Apple's High-Tech Gamble
In the 1970s, VisiCalc was the killer application that made the Apple II a hit. In the 1980s, desktop publishing did the same for the Macintosh. In the 1990s, Apple is hoping once again that breakthrough technology will help it leapfrog the competition.
By Tom R. Halfhill
BYTE
December 1994
Apple is gambling its future on three new technologies that could transform the computer industry and redeem Apple's destiny. But if those technologies should fail, or unfold too slowly, Apple probably won't finish the century as a leading computer company.
If that judgment sounds too harsh, consider recent events. Apple has slashed prices to stay competitive with companies selling commodity PC clones, at the expense of watching its once-luxurious profit margins tumble by half. Thousands of employees have been laid off, including high-profile CEO John Sculley, who was replaced with a more down-to-earth manager, Michael Spindler. New operating systems from Microsoft are catching up with, and even surpassing, the Mac OS in key areas of technology, while PCs are finally acquiring some long-standing hardware advantages of the Macintosh, such as plug and play. Most of all, in a risky effort to increase its stagnant worldwide market share--10 percent, according to market-research firm Dataquest (San Jose, CA)--Apple is turning to a drastic strategy: licensing other vendors to clone the Mac.
Of course, dire predictions are nothing new for Apple, and few companies have such a remarkable record of confounding the experts as Apple does. In the past, some new technology has always come along to save the day. That's because Apple is a technology-driven company that relies on innovation to differentiate its products from those of the bargain-basement clone makers. Unique technology is what allowed Apple to charge high prices in the Sculley era, when profit margins took precedence over market share.
Although Apple is now pursuing a market-share strategy and has altered the price/performance equation in its favor, make no mistake: Apple still considers itself to be a technology company. It has no more intention of competing on price alone in the 1990s than it did in the 1980s. Once again, Apple is gambling on technology--actually, three new technologies--to rescue its fortunes.
By far the company's biggest bet is on RISC. Apple believes that the horsepower of RISC processors will be required to run the system software and applications of the future, and that anyone who invests in CISC hardware and software is on a dead-end road. RISC chips, such as the PowerPC, says Apple, will keep getting faster and less expensive, while CISC chips, such as the x86, will run out of gas. Operating systems closely tied to the x86, such as Windows 95, are equally doomed in this view. Eventually, says Apple, PC users will realize they must move to RISC, just as they eventually accepted the GUI and numerous other technologies pioneered by the Mac.
Apple's second bet is on the 11-year-old Mac OS. Although the operating system is falling behind the competition in some important categories--most notably, preemptive multitasking and memory protection--it still offers features that are just starting to appear in Windows, such as 32-bit addressing and color matching. Its collaboration software, digital signatures, speech recognition, and telephony functions have been shipping for a year or more. Plug and play has been standard since 1987, and Apple is applying this experience to new device interfaces, such as PCI (Peripheral Component Interconnect) and PCMCIA.
Apple's third bet is that large, monolithic software will give way to small, efficient componentware. This is another obvious shot at Microsoft, which happens to be the market leader in large, monolithic software. Apple is counting on new technologies such as OpenDoc and the Taligent operating system to catch on.
The idea behind componentware is that users can build their own customized solutions by linking together software parts optimized for specific tasks. Microsoft's OLE is a similar technology, but Microsoft is trying to protect its large applications (e.g., Excel and Word) by promoting them as components. Apple is betting that componentware will not only change the way people use computers but will also change the software industry by lowering the barriers that now discourage small, innovative developers.
In addition to these three technologies, Apple is making a wild-card wager on mobile computing. Laptops, such as the PowerBooks, are definitely part of this strategy, but Apple is preparing for the day when PDAs (personal digital assistants) are as ubiquitous as calculators. Despite the almost unprecedented storm of negative publicity that drowned the introduction of the Newton MessagePad, Apple has not retreated from this technology. On the contrary, Apple is quietly laying a foundation that could give the Newton a crucial advantage when the hardware catches up with the vision. (See the text box ``The Newton and the Hare'' on page 52.)
Several other technologies and products are important to Apple--indeed, the company constantly wrestles with overdiversification--but the PowerPC, the Mac OS, and componentware appear to make up the fulcrum on which Apple's future is balanced. Apple might survive a wrong bet in one or two of these areas, but if all of them tilt the wrong way, the company will find itself in serious trouble. Although outright collapse is unlikely, Apple could succumb to a fate almost worse than death: irrelevance. That's what happened to Commodore after its disastrous fall in the mid-1980s, despite the cutting-edge technology of the Amiga.
