The Paradise Tree
(Non-allegorical version for the mythologically challenged)

Copyright 1998 Roger E. Ison. All rights reserved.

The outrage of plaintiff’s attorneys is easily had, and plenty of them are having at Microsoft lately. Microsoft stands accused of unsavory things, as when they imitated another firm’s disk compression software, perhaps in violation of patent, and could hardly plead ignorance. They have used marketing muscle inappropriately in some of their licensing and distribution contracts.

But those obnoxious practices didn’t make Microsoft what it is today. Eliminating them will not unseat Microsoft or rescue Sun Microsystems and Netscape from their accumulated mistakes and misjudgments, nor make consumers and software developers much better off or unleash a new world of suppressed creativity.

Microsoft and Intel were endowed by the disaggregation of IBM’s old, vertically integrated business model, and the emergence of a new model that broadly separates information businesses into infrastructure and adaptive segments. It’s important to understand the technological and economic forces that drove this change, because those same forces will also determine the real effect of any "remedy" now imposed by a court.

Ancient history

Before 1964, each generation of computers from every manufacturer was fundamentally incompatible with previous models. This problem desperately needed solving, and IBM fixed it once and for all with the introduction of System/360. Few people outside IBM now remember that Tom Watson received a Presidential award from Lyndon Johnson for that achievement. The company defined, created and controlled every aspect of their computer family: architecture, memory, semiconductors, disk drives, operating systems, terminals, networking, application software and support.

Every part of this vertically integrated "tree" of technologies was carefully cultivated: the roots were fertilized and irrigated with R&D investments, the branches carefully pruned, the harvest timed and controlled. IBM trucks carried the fruit to market. IBM employees staffed the kiosks at the grocery stores.

This grand vision forced all computer companies to play by a formula: "Our computers are all compatible, and they are neater than brand X". But Big Blue did it first, executed with military precision, marketed superbly, and locked customers into architectural compatibility. Amdahl, Fujitsu and other mainframe cloners bit out pieces of the business, but they were never able to compete from bottom to top of the vertical technology stack.

However, IBM’s was an orchard with only one type of tree, bearing one type of fruit. Some trees were big, some were small, but they were all basically the same, and too bad if you didn’t like them. There were holes in the product line, notably minicomputers, but in retrospect we can see that most other computer families introduced by IBM during this period were essentially unimportant experiments in niche markets.

Why the model broke

This business model required the maker of a computer family to innovate simultaneously at every architectural and technological level, and only IBM, having gotten there first, could afford R&D and marketing on that scale. The Justice Department’s long, flawed, antitrust litigation hobbled the company without fundamentally changing the competitive situation. IBM became more circumspect and developed an absurd internal bureaucracy of lawyers and overseers, but until the industry’s structural model was redefined there was little meaningful change in the competitive reality.

Change did not arrive until advancing technology disaggregated this vertical business model into a horizontal tangle of independent microprocessor, disk storage, memory, system software and application industries. To hold its position, IBM would have had to invest and innovate, at each level of its vertical stack of technologies, as quickly and efficiently as all those independent firms could do in their independent segments. It was simply impossible for a single firm to win against vigorous, simultaneous Darwinian competition at all levels.

The economic definition of an "open system"

In a strange way, Microsoft’s original MS-DOS operating system became an economic success because of its inadequacies. I was told, by a person present at the creation, that this weakness was one of the reasons IBM went "outside" for an PC operating system. Apart from the internal politics, the company was afraid something created by their own teams might be good enough to threaten their broader business.

Yet that technical weakness was a great economic strength in disguise. MS-DOS was a totally "unprotected" operating system. The memory, interrupt vectors, I/O cards and other features of the PC were completely open to any programmer. Driven by some reflex developed during the antitrust lawsuit, IBM even published the source code of its PC BIOS chip, the basic input/output system that controlled the machine.

And this in turn meant that any individual with gumption could independently improve the system, add value, and make a profit – without a by-your-leave from Microsoft or IBM. The Lotus 1-2-3 spreadsheet would have been impossible but for this openness: 1-2-3 had to bypass the IBM BIOS and write directly to the PC’s screen to achieve acceptable performance. Suddenly the world was full of elves writing software!

No doubt the cheapness of PCs and their mass market potential attracted lots of developers. But a "closed" architecture, which is to say a protected, controlled, safe and secure operating system, would never have succeeded on the scale of the PC. IBM, Intel and Microsoft had stumbled into a completely new business model where, at every level of technology, independent players could add value and make an independent profit.

