A microcomputer mess is on the horizon -- worse than any we've seen. Here's what your company should do

The microcomputer industry today is facing its most chaotic time since 1982, when the IBM PC took over the business market. No company that uses microcomputers will be spared from the confusion, but those using an IBM PC model or a PC clone will be the hardest hit. Understanding the reasons behind the chaos will help you guide your company through the coming years of instability.

For a while the industry seemed relatively settled, and buying a microcomputer for a business was often a simple matter of getting an IBM PC or compatible. Fast or slow, all these models worked essentially the same way. They all ran the same application software -- Lotus 1-2-3 for spreadsheets or WordPerfect for word processing, for example -- and everything was compatible.

Not for long. The IBM PC standard is fragmenting into three classes of hardware and software, based around three different central processing chips that are only partly compatible. The breakup is inevitable because the standard was based on IBM's original PC, and continuing to follow it would impede the development of advanced software that could take advantage of newer, more powerful microcomputer models.

The problem began at the heart of IBM's first PC -- with the central processing unit (CPU), the 8088 chip made by Intel Corp. (This chip is virtually identical to Intel's 8086 chip, used in many current low-end IBM models and clones.) To manage the interactions among the PC's components -- including the CPU and other hardware, memory, and application software -- IBM chose MS-DOS (Microsoft Disk Operating System; called PC-DOS by IBM). Software written for MS-DOS and IBM's hardware design can run on any IBM model or clone.

A critical measure of a CPU's power is the amount of random-access memory it can address; more memory makes more sophisticated software possible. The 8088 can address one megabyte of memory. Out of the one megabyte on an IBM PC, 360 kilobytes are reserved for such system functions as managing the video display; 640 kilobytes remain for software. This "memory barrier" limits what software can do. At its simplest, it restricts the size of spreadsheets you can work on or the number of files you can view simultaneously. At a more complex level, tremendous memory will be required for advanced software that can actually adapt to your style of working by automatically recalling how you have performed similar tasks in the past.

In 1984, IBM introduced the PC/AT with an Intel 80286 CPU (286 for short), which can address up to 16 megabytes of memory. Software written for an 8088 CPU can run on a 286, but software written specifically for a 286 cannot run on an 8088. IBM maintained compatibility with its earlier models simply by ignoring the extra capability of the 286. For the PC/AT, MS-DOS remained unchanged, so application software was still constrained by the 640-kilobyte limit. The 286 was and is used for its higher computing speed rather than for its additional power.

The newest, most expensive class of PC compatibles, first introduced by Compaq Computer Corp. in 1986, are built around Intel's 80386 chip (the 386). The 386 runs software written for the 8088 and 286, but software specifically designed for the 386 can not run on the earlier CPUs. A handful of programs has been written for the 386, but most people use a 386 machine merely to run 8088 software faster than a 286 machine can. Like their predecessors, 386-based models are still running the same MS-DOS software with the same 640-kilobyte limit, even though the 386 can address up to four gigabytes -- 4,096 megabytes -- of memory. (With the circuits available today, only a few dozen megabytes can fit into a desktop machine.)

It all seems crazy: to maintain compatibility between the new chips and their less powerful cousins, IBM retained MS-DOS and thereby saddled advanced microcomputers with an obsolescent operating system. The key to unlocking the power of the 286 and 386 processors, then, is a new operating system. The one IBM has chosen is Microsoft's OS/2, which does away with the 640-kilobyte limit and adds many new features, including the ability to run several programs simultaneously.

Right now, applications that run under OS/2 (version 1.0, or OS/2 plain) work the same way as their MS-DOS counterparts, although some OS/2 programs can process more data. The real development of software will come with OS/2 version 1.1, which will incorporate a graphical user interface called Presentation Manager. Written by Microsoft and supported by IBM, this new interface works something like the Apple Macintosh: basic procedures are done the same way in all programs. Whether you are in a word processor or spreadsheet, a database or an accounting program, you can enter and edit text, open and close files, and quit the program with the same commands; moving information from one program to another is easy. All of this represents a major change from MS-DOS programs and means that nearly all MS-DOS programs will have to be thrown out or completely redesigned to take advantage of Presentation Manager.

Faced with the limitations of MS-DOS and the advent of OS/2 with Presentation Manager, the most ambitious software companies have already given up the idea of trying to put their new ideas into MS-DOS programs. New versions of established MS-DOS programs are still appearing, and a few new programs have yet to make their debut, but development of MS-DOS software is already slowing and will probably stop abruptly when Presentation Manager and OS/2 become established in 1989. The MS-DOS software you see in 1988 is pretty much all the MS-DOS software you will ever see.

OS/2 with Presentation Manager will not put an early end to the chaos, however. For one thing, the initial version of OS/2 and the programs written for it are designed for 286 micros; software can occupy no more than 16 megabytes (less the 360 kilobytes reserved for system functions). These programs will run on a 386 micro but will not take advantage of a 386 processor except to run faster. For another thing, the 386 is technically much better than the 286, quite aside from its memory-addressing capacity -- so much better that many software developers wish that all 286 micros would disappear. Developers know that their best ideas really need a 386, not the awkwardly designed 286. So OS/2 and its software will split into two forms -- simpler products for the 286 and, later, more sophisticated ones for the 386. The large number of 286 micros now in use will retard development of good software needing a 386.