The RISC Debate
Apple believes that users who buy x86-based PCs are stampeding headlong into a brick wall: CISC processors, such as the Pentium and its successors, won't deliver the raw performance demand ed by tomorrow's advanced operating systems and applications. Eventually, PC users will realize their folly and turn to RISC, rendering most of their hardware and software obsolete. Why not switch now rather than later?
"Apple has already made the transition," says Ian Diery, executive vice president and general manager of Apple's Personal Computer Division. "We're already a known quantity. That's going to help us gain market share. I'm not suggesting that Intel and Windows are going to disappear, but I do think Apple will prosper because we are making the RISC transition sooner."
It's a compelling argument--one that's widely supported by other companies with a vested interest in RISC: DEC, IBM, Motorola, SGI/Mips, and Sun Microsystems. And they certainly have a lot of cards in their favor. All the latest and fastest microprocessors are pure RISC designs: DEC's Alpha 21164, IBM/Motorola's PowerPC 604 and 620, Mips's T5, and Sun's UltraSparc. The newest architectural innovations usually appea r first on RISC, and RISC chips generally have smaller dies and fewer transistors than comparable CISC processors, so they cost less to make.
Furthermore, say RISC proponents, the price/performance gap between CISC and RISC is widening. They claim that CISC in general, and the 16-year-old x86 architecture in particular, is on a flattening curve that will not keep pace with RISC. Apple says its first Power Macs with PowerPC 604 and 620 chips--which are planned for 1995 and 1996--will blow away the Intel-based boxes and make this fork in the road crystal-clear to even the most loyal PC devotee.
The truth is not quite so clear-cut, however. For one thing, the manufacturing costs of microprocessors are determined by volume as well as die size, and nobody approaches Intel's volumes. Dataquest says Intel holds 74 percent of the worldwide microprocessor market, followed by Motorola, which has 8 percent. Part of the reason why Pentiums cost more than RISC chips may be that Intel is simply amortizing its expensive chip factories more quickly. At $1 billion per factory, that's no small factor.
What's even more important is the fact that the architectural differences between CISC and RISC are becoming rather fuzzy. The newest x86-compatible chip is the K5 from Advanced Micro Devices (Sunnyvale, CA). It's virtually a pure RISC design: quad-issue superscalar pipelines, five parallel functional units, dynamic branch prediction, speculative execution, out-of-order execution, an expanded register file, and large primary caches. To feed this core, the K5 has an ingenious decoder that reduces complex x86 instructions to RISC-like operations that issue in parallel (see "AMD vs. Superman," November BYTE). The latest x86 chips from Cyrix (Richardson, TX) and NexGen (Milpitas, CA) are similar CISC/RISC hybrids.
CPU design is so wide open that there is no architectural technique used in RISC processors that doesn't find its way into CISC chips. The only real difference goes back to the original definitions of CISC and RISC, which stated the relative complexity of the instruction sets. CISC chips still must devote more cycles and transistors to decoding complex, variable-length instructions. But even that distinction is minimized by modern compilers, which generate most of their code using a subset of faster, simpler instructions.
If CISC isn't dying, RISC fans counter, then why did Intel recently form a partnership with Hewlett-Packard to develop a new microprocessor based on HP's Precision Architecture RISC technology? Again, this question dodges the basic issue, which isn't really RISC versus CISC at all. For one thing, the jointly developed Intel/HP chip may not fall neatly into either category; rumors indicate that it will use VLIW (very long instruction word) technology. More to the point, the chip will be backward compatible with the x86, underlining that this is really a battle between instruction sets, not architectural styles.
Right now, Intel is winning the instruction-set battle. Apple has done a commendable job of building a solid base of native PowerPC software--more than 200 applications, including most of the mainstream programs that Mac users really need. But so far, there aren't any truly unique programs that aren't also available for the x86. There's no killer application like Aldus PageMaker, which can't run in a character-oriented environment such as DOS. Although Apple has established itself as the leading vendor of RISC-based personal computers, the x86 instruction set continues to attract more users and software, not less.