In retrospect it is clear that the true definition of an "open system" is economic, not technological. A system is not "open" because it conforms to standards; it is open when independent players can independently make it better and make a profit, regardless of who establishes and maintains its definition. In this profound sense, the PC operating system defined and controlled by Microsoft was actually more open than UNIX with all its incompatible versions.

With economic openness comes an avalanche effect: the more people improve the platform and add value to it, the more attractive it becomes for further improvements. The platform becomes a sort of magical Paradise Tree: the more it’s harvested, the more fruit, and more varieties, it bears. Where IBM’s orchard grew only plums – big plums, little ones, sweet and sour, but all of them plums – the Paradise Tree is a sprawling, reaching tangle of innovation. All those independent developers are constantly grafting on new kinds of fruit. Things that taste good together are often crossbred into new hybrids that are a bit of both. The most successful branches become permanent parts of the tree.

How Apple and Sun went wrong

Apple, with a similar opportunity, a head start, and a better product than WinTel, failed to dominate because its system was fundamentally closed. The company kept it that way to maintain, as much as possible, a vertically integrated business model with profits derived from selling hardware.

Sun Microsystems went astray much like Apple. Although Sun adopted UNIX, their computer platform was not very open in the deepest, economic sense. Sun insisted on defining its own "SPARC" CPU architecture, and was deeply conflicted about encouraging clones. SPARC clones were technically legal and possible, but Sun made the business reality as limited and unattractive as possible.

All the UNIX vendors suffered from this disease. To them, the operating system was a necessity required to sell hardware, and an opportunity to lock in customers. The OS was an expense subsidized by the hardware. Their pitch echoed IBM’s old line: "Mine’s better than yours. Why don’t we standardize on mine?"

Sun’s arrogance was particularly striking. I once stood beside Bill Joy (a Sun founder and chief technical wizard) in the lunch line at an "Open Software Foundation" meeting. Sun, of course, was "outside" the OSF and opposed to it. Sun wanted the world to standardize on their version of UNIX.

Joy couldn’t be troubled to talk with his neighbors in line. He ignored us and shot Nerf ball baskets in machine-gun succession. I don’t really want to be here, he seemed to be saying, we’re going to out-shoot and outrun you.

The posturing at OSF was incredible, but its founders never lost touch with their vertically integrated business models. At one point OSF had a technology essentially in hand that would have allowed application developers to compile a program once, then run it efficiently on any OSF-compliant computer, regardless of the underlying architecture. Quiet opposition from some of the founders killed it. Terrified of making it easier for customers to switch vendors, these giants of industry completely missed the importance of enriching their application offerings.

Bill Gates went and did it. To Microsoft the operating system is a profit center, so there could never be any question about what to do. Without applications, who would buy MS-DOS or Windows?

It wasn’t genius. Gates was already operating in a disaggregated business model. For Microsoft there had never been any other way.

Innovation and absorption

The computer industry’s disaggregated business model, for software and hardware alike, is defined by two facts. While the first copy of a complex software product or silicon chip is enormously costly, subsequent copies are almost free (except for marketing costs) and fairly easy for competitors to replicate ex post facto. Continuous innovation is therefore the only way possible to stay in business, so hardware and operating systems necessarily absorb inventions at the technological periphery. Otherwise, a company will die as soon as somebody copies its product.

This process of absorption is a very good thing for application developers and end-users. Even with all the work Microsoft puts into compatibility – the company has a truly remarkable record of carrying users forward without breaking installed applications – installing and administering PC software is a daunting task. It would be impossible if all the pieces came from different vendors.

(If you don’t believe me, try installing Windows/NT 4 with networking on an empty hard drive. Then add Internet Explorer 4.0. Only a software-era Renaissance man can get it to come up "clean".)

Within large businesses, continuous integration and absorption is pursued as a cardinal virtue. Think, for example, about corporate databases. Today no IS manager in his right mind would try to run a company using hundreds of independent little applications that must somehow be maintained. A big, uniform relational database that integrates everything is the only sensible answer, which is why SAP is well on the way to becoming the Microsoft of enterprise systems.

So I repeat: it is a Good Thing For Most Of Us that software, the most malleable of mankind’s inventions, constantly absorbs new features. If Microsoft didn’t do it, some other firm would be invented to do so, and that firm – lacking control of the operating system – would not be able to do as good a job, nor do it as cheaply.