And the even larger number of 8088 and 8086 micros sitting on desks will remain in use for years. As a result, many companies will be struggling with a jumble of three coexisting processors and operating systems, none of which are fully compatible. Programs and files will no longer be readily transferable between systems, and people trained on one system will have difficulty with a second. The disorder is starting to paralyze software developers, who do not know which combination of hardware and operating system to put their major efforts into.

So what can you do? The best strategy might be to start with or switch to another kind of microcomputer. The most serious contenders are the Apple Macintoshes.

The Motorola CPU chips used in the Macintosh microcomputers do not suffer from the incompatibilities that plague Intel's CPUs, which means that the same software can run on any existing Macintosh model. The operating system and user interface is uniform across models and across applications, and switching from an early version of the operating system to a later one requires no relearning. Overall, software for the Macintosh is about two years ahead of software for the IBM PC. Indeed, it is precisely the advantages of Mac software that IBM and Microsoft are attempting to mimic with Presentation Manager in OS/2. These days, many developers of advanced software are writing programs for the Mac instead of PCs, planning to produce IBM versions when the dust settles. For example, ON Technology, a new company started by Lotus Development Corp. founder Mitch Kapor, is developing its product for the Macintosh first.

The main disadvantages of the Macintosh are its price and Apple's status as the sole supplier of hardware. Macintoshes are expensive, and Apple produces only a few models. Another drawback is the lack of a satisfactory portable Mac, and when Apple finally produces one, it too will undoubtedly be expensive.

As for other systems, they lack the range of software available for the IBM PC and the Macintosh. Someday UNIX could become a competitive operating system in the business marketplace -- although UNIX machines and software suffer from vastly greater incompatibilities than the IBM PC -- but for at least the next few years, the software choices will be limited and very expensive.

If sticking with IBM is your only alternative, the following plan should help minimize the confusion of the next five years:

* * *

* 1988. If you need to buy hardware, don't buy any microcomputer based on the 8088 or 8086 processors; they have no future. At current prices, it's a good idea to buy micros built around the 286, which are selling for only a few hundred dollars more than basic PCs. Overall, it will probably cost you less to buy an AT-like machine now and trade it up for a 386 machine later than it costs to buy a 386 machine now. But because even 286-based models will have a limited useful life, go for the less expensive brands. On the other hand, if your business can use the speed of a 386 micro now (as a central computer on a network, for example), stay with the major brands; while OS/2 is still in flux, larger companies are more certain to support their 386 machines. If you absolutely need more memory for MS-DOS software now, you might consider the Lotus-Intel-Microsoft Expanded Memory Specification (EMS). Designed to work on an 8088 CPU, this scheme is at best an awkward stopgap, but if your software supports it, EMS is better than nothing.

On the software side, keep the same MS-DOS word processor, spreadsheet, and database that you are currently using. There is no point changing now to another MS-DOS program, even if it offers better features, because you will soon have to retrain everybody for OS/2 programs.

* * *

* 1989. Most of your microcomputer users will probably continue with MS-DOS through 1989. A key planner or a small group in your company that has a 286 or 386 model should now buy OS/2 with Presentation Manager. This person or group should start looking for OS/2 applications to replace your MS-DOS programs and for utility programs to convert data stored by MS-DOS programs.

If you need more hardware, you should buy some 286 and some 386 machines. The ratio will depend on the price of the 386-based models; if their price has dropped significantly, buy more of them. By mid- to late 1989, you should know if OS/2 with Presentation Manager will run on the less expensive 386-based models.

* * *

* 1990-1991. By now, OS/2 Presentation Manager programs should be appearing in all major software categories. When your planning team has found a good set of OS/2 products, change over from MS-DOS as quickly as you can; avoid running both MS-DOS and OS/2 in your company at the same time. Retire all MS-DOS software and any old hardware that can only run MS-DOS. At this point, 386-based micros may be cheap enough that you need not buy any more 286-based models.

* * *

* 1992-1993. Application software specifically designed for 386-based microcomputers should start appearing at this time. If a 386 program is an enhanced version of an OS/2 program written for a 286 micro, you may be able to run both versions within your company without much difficulty. The very best software may be available only in a 386 version, however, so you will have to make some choices. Presentation Manager should at least minimize the retraining involved in switching from one version to another.

By now, hardware based on the Intel 80486 CPU should be on the market, and it should be more compatible with 386-based hardware than these two chips are with Intel's earlier chips.

Toward the end of this century, we will probably look back at the Macintosh and Presentation Manager as the first period of real stability in the microcomputer market. Then, just about the time we are comfortable with that idea, new technologies based on voice recognition and artificial intelligence will make us want to throw everything out again.

* * *

Cary Lu was the founding managing editor of High Technology magazine; he is now technology editor of Inc.