This is especially true in the business world, where the Mac has only about 6 percent of the market, compared to 14 percent in homes and 28 percent in schools worldwide, according to Dataquest (see the text box "Macs in the Enterprise" on page 64). Even Apple bows to this reality by promoting x86 compatibility on the Macintosh, either with emulators, such as SoftWindows, or plug-in x86 boards.
Apple's experiment last spring with the 486 Houdini board for the Quadra 610 was so successful that a new line of 486 boards for the Power Macs was to be announced in November (see the text box "Houdini Reappears" above). Apple is also rumored to be looking at IBM's PowerPC 615, a RISC chip that will execute x86 instructions--virtually the same concept as that used for the Intel/HP hybrid processor.
Apple embraced RISC not because CISC is inherently inferior, but because Motorola's CISC chips were falling behind Intel's and there was no viable alternative. If Motorola had shipped the 68040 and the 68060 a year before Intel released the 486 and the Pentium, Apple might still be in the CISC camp.
Finally, while it's true that the latest CISC chips are not as fast as their RISC counterparts and also tend to cost more, the price/performance delta does not directly translate into street prices for complete systems. For instance, the PowerPC 601 chip costs about half as much as a comparable Pentium chip, but a 601-based Power Mac does not cost half as much as a Pentium-based PC. The Mac is slightly less expensive and slightly faster, but currently the differences are not dramatic enough to lure significant herds of buyers away from the x86.
Widening the Gap
Part of the reason why the PowerPC's edge over the Pentium does not translate into desktop performance is that Apple's transition to RISC was tied to backward compatibility, not maximum performance. The primary goal of Apple was to preserve its users' investment in existing software and hardware.
Moving to a new microprocessor architecture is so fraught with danger that only one personal computer company has ever successfully made such a transition: Apple. In the 1980s, Apple successfully switched from the 6502-based Apple II series to the 680x0-based Mac. Competitors who attempted the same metamorphosis (e.g., Commodore and Atari) failed to repeat their past glories. Tandy appeared to survive its move from the Z80 to the x86, but the company later sold off its computer manufacturing to AST Research.
Operating-system transitions are hazardous, too. When IBM and Microsoft first tried to move PC users from DOS to OS/2, very little software was available, and DOS programs couldn't be multitasked. After the companies' divorce, OS/2's slow takeoff opened up a window of opportunity for Microsoft.
Recognizing this danger, Apple made several compromises while designing the Power Macs. The biggest compromise was not porting the entire Mac OS to PowerPC-native code. Of course, time-to-market considerations also played a major role in this decision, but the result is that most of the Mac OS is still 680x0 code. The operating system relies on the 68LC040 emulator--the same one used by 680x0 applications software--to run the nonnative portions of itself. In fact, only about 15 percent of the Mac Toolbox--the low-level functions called by the Mac OS--runs natively.
Skeptics (including this writer) questioned the wisdom of this scheme, but Apple miraculously pulled it off. Indeed, the difficult shift to a new CPU architecture has broken less software than earlier, relatively minor, transitions--such as the move from the 68030 to the 68040 or from System 6 to System 7 (or, for that matter, from Windows 3.0 to Windows 3.1). But now Apple needs to pay more attention to performance.
The Mac OS is so vast (with more than 4000 Toolbox calls), and so much of it is written in hand-tuned 680x0 assembly language, that it's unlikely the whole thing will ever be ported to the PowerPC. Instead, Apple is concentrating on the most frequently called and the most time-consuming routines. This makes sense, but the process is progressing slowly. BYTE and others think Apple should release native Mac OS components in stages to lengthen its performance lead over the Pentium.
Hardware improvements are also required. Apple is moving toward a more industry-standard platform that will provide more headroom for better performance, but its laudable desire to protect the investment of existing users will spread the project over a period of two or three years (see "Apple Redefines the Macintosh," November BYTE).
In any case, a RISC victory still won't guarantee a triumph for Apple, because Microsoft is covering both bets: If CISC sinks, Microsoft has Windows NT, which runs on the Mips R4x00, DEC's Alpha, and, soon, the PowerPC. IBM's OS/2 is moving to RISC as well, and Unix was there all along. In other words, Apple isn't the only RISC game in town.
Microsoft's "Windows everywhere" strategy erects a safety net for software developers, too. By writing for the Win32 API and avoiding x86 dependencies, developers can ship to the huge x86 market now and easily recompile for a variety of RISC chips if CISC goes sour later. Options for Mac developers are more limited because the Mac OS runs only on the Mac.