Nor is absorption just a software phenomenon. Compare the guts of a new 400 MHz Pentium II computer to a machine five years older. The new box is smaller, its "mother board" almost empty by comparison, with fewer chips doing much more. When the heat constraints are eventually solved, there will only be two or three chips doing everything including video and audio. (Two-chip computer systems based on older versions of the Intel architecture have been available for years.)

Infrastructure, adaptation and the Internet

Modern information systems are assembled from large chunks of relatively independent infrastructure technology. Intel, Oracle and Microsoft are infrastructure providers. The other part of the industry structure is an "adaptive" segment that takes these pieces of infrastructure and efficiently assembles them into products. PCs themselves are products of the adaptive segment. So are enterprise software systems.

This new structure defines a new class of technology investment and invention. The challenge for an adapter is to reduce the cost of dealing with variety in his customer base, making it cheaper and faster to describe and assemble exactly the product each customer needs. The vigorous, completely open market in "re-usable software components" for Visual Basic, Delphi and Java reflects the economic potential of this approach.

Borland (now called Inprise) has deep expertise in this area. Microsoft’s Visual Basic for Applications (VBA), Object Linking and Embedding (OLE) and related facilities are all innovations that make it easier to assemble sophisticated applications from large infrastructure components. Relatively unsophisticated users can construct complete, customized office information systems using only VBA, OLE, a spreadsheet, and a database. Sophisticates regularly build stuff no MIS department could have produced five years ago.

The Internet itself is the world’s largest, most economically open infrastructure, and web applications may be the ultimate adaptive segment. Java is an important enabling technology for the Internet vision, but an imperfect and incomplete one, and obviously not the only enabler that will be needed. For example, I've been using a pre-release version of Windows 98, and it’s a decent product. I also like and use a web browser called Opera. But I would not want Windows 98 without Microsoft’s Explorer browser, nor will many other users.

Why? When I want to update my copy of Windows 98, I just click a few buttons. IE fires up and goes to a special web site where the state of my operating system is examined, and any necessary new patches and features can be downloaded and installed. The savings in time and effort are enormous. Microsoft didn’t invent this idea, but I’m sure glad they put it in Win98. That one feature, by itself, justifies moving to Windows 98. Most customers won’t want to be without it, and every large corporation will use some version of this technology to manage the software on all their PCs.

This update function can't be performed in "pure" Java. Java is a "secure" environment, carefully isolated from the rest of the computer system so a Java terrorist can't destroy your files or otherwise inconvenience you. It certainly can't revise your operating system. It could download a new Java operating environment, but that environment can't run your current Windows apps, nor will it ever be able to.

Could Microsoft do the same thing without IE? Sure. But they’d need a piece of software with a user interface, networking, and many other capabilities already present in Internet Explorer. It would be stupid to turn a straightforward piece of engineering into a more complex, unnecessarily fragmented design, as the Department of Justice now demands.

How many do we really need?

Sometimes we can gain insight into hard problems by asking what would happen if some parameter or condition were pushed to an extreme. So I ask: how many operating systems does the world really need? There will always be niche requirements and products to satisfy them, but what about the great public river?

Right now the economic answer seems to be two: one for people, one for giant servers. The ultimate answer may well be just one, at least until heavy-duty multiprocessor applications become important.

This would not be the answer if software and hardware lacked the capacity to absorb innovations; but that would be a different world. In our world, innovations absorbed into the infrastructure are invariably more useful, more valuable and more marketable than they could ever have been standing alone.

Then more people use, expand and further decorate the infrastructure, making it even more valuable. Economists call this avalanche effect an "increasing returns regime", contrasting it with the more traditional "decreasing returns" on most manufactured goods, where competition drives profitability down to an equilibrium value.

The quiet-spoken Stanford economist Brian Arthur has popularized the notion of increasing-returns, and reports suggest his theories carry great weight at the Justice Department. Arthur says that because of increasing returns, one technology may become so dominant and established that better alternatives cannot attract enough adopters for innovations to succeed. If that doesn’t justify an antitrust lawsuit, what does?

Obviously somebody in the Star Trek scripting team read Brian Arthur. "We are the Borg. You will be assimilated. Resistance is futile."

It’s an elegant theory that goes far toward explaining the exponential growth of personal computing. But to be complete, the theory must also talk about what happens when innovation ceases. How long would Microsoft survive it if stopped improving its applications and operating systems? Maybe five years. The theory also must explain why the prices of operating systems and applications have steadily declined, both in absolute terms and in relation to their growing capabilities and feature sets. Arthur's work did not address these issues when I last examined it, in about 1990.

