s AIX on its workgroup servers instead of developing its own new flavor of Unix.
In 1994, Apple announced efforts to host PowerMac applications on top of AIX and PowerPC using middleware software called MAS, Macintosh Application Services. The confusingly similar MAE, Macintosh Application Environment, earlier enabled non-PowerPC native Mac apps to run on HP and Sun Sparc workstations.
After giving up the idea of licensing the Mac software environment for other platforms, Apple floated MkLinux as an experiment in 1998; it ran the Linux kernel on top of an OSF Mach 3.0 microkernel. The idea was intended to sell Mac hardware to users interested in Linux.
Novell Buys Up A Defense Against Microsoft: 1991 – 1996.
As Windows began gaining significant ground in 1991, Novell bought up Digital Research and began selling its DR-DOS as Novell DOS. This allowed Novell’s NetWare customers to avoid licensing Microsoft’s MS-DOS.
Worried that Microsoft’s increasing monopoly power with Windows and its Office applications would threaten its established networking business, a very spooked Novell–run by Ray Noorda–began working to build up a competitive position for the future through a series of other acquisitions.
To compete with IBM’s OS/2 and Microsoft’s promised NT/Cairo in the post-DOS market, Novell bought up Unix System Labs from AT&T in 1993. Novell hoped to pair AT&T’s UnixWare with its own NetWare business to take on Windows as a server product.
It was Noorda who ended the BSD lawsuit AT&T had started; after buying Unix, he quickly worked to settle the dispute and get back to business. Novell then bought up WordPerfect and Borland’s Quattro Pro spreadsheet to have a competitive office suite to rival Microsoft’s Office.
Viewing Linux as the future, Noorda began a project within Novell to use its assets to compete against Microsoft using open software. However, the 70 year old Noorda was removed as CEO in 1994 and the company almost immediately began selling off his acquisitions.
- In 1995, Novell sold distribution rights to the Unix code it had acquired from AT&T, then called UnixWare, to the Santa Cruise Operation, which began selling it alongside Microsoft’s old Xenix, since renamed SCO Unix.
- In 1996, Novell sold its office apps to Corel.
- In 1996, Novell sold off DR-DOS to Caldera Systems, which was being run by Novell’s former CEO, Noorda.
Caldera Sues Microsoft: 1996 – 2000.
Caldera was a Linux distributor with little love for Microsoft. Along with the then nearly worthless old DR-DOS, Caldera acquired the ongoing dispute between Novell and Microsoft over the market monopolization, illegal tying, exclusive dealing, and tortious interference that had occurred over the previous decade and a half from 1981 to 1996, as Microsoft shut Digital Research out of the market it had created, using a copy of its own product.
Armed with piles of evidence assembled by Digital Research and Novell, Caldera intended to parley the DR-DOS case against Microsoft as a way to fund ongoing development of Linux as a new competitor in the desktop PC operating system market.
Over the next half decade, Caldera continued to argue its case against Microsoft, which ended in 2000 with a $275 million settlement for Caldera in exchange for Microsoft’s demand that all the evidence in the case be destroyed. That was intended to help prevent other companies from developing similar claims against Microsoft’s predatory practices throughout the 90.
Apple Leaves AIM, Shops for a New OS: 1996.
Apple’s partnerships with IBM largely unraveled in 1995. Taligent was taken over by IBM, and the PowerPC partners failed to come up with common platform that could rival the Intel-based PC. Apple ended up the only significant customer of the PowerPC in desktop computers.
The rest of the industry had largely lined up behind the x86 PC and Intel’s promise to deliver its own brand new processor architecture to carry on the torch of the 32 bit x86: the Itanium IA-64. Instead, Itanium turned into a disastrous fiasco that also destroyed MIPS, PA-RISC, and Alpha.
The advantages of PowerPC’s fast and clean design helped Apple to account for its failure to deliver a new generation of software by keeping the Mac OS competitive. After leaving Taligent, Apple then aborted its internal Copland project as an unworkable mess. Apple desperately needed to acquire new, ready to use OS technology that could really deliver the potential of PowerPC Macs.
In 1996, Apple looked at and dismissed Microsoft’s NT, then considered Be, a project developed by Jean Lois Gassée’s, who ran Apple’s Mac development through the late 80s after Steve Jobs left.
Gassée’s BeBox could do great multimedia demonstrations, but in 1996 the BeOS was in early developer release. It lacked basic modern OS features such as multiuser file permissions and security, and even the ability to print. It remained in “preview release” through early 1998.
NeXT Reinvents Apple: 1997 – 2000
Jobs pushed Apple to instead buy NeXT, demonstrating the potential for Apple to inherit NeXT’s Enterprise business in addition to using NeXTSTEP to inject modern technology into its consumer Macs. After the sale, the migration to NeXT software was discovered to be a bigger challenge than first anticipated. Additionally, NeXT’s Enterprise clients shrank from doing business with Apple, preferring instead to align with WinTel.
NeXT management reinvigorated Apple, shaking off its extraneous business to focus on its core markets in education and consumer systems. Under Jobs, Apple also ended its litigation against Microsoft, which had stretched along for over a decade. That refocusing left Apple in a position to compete on its own merits.
In 2000, the Apple revealed plans for Mac OS X, which would bridge the Mac platform into a new generation of software delivering upon the advances NeXT had pioneered a decade earlier. That technology would be aimed squarely at the consumer market.
Doomed to Repeat it: the 2000s.
Microsoft immediately announced plans to deliver a copy of Apple’s technology, even while the majority of its users were still running Windows on top of DOS.
The 1991 unfulfilled promise to deliver Cairo and catch up with NeXTSTEP was recycled into a new promise for 2001 to deliver Longhorn with the features of Mac OS X in just a couple years.
[next: things really get going in the 2000s]
Like reading RoughlyDrafted? Share articles with your friends, link from your blog, and subscribe to my podcast! Submit to Reddit or Slashdot, or consider making a small donation supporting this site. Thanks!