Operating-System Wars
Behind Apple's argument for RISC is the promise that extra horsepower will raise system software and applications to a new level of performance unattainable by CISC. But staying ahead of Microsoft and other competitors will not be easy.
Already, millions of users don't perceive a qualitative difference between Windows 3.1 and the Mac OS. Apple's attempt to defend the Mac's uniqueness by suing Microsoft in federal court backfired badly; when the court ruled that almost no aspect of the Mac's GUI is protectable, the doors opened wide for Microsoft to blur the differences even further. As a result, Windows 95 sports a radical, Mac-like face-lift, and the Mac is losing its most visible advantages.
Of even greater importance, however, are the less-visible features. Windows 95 and Windows NT (not to mention OS/2) are pulling ahead of the Mac in some critical areas: preemptive multitasking, multithreading, memory protection, multiprocessing, security, and portability. The last three features are provided only by NT, and Windows 95 has yet to ship, but Microsoft still makes a strong case that RISC isn't a prerequisite for state-of-the-art system software.
Although Apple added a thread manager to System 7.5 (which was released in September), other advanced features require architectural changes that will shake the foundations of the 11-year-old Mac OS. The transition to RISC and other diversions have delayed this critical project. The next major release of the Mac OS, code-named Copland, is not expected until late 1995. It will be followed by another major upgrade, Gershwin, in 1996.
Among the improvements in Copland will be a microkernel, limited memory protection, and a new I/O architecture. These are the first steps toward full preemptive multitasking and memory protection, which won't arrive until Gershwin's debut.
Copland's microkernel will allow such low-level operations as concurrent I/O and networking to multitask preemptively; it will also allow applications to spin off preemptive tasks in protected memory partitions. But the applications themselves will continue to multitask cooperatively in a shared memory space. The limitation is that most Toolbox routines won't be reentrant in Copland, and the only tasks that can take advantage of protected memory and preemptive scheduling are those that call reentrant code. Fortunately, because device drivers and network services call lower-level routines that will be reentrant, Copland should measurably improve the Mac's reliability and performance.
Gershwin will convert still more of the Toolbox to reentrant code, allowing all tasks to run preemptively in protected memory. Any of the less important Toolbox calls that aren't reentrant could be replicated in protected memory to avoid scheduling conflicts.
A lot more code in Copland will be native to the PowerPC. Apple is vague on details but says more than half the code will be native. However, much of this native code will be in new parts of the Mac OS, not the existing Toolbox. As mentioned before, some Toolbox routines simply aren't called often enough to warrant a rewrite. Another consideration is the code expansion that's inevitable when finely tuned CISC code is ported to RISC.
Other new features of Copland will include improved networking, a revamped user interface, and integrated OpenDoc (which will be released as an extension for System 7.x before Copland ships). The improved networking will come via Open Transport, a new abstraction layer that shields applications from network protocols. By communicating with this API, programs don't need to know whether the underlying network protocol is AppleTalk, NetWare, TCP/IP, or whatever.
But the new user interface will not be the kind of major face-lift that sets Windows 95 apart from Windows 3.1. Apple is coy about this because it doesn't want to tip its hand to Microsoft, but the general direction is toward a GUI that is scalable for users of different skill levels and that offers much more active assistance--Apple's term for interactive on-line help assisted by smart agents or scripts. The first examples of active assistance appeared in the Newton operating system and the Apple Guide extension to System 7.5. It's impressive technology, going beyond static on-line help to walk users through complex tasks or to automatically perform tasks delegated by the user.
Apple's new GUI may take on some of the look and feel of the company's At Ease software, which is bundled with home-market Performas. The Finder is hidden behind a simplified GUI that makes it easier to launch programs and more difficult to delete important files. With a password option, it can present different working environments to different users on the same computer. At Ease is a remarkably elegant solution to the problem of sharing a Macintosh among family members, some of whom (either the adults or the children) may be computer neophytes (see the two screens on page 58).
To get some idea of where Apple is going next, simply imagine At Ease and the traditional Finder as two disparate points on a spectrum of usability. Now imagine a highly customizable environment that can be adjusted to any point in between. Add active assistance, and you've got a rough picture of Apple's future GUI.