Still, his theory is valuable and important as far as it goes. Increasing returns from the information infrastructure are a vital and underappreciated factor powering America’s economy today. Many of the best, recent innovations make it easier and cheaper to absorb and adapt new technology. Many software inventions could never sustain a profitable existence outside the sheltering environment of that infrastructure.

And all that Java

Even if the unexpected happened – some world-changing computer innovation that cannot be absorbed into the world’s most widely used operating system – I rest easy in the confidence that gumption, intelligence and free markets will find ways to make it cheap, widely available, and fun.

Certainly Java is not that earth-changing innovation. Most people would be much the worse off if Windows were overturned and replaced by a Java based operating system. The cost of change would be enormous, the benefits negligible. After all, we already have Java in Windows. I can run Java apps and Windows apps. I can even use them together.

The market has already done the right thing!

Java without Windows? Give me a break! Why give up all the capabilities that will have to be replaced if Java assumed all the technical and economic functions of Windows? That blunder is why NetPC’s were non-starters.

For all these reasons, I have no sympathy for whining about Microsoft. Sun misunderstood and misplayed the game, and now wants Joel Klein to bail them out. So too Novell, so too Netscape. The scary thing is that Justice seems to think it can, and doesn’t care about the impact on the rest of us. The Macintosh experience suggests that not many application developers will make products for more than one PC operating system.

Give a lawyer a law and he’ll find a way to use it. Joel Klein wants to cut down the Paradise Tree.

But Microsoft is, truly, a creation of technological and economic forces. Another Paradise Tree would spring up to replace it, different but not better than the original. And untold millions of dollars – perhaps even billions, when we count the impact on users – will have been wasted by an ambitious lawyer making Government the agent of competitors who blew enormous opportunities through arrogance and incomprehension.

And do I trust Klein more than I trust Bill Gates? Gates got rich by making it possible for others to get rich adding value to his creations. He understands how this business works. Joel Klein may have the best intentions, but what thing of value has he ever created?

Anyway, life isn't all about infrastructure. The Internet is really about content, and there Microsoft has no particular advantage. The Microsoft Network is just another Internet service provider, not the biggest nor the best. The company's other Internet enterprises have been disappointing too. If any succeed, it won't be because the company controls PC operating systems. It will be due to finding a way forward through dogged persistence.

Nor is there any reason to fear Microsoft will set up some kind of Internet toll booth or file the milling off every electronic coin that changes hands. Electronic commerce applications are built using common infrastructure facilities, but they are just that: applications. There is no possible way for Microsoft to prevent you using any E-commerce service you wish, any more than they could prevent you from running the Netscape browser, WordPerfect, or Quicken.

Which leaves just one problem

But where do we draw the line between absorbing new technology, and stealing stuff? This is a very tough problem, and hard cases make bad law. As I said at the outset, Microsoft appears to have done some unsavory, muscular things. If the reports are accurate, some of those tactics are not things I would have done in the same situation. I would not want to be on the receiving end of that stuff.

But is innovation being suppressed, because nobody can make money, because Microsoft steals and perverts all the great ideas? The sequential failures of Netscape certainly don’t prove the case, because the company had an inborn flaw: it had little viable, proprietary intellectual property and no realistic business plan.

Browsers were already publicly available. Netscape’s and Microsoft’s both derive from the predecessor called Mosaic. This fact put Netscape head-to-head against a well-financed Microsoft team with years of experience in the art of making barely adequate products get steadily better. Realistic business plans take competitors into account, including the size of the necessary investments. Netscape was launched by giving away its browser.

Browsers aren’t even very hard to write, although a lot of detailed work and testing goes into a good one. I can build a simple browser in a very short time using Borland’s Delphi software development environment and a few over-the-counter software components. Opera, a very nice if yet unfinished browser, was built from scratch by a small team in Norway. I happily paid the Opera guys $35 for their work, and I’ll do it again when they improve it sufficiently.

Might things have come out differently if Netscape hadn’t gotten caught up in Sun’s grandiose dream of replacing Windows with a Java based operating system? The strongest market and economic approach would have positioned Netscape to complement and enhance Windows. Who knows, Microsoft might have bought them out at a handsome price.

My point is that intellectual property ownership is absolutely vital when a small innovator wants to extract value from an invention in the face of giant, cold-hearted infrastructure firms. And the rules governing software intellectual property are a mess, even worse than antitrust law.

But that's a topic for a different Thinking Allowed.