Something is happening in our industry, something very important, and we would be wise to recognise its significance and take account of it in our plans. The advent of the IBM PC in 1982 forever changed the nature of the PC software business, even though many software companies didn't realise it at the time. (Autodesk certainly didn't: we spent at least as much effort on CP/M-80 and CP/M-86 machines in the first two years as we did on the IBM). Today, the shape of the industry is again being changed by the emergence of a new standard application platform defined, this time, not by hardware but by software--Microsoft Windows.
Just as there was nothing ``new'' about the IBM PC in 1982; 8086 and 8088 machines with similar capabilities existed years before, there is nothing at all new about Microsoft Windows except the way the market has embraced it. But that's all that really matters in the long run. Whether you're a Macintosh fanatic, a committed NeXT developer, an OpenLook advocate, or a Silicon Graphics believer doesn't change this fact: more than eight million copies of Windows are expected to be sold this year, and that estimate may prove low since rumour has it more than half that number were sold in the first 90 days of 1991.
Whether it's ugly or beautiful, naughty or nice, good or evil, when a product gains that kind of momentum and enlists that many users on its side, software developers had better start paying attention to it. Consider this: foremost among the companies who desperately hoped Windows would never take off was IBM--an outfit known to have some clout in the marketplace and reputed, by those who dislike it, to be able to persuade people to buy almost anything. Yet even in the Fortune 500, the very heart of IBM's market and the segment it influences most strongly, Windows is spreading like wildfire, brushing away OS/2 as if it had never existed.
Unlike the Macintosh, which has suffered from a premium price, single source, and worries about connectivity, Windows allows a DOS user to upgrade for less than $100 and continue to run all his old software. This both contributes to the rapid adoption of Windows and reinforces users' demand for truly integrated applications; once Windows is on your machine, the distinction between programs that understand the clipboard, system fonts, system printer, and all the other Windows services and those that go boinggg!!! and blop out to a dumb old DOS character screen becomes glaringly evident. This makes getting caught out without a Windows version of your program just about the worst possible thing that can happen to a DOS application vendor these days. Just look at the increasingly desperate and strident promotions being unveiled by Lotus to try to maintain sales of 1-2-3 as they feverishly debug their Windows-based reply to Excel.
All of the dynamics that made the Macintosh market so special, that made Macintosh users so unwilling to consider any alternative, are now being ignited by Windows in a market ten times larger. A community of users ten times the size of the Macintosh, Amiga, Sun, and NeXT user base combined is now beginning to discover, albeit in a cruder way, what possessed people to buy those other systems. You don't have to predict the future to see the Windows phenomenon, you need only open your eyes. Consider this: Windows, like the Macintosh, lets you attach a little icon to an application. Users can make their own icons and customise their systems that way. Proud of their artistry, many users upload their spiffiest new icons to CompuServe and other networks so others can share them. The last time I looked there were, sitting on CompuServe, a total of 1.3 megabytes of Windows icons ready to download. These aren't the applications, just the icons--a total of 1700 of them, including five each for AutoCAD and Generic CADD!
A market ten times the size means ten times the money to be made by application vendors, and if that weren't incentive enough, it's a market that, with each installation, displaces a raw DOS machine. Certainly Windows continues to suffer some technical shortcomings: it only allows 16 bit applications (unless heroic effort is exerted, as in the case of Wolfram Research's Mathematica), it is essentially a single-tasking system, and it inherits all the shortcomings of the MS-DOS file system. All of these limitations are, however, scheduled to be remedied over the next two years. Given the importance of Windows to Microsoft's strategy and the resources they commit to such projects, Windows buyers can be reasonably confident the schedule will be met. And even if it isn't, Windows 3.0, as it stands today, is far and away the best environment you can choose without throwing away your DOS hardware, and that's an option most users can't afford, even if they were inclined to.
In addition, there's an effort underway by Microsoft, Compaq, and others to make a machine-independent version of the Windows-OS/2 environment and use it to enter the workstation market. While there's room for many a slip in such glib and grandiose plans, if I were Sun I'd be more than a little worried about the prospect of twenty or thirty companies cranking out RISC machines that ran a user interface already known by twenty or thirty million people, one to which application vendors could port their software to simply by recompiling.
What I'm suggesting is that Windows is a Big Event--the kind of thing that happens every decade or so in our industry that establishes a new baseline from which future evolution builds. Events of this nature reward those who move quickly enough to exploit them and winnow out others whose attention is elsewhere, who underestimate the significance of the change, or cannot react in time. Big Events force those who wish to survive to revisit their strategies and question long-term plans. This process requires flexibility in an organisation which is difficult to maintain after it has grown enormously in size and become set in its ways.
One central and virtually unquestioned tenet of Autodesk's strategy has been ``platform independence.'' That means, with very few and very limited exceptions, we do nothing on any one machine that cannot be done on every machine that runs AutoCAD. This forces us (some would say ``gives us a handy excuse to'') exclude support for many facilities on machines like the Macintosh which virtually all other applications, even the least expensive, furnish. The facilities provided to the AutoCAD user become, in a large sense, limited to the least common denominator of those provided by all the machines we support and look, consequently, crude next to applications closely tailored to a specific system. When we feel compelled to address a glaring shortcoming, such as lack of support of system menus and dialogues, we're forced into a much larger project such as Proteus, since our solution must work on every machine and operating system, not just one.
Severely limiting the integration of AutoCAD with various operating environments hasn't hurt us so far, I believe, primarily because the systems that account for the overwhelming percentage of our sales: DOS and 386 DOS, also happen to be the least common denominator in virtually every aspect. Consider all the things we could have done to make AutoCAD faster and easier to use if we hadn't required everything to run in 640K of RAM. Regardless of your opinion of our Macintosh version of AutoCAD, or your view as to how we might have proceeded in that market, the fact is that regardless of whether we succeeded beyond our wildest expectations or failed to sell a single copy of AutoCAD for the Macintosh, our financial results wouldn't have changed much. That doesn't mean we shouldn't have put AutoCAD on the Macintosh, or that we shouldn't continue to strive to better integrate AutoCAD into that environment (I'm the guy who first managed to get AutoCAD running on the Macintosh, you'll recall); all I'm saying is that Apple's market share is simply too small to have much of an effect on our sales (especially when you only count machines capable of running AutoCAD), and rapid expansion of the market generated by AutoCAD is unlikely as long as the Macintosh suffers in price/performance against 386/486 and Sparc machines.
Windows, however, is changing the rules in the very heart of the AutoCAD market. We will, certainly, ship a Windows version of AutoCAD before too many months pass, and we will upgrade that initial product to take advantage of the new versions of Windows now in the pipeline, but I think we have to revisit the level of support we're planning for Windows. Our Windows product will be integrated with Windows roughly to the extent our Macintosh product conforms to the Macintosh environment, which is to say somewhere between ``somewhat'' and ``moderately.'' Certainly it will be obvious to any user that many rules change when the mouse strays into the AutoCAD window. This situation has almost certainly hurt us in the Macintosh market, but due to the limited size of the market and the fact we didn't have any sales there to begin with, hasn't become a company-wide priority to fix. I believe that a similar failure to comply with ground rules for Windows applications may hurt us severely, and every week that passes without our thinking about how to address this problem adds to the danger.
(I would hope that whatever we do to allow AutoCAD to fit comfortably into Windows will also let us conform as closely on the Macintosh, OpenLook, NeXT, etc. However, if forced to choose between close integration with Windows now and all-platform user interface support in 18 to 24 months, I'd do Windows first and worry about the others afterward.)
Editor: John Walker