This could be Apple's best chance to reassert the Mac's uniqueness. Apple is justly famous for its human-interface design and traditionally leads the industry in that area. It's no coincidence that every new version of Windows looks more and more like the Mac. While Windows users are celebrating the end of their 10-year wait for long filenames, Apple is pushing forward into bold new territory.
But the flashy new GUI still needs an industrial-strength foundation, especially if Apple wants to make inroads into corporations. Full preemptive multitasking, memory protection, and advanced 3-D graphics won't arrive until Gershwin's debut. Mac users are fairly well served by the robust cooperative multitasking and crash recovery of System 7.x, but Apple would be playing a stronger hand if it could deliver these basics sooner.
The Clone Question
One significant benefit of Copland's microkernel will be hardware independence. The Mac OS has always been tightly integrated with the Mac's underlying hardware. Removing those dependencies will make it easier for Apple to license Mac clones and will also move the Mac system architecture toward a more industry-standard platform. Both objectives are critical to Apple's future. Indeed, how this process unwinds will largely determine whether Apple ends the decade as a mainstream computer vendor or as a niche player.
The Mac's tightly integrated architecture is perhaps its greatest and most invisible strength. The advantages of tight integration didn't become obvious until the multimedia craze spread to PC users, who suddenly faced the dilemma of installing sound boards, video cards, CD-ROM drives, and other conflict-prone paraphernalia. If the resulting frustration didn't trigger a blip in the national suicide rate, it at least forced Microsoft and Intel to patch the PC system architecture with a catch-up technology called Plug and Play (see "Transforming the PC: Plug and Play," September BYTE).
Mac users have taken plug and play for granted since the introduction of the Mac II in 1987, and it probably accounts for Apple's strong showing in the home and educational markets. But the Mac OS and hardware are so interdependent that they pose problems for licensees working on Mac clones.
Licensing the Mac OS isn't just a matter of releasing the system software; everything is closely tied to custom ROMs and ASICs (application-specific ICs). In fact, Apple is sharing Verilog models of its ASICs with licensees and is offering them the same volume discounts that Apple gets from its own suppliers. But this does not give the clone makers much latitude for differentiation. Apple's challenge is to break the system dependencies without losing the advantages of integration.
The first PC clone makers in the 1980s didn't confront this problem because the PC system architecture and MS-DOS were never so tightly integrated. Also, cloning the IBM PC was much easier because it wasn't nearly as sophisticated as the Mac. The biggest hurdle was writing a clean-room BIOS, which back then contained less than 32 KB of low-level code. Incidentally, the clone makers did this without IBM's help or blessing. Contrary to popular belief, IBM did not openly license the PC system architecture in those days; it was as proprietary as the Mac's. Only the operating system, Microsoft's MS-DOS, was openly licensed.
No one ever successfully cloned the Mac because the task of reverse-engineering its architecture was too daunting. Apple's decision to license the complete Mac architecture and overhaul it to help clone makers is therefore unprecedented.
Toward Common Ground
Licensing is a vital part of Apple's strategy to gain market share, but there is a more important motivation for decoupling the Mac OS from the metal. Apple wants to evolve the Mac toward a common PowerPC platform supported by IBM and other PowerPC system vendors.
While Apple has been shipping hundreds of thousands of Power Macs, IBM has been conspicuously quiet. IBM is selling PowerPC-based workstations, but no PCs. And while Apple is steadily accumulating a base of PowerPC applications software, there is no software for IBM's PReP (PowerPC Reference Platform). In fact, an industry joke is that PReP has more operating systems than applications (this is not quite true: Windows NT and OS/2 for PowerPC aren't available yet, either).
In great secrecy, Apple and IBM are negotiating the future of the PowerPC platform. The problem is that Apple's Power Macs and IBM's proposed PReP systems are fundamentally incompatible. There are several reasons for this, but the most vexing one comes down to the order in which bytes are arranged. Apple, with its 680x0 heritage, is big-endian; IBM, with its x86 heritage, is little-endian.
PowerPC chips can switch between the two modes, but hardware dependencies in the Power Macs prevent them from cold-booting a little-endian operating system, such as OS/2 or NT. That's why SoftWindows must be launched on top of the Mac OS, wasting megabytes of RAM and precious processor cycles to simultaneously run two operating systems. Similar dependencies also prevent the Mac OS from running on PReP systems.
If the two sides don't reach some common ground, buyers will be confused by two mutually incompatible PowerPC platforms. It would be as if some x86-based PCs ran DOS and Windows software, while others required a whole different software library. Such a rift could prevent both PowerPC platforms from winning substantial market share.
Apple says that it needs at least a year to disentangle the Mac OS from the metal without breaking too much existing software. IBM, obviously, doesn't want to wait that long to launch PReP. Apple would seem to have a stronger position in these negotiations; unlike IBM, Apple has a growing installed base of PowerPC systems and software.
The ideal solution, from Apple's viewpoint, would probably be for IBM to manufacture Mac clones under license while the two companies jointly evolved the architecture toward a more standard platform. Working together, they could probably accelerate this process and quickly reach something very much like PReP. Apple and IBM proved they can cooperate in this fashion by developing the Power-PC with Motorola. This new project, like that one, would serve the interests of both parties.
If Apple and IBM cannot agree on a course of action--and do so soon, ideally before the release of Windows 95--the PowerPC will pose a much smaller threat to the dominance of Intel and Microsoft. Apple and IBM could both end up as minor players in a high-stakes game.
Reinventing Software
The snowballing success of Windows over the past four years stole the Mac's uniqueness in more ways than one. In addition to retrofitting the PC with a Mac-like GUI, Windows lured away the software developers whose revolutionary applications set the Mac apart from any computer that had come before it.
None of the major developers writes exclusively for the Mac anymore. Nor do they automatically release their Mac versions sooner, as they once did. Programs such as QuarkXPress, PageMaker, and Adobe Photoshop bring equal capabilities to both platforms, and the lucrative Windows market has attracted some big-league competition of its own, such as Aldus PhotoStyler and Corel's CorelDraw. Apple clings to a slim lead in some areas--for example, HSC Software's Live Picture is a breakthrough image editor available only for the Mac--but the old days of Mac supremacy are gone.
If RISC comes through and Apple can sell Power Macs that are insanely greater than x86-based PCs, there's a chance that Apple could reclaim its advantage. But the performance gap would have to open so wide that the new programs wouldn't work under Windows. And, of course, developers would have to find the inspiration to write those programs.
Apple is hedging this bet with another strategy: component software. If this concept pays off, big developers and applications might not be as important as they are today, and the Mac could attract some compelling software from a new generation of smaller, more inventive developers. Since Apple doesn't have a significant applications software business, it has little to lose--and possibly much to gain--by shaking up the software market.
"The software industry is in terrible shape," explains Apple's Ian Diery. "The market is dominated by one huge software company and by a small number of huge software applications. We have to move away from these huge programs to make room for more innovation."
If these big, monolithic programs could be broken up into smaller components, or if their features could be imported into other applications, the fatware trend might be halted. As a by-product, computing would be reoriented around the task, not the application. Users would work on compound documents, and systems would switch seamlessly among different programs and components (see "Componentware," May BYTE).
Microsoft is moving in this direction with OLE. Apple's answers are OpenDoc and Taligent. All three technologies have a similar goal: to reinvent the way software is constructed and used. However, Microsoft is taking a more conservative approach, promoting its huge programs as being both exportable components and containers for other components. Apple has a somewhat different model in mind and is portraying OLE as the more limited, proprietary solution. (It's proprietary because Microsoft controls OLE, while Taligent and OpenDoc are steered by open alliances of several different companies.) The respective roles that Taligent and OpenDoc will play in Apple's future are not entirely clear, but it looks as if Taligent will be more specifically an enterprise solution, while OpenDoc will address the full range of users.
Components Everywhere
Taligent is an outgrowth of Apple's secret next-generation operating system, formerly known as Pink. Originally it rested on a microkernel, called Opus, that was 680x0-specific, but it was switched to a Mach 3.0 microkernel when Apple and IBM (who were later joined by HP) formed the partnership that became Taligent. Opus wasn't extensible to other platforms, and it also lacked industrial-strength security and multithreading. Mach is both more powerful and more portable, and it's the same microkernel that lies beneath Workplace OS.
TalAE (Taligent Application Environment) is the object-oriented framework that rests on Mach. It contains 1730 public object classes (and an equal number of nonpublic classes) and a stunning 53,000 methods, compared to about 4000 calls in the Mac API and 1500 in Windows. Indeed, TalAE is so rich in objects that Michael J. Potel, Taligent's vice president of technology development, describes it as "a whole OS of nothing but hooks."
Yet TalAE has just over 600,000 lines of code, compared to about 4 million lines in Windows NT and an estimated 16 million lines in Cairo, an object-oriented version of NT scheduled for 1995. In Tal AE, objects can be as small as a single character and can even sit on the processor stack or in a register.
As Next discovered, convincing the world to switch to a whole new operating system isn't easy, despite the promised benefits of object-oriented computing. So, in addition to TalAE, Taligent is offering its technology in the form of TalOS (Taligent Object Services), which provides portable system-software services, and TalDE (Taligent Development Environment). TalAE is portable across several CPU architectures (the PowerPC, x86, and PA-RISC) and operating systems (OS/2, AIX, PowerOpen, HP-UX, NT, and, eventually, the Mac OS).
TalAE, TalOS, and TalDE are scheduled to be released in phases throughout 1995. But TalAE won't run on the Mac OS until Gershwin's arrival in 1996, because it requires preemptive multitasking.
OpenDoc delivers componentware by another path. It is scheduled to ship as an add-on to the Mac OS, Windows, and OS/2 in 1995, and it will be integrated into Copland and O S/2 later. OpenDoc's evolution is controlled by an independent alliance known as Component Integration Laboratories (Sunnyvale, CA)--CI Labs for short--whose backers include Adobe, Apple, Borland, IBM, Lotus, Novell, Oracle, Sun, and Xerox. Because OpenDoc appears to be a less radical make-over than Taligent, it will probably require less commitment from developers and users.
OpenDoc also promises several advantages over OLE, including wider cross-platform support (Mac, Windows, and OS/2); easier development (the OpenDoc API is claimed to be one-third the size of OLE's, and it handles the multitasking and event messaging that OLE pushes on developers); version control (successive changes to a document are stored as part of OpenDoc's Bento format); and more functionality (unlike OLE components, OpenDoc parts can be nested more than one level deep and aren't limited to rectangular screen regions).
But OLE, on the other hand, is already here and is gathering broad support from developers. And Microsoft is doing a good job of supporting OLE with its development tools and foundation classes.
Even OpenDoc's supporters know that OLE cannot be ignored, so CI Labs is working to make OpenDoc OLE-compatible. In fact, it is trying to turn OLE into a subset of OpenDoc; by writing to the OpenDoc API, developers could automatically gain OLE operability. CI Labs demonstrated this at Apple's Worldwide Developer's Conference last spring.
Technology demos are one thing, but do users really care about components? There is strong evidence that they do, often without realizing it. Photoshop Plug-Ins, QuarkXPress XTensions, Aldus Additions, and VBXes (Visual Basic custom controls) are all examples of successful componentware. But those components are specific to their applications. OpenDoc and Taligent promise a richer vein of components that will work across all applications and platforms.
"We didn't just stumble into component software," says Karl May, manager of solutions tools at Apple. "Customers may not be clamoring for components, but they are complaining about the problems they have, like increasing system requirements for software."
Eventually, users will embrace components, but that doesn't guarantee the software industry will change in Apple's favor. The big software publishers (who are almost unanimously supporting componentware) might still dominate the market, and Windows might still end up with the most components.
Apple's bet on componentware is a good one, though. If OpenDoc and Taligent live up to their claims, Apple could stir up the market, stem Microsoft's momentum, and gain lots of new software for the Mac.
Can Apple Change?
This article barely explores the many fronts on which Apple is fighting for its future. No mention has been made of Apple's strategies for enterprise servers, PowerBooks, printers, digital cameras, or system software for interactive TV set-top boxes. In BYTE's assessment, the outlook for all these products depends on the success of the core technologies discussed here. If the core technologies flop and an auxiliary product line emerges, it will be Apple that is transformed, not the computer industry.
Nor is this analysis intended to spread doom and gloom. Although Apple's market share is small, its volumes are quite healthy. Apple, Compaq, and IBM are virtually tied for the position of the world's leading computer vendor. Apple is pushing ahead with some exciting new technology, and the industry would be much poorer without Apple's technical leadership.
Indeed, it would not be an exaggeration to describe the history of the computer industry for the past decade as a massive effort to keep up with Apple. In 1984, critics derided the Mac for its appliance-like simplicity, but it went on to pioneer or popularize almost every innovation in personal computing: the GUI, desktop publishing, built-in networking, plug and play, integrated multimedia, API-based software development, visual programming, hypertext, 24-bit color, the global clipboard, undo, voice control, built-in color calibration, dynamic memory allocation, SCSI, and even 3 1/2-inch floppy drives. Apple's R&D lab, located in Cupertino, California, is the inspirational R&D center for the entire industry.
If Apple deserves blame for keeping the Mac proprietary for so many years, consider that few other companies have had as much unique technology to protect. It's easy to be open if your technology breaks no new ground. However, don't read too much into Apple's decision to license the Mac. It is not so much a change of corporate direction as it is an attempt to defend the Mac's viability as an alternative platform. Apple still seems bound by a mind-set that in some ways is the opposite of Microsoft's.
For instance, Apple needs to strengthen its development tools. HyperCard's debut in 1987 brought visual programming to the masses for the first time, but Apple let HyperCard languish for years. Not until Microsoft released Visual Basic was the HyperCard concept advanced toward its full potential.
Apple's componentware strategy would be a lot more credible if it offered a tool like Visual Basic, which has sold well over a million copies, mostly to the corporate developers Apple hungers for. HyperCard and AppleScript let you control OpenDoc containers and parts, but not create them. Apple hints that this gap may be filled by a future tool or by an enhanced version of HyperCard: "Apple is aware of Visual Basic and has analyzed what Visual Basic does--not just what it does technically, but also the market perception of what it does," says Apple's May. "That is part of what goes into our thinking when we plan what we'll do with HyperCard."
Apple has also been slow to support the Power Macs with good tools. In fact, all the software released at the introduction of the Power Macs last spring was built with beta versions of compilers because final versions were not available. There would be a lot less Power Mac software in existence today if a small Canadian company, Metrowerks (St.-Laurent, Quebec, Canada), had not plugged the gap with CodeWarrior.
The next missed opportunity could be Apple's QuickTime VR, an astounding virtual-reality technology that can meld a series of photographs or rendered images into a seamless 360-degree panorama. Without any special hardware, you can pan the wraparound image in complete circles, zoom in on details, and examine objects from any angle. The software automatically corrects distorted perspectives and smooths out jagged lines with real-time antialiasing. QuickTime VR is practically a reinvention of photography, and it's also a prime example of the innovation that continues to emerge from Apple's Advanced Technology Group.
Another company might put this wonderful creative tool in a box and sell millions of copies at a few hundred dollars a pop, as Microsoft did with Visual Basic. Within months, there would be thousands of commercial and freeware panoramas available on CD-ROMs, BBSes, and the Internet. But Apple is restricting the QuickTime VR tools to a relatively small group of registered developers--a marketing strategy that seems perversely devised to capture the smallest possible market share.
Technology has never been a problem for Apple; marketing is another story. You would expect a company that has scored as many technological firsts as Apple to dominate the industry, yet the PC industry is swamped by clone makers who rarely, if ever, invent anything new.
Now, once again, Apple is gambling its future on new technology. Can it shift market forces, cut through fierce competition, and regain momentum? One thing is for sure: Apple will raise the technological ante for the whole industry.
Acknowledgements
BYTE senior technical editors at large Tom Thompson and Jon Udell also contributed to this article.
Apple's Technology Milestones
April 1, 1976:
April 1977:
October 1979:
January 1983:
January 1984:
January 1985:
January 1986:
March 1987:
August 1987:
March 1988:
September 1988:
April 1989:
January 1991:
January 1991:
May 1991:
May 1992:
November 1992:
January 1993:
July 1993:
August 1993:
October 1993:
October 1993:
March 1994:
March 1994:
March 1994:
August 1994:
Fall 1994:
Illustration: At Ease, a simplified user interface that's bundled with Macs for home and educational use, provides a glimpse of Apple's future GUI. This GUI will be adjustable for users of different skill levels, ranging from beginner to expert. The log-in screen (top) allows At Ease to present different environments to different users of the same computer; the other screen ha s single-click buttons that make it easier for neophytes to launch programs. Note the absence of the Trashcan and drive icons; this prevents neophyte users from accidentally deleting important application or system files.
Tom R. Halfhill is a BYTE senior news editor based in San Mateo, California. You can reach him on the Internet or BIX at thalfhill@bix.com.
Copyright © 1994 CMP Media LLC