(NB: I am moving some years-old writings into Due Diligence for future reference. This piece originally appeared in early 1997 on Howard Rheingold's old Electric Minds site. Little or no effort has been expended on cleaning up links broken in the intervening decade.)
The Way It Was Supposed to BeA Microsoft strategy statement in the early nineties could have gone something like this:
"We have achieved dominance in desktop operating systems, and a large share in the associated programming tools and user applications. We need to keep growing the business, and to anticipate threats to our current market. Moore's law shows that 'desktop' machinery will soon have the power necessary to fill the roles of workstations and servers. The dominant operating system in these markets is UNIX, and the network capability implied by servers is supplied by TCP/IP. Collision is inevitable. We need a platform that we control, which has similar or superior capabilities, and is leveraged from our current competitive position."
The product plan that emerged might have read like this:
"We will use TCP/IP as our networking layer as well. It allows us to partially co-opt the existing workstation market, and also commoditizes a platform layer in which our competitor, Novell, has a strong position in LANs. Our primary franchise is in operating systems. We need a multitasking version of Windows that's robust enough to run a server, but close enough to the desktop version to exploit the existing base of programming expertise and user experience. Call that Windows NT. Since another of our strengths is in programming languages and tools, we'll construct a toolbox layer that lets developers easily modularize and mix application components -- 'objects' -- in a user interface, first on a single machine, and later distributed among desktops, servers, and workstations. We'll call that OLE, and later on Distributed OLE. With our Office application suite as the leading exemplar of OLE technology, and our superior installed base and volumes, we will quickly dominate the workstation application market as well. We'll know we've won when the network and server platforms are defined at the application layer in terms of Microsoft APIs."
The plan outlined was a combination of accretion (adding modular object-like elements to Windows through OLE) and a change of the conversation in scope, from desktops only to network-integrated elements. This seemed straightforward enough ...
The Boys from Chambana
The first moves came off according to plan. OLE reached market and its second version in April, 1993 attracted substantial support (and was later renamed OCX). A version of NT came out to complement Windows 3.1 in July, 1993. Then things got strange.
In trying to commoditize the network layer, primarily in LANs, Microsoft had adopted TCP/IP. That is also the protocol used in the Internet, where many preexisting networked applications were layered over TCP/IP; informal programming groups maintained and extended them on the UNIX platform. Without meaning to, Microsoft had invited them all to the Windows party.
The Internet grew by incrementalism. Grand visions of programming platforms like CORBA sputtered along for years, while just-good-enough hacks like Gopher took off like wildfire. Having no central brain trust, the net might be said to have naturally evolved accretion as a survival strategy. Multiple attempts at adding functionality ran in parallel, and the winner was acclaimed according to whose work got used.
The Web protocols were designed by Swiss physicists, originally for use by other physicists, and have that "just barely there" flavor. HTTP is a very limited stateless protocol. Its biggest win was in creating a unifying syntax for several other protocols -- ftp, news, gopher -- and thereby co-opting their installed server base. HTML is a cutdown version of a text markup language called SGML, simplified enough that documents could be turned out in a plain-text editor, but with enough expressive power that the texts would have some uniformity across various display devices.
Technically, this was very ugly. The Web was like suggesting that a word processor should be the basis for a global operating system. But the physicists, and the other scientists who noticed the early Web, didn't care about elegance; they started writing a lot of texts in HTML on UNIX. As it spread in the early nineties, the Web came to the attention of the National Center on Supercomputer Applications (NCSA). Under the rationale that Web clients that ran on the most popular desktop machines would further expand the Internet's value to researchers, NCSA funded the creation of Mosaic, the first graphical Web browser, on both Windows and the Mac, along with a set of free server software.
Mosaic simultaneously expanded HTML's utility and made it independent of the underlying operating systems. Web clients and servers could interact regardless of whether they were based on Windows, UNIX, or Macintosh. In effect, NCSA's move had hollowed out the nascent Web platform, which could now leverage the entire base of TCP/IP Internet-connected machines, worldwide. Web content exploded into exponential growth.
The Web platform had achieved positive returns; its commercial potential was obvious, but no one was in the driver's seat. NCSA sold some commercial licenses to its technology, to small companies like EIT, Spry, and a startup called Mosaic Communications. Soon the entrepreneurs came back and hired away the programmers who had created the original Mosaic, and NCSA stepped out of the picture, assigning its rights to Spyglass. [Update 2003: The record is fading quickly, it's getting hard to find stable URLs for these early days.]
Mosaic Communications was backed by former Silicon Graphics executive Jim Clark, and some serious Valley venture money. After a squabble with NCSA over names, it was rechristened Netscape. Netscape quickly made two significant moves. In re-engineering its own graphical Web client, it unilaterally added new features to HTML and HTTP, adding important
capabilities such as tables and security, but also moving to become the de facto controller of parts of the standard.
Second, Netscape made the stunning move of giving away its first product, simply putting it up on the Internet for free download. Soon 80 percent of the Web browsers in circulation were some version of Netscape. This brilliant strategic insight drove the Web further into positive returns, caused the Netscape extensions to become part of the platform, and brought the speed of electronic distribution to bear against competitors who were still using physical channels. Netscape was starting to change
the conversation on Microsoft: it was creating a layer that was closer to the user than the operating system, and its scope was the entire Internet, not just the desktop.
Netscape made sure that some of its extensions to the Web browser, such as SSL, could best be supported with its own servers. It then used its share dominance in browsers to aggressively sell its servers and these features against the largely freeware competition. It also exacted revenues from commercial enterprises that wanted to distribute the Netscape browser, or use it on corporate LANs, and later began to charge per-copy fees to end users.
Netscape created plug-in browser interfaces that allowed addition of media types such as animation and sound, and also server interfaces which allowed webmasters to add capabilities through Perl scripts or
compiled code. What had started as a glorified text display interface was adding enough APIs that it began to look like a programming toolkit, one with no allegiance to Windows or any other operating system. Jim Clark began to call Netscape a platform in public forums, perhaps a little too loudly. Microsoft woke up.
Turning the Battleship
Microsoft was bearing down, focused on getting Windows 95 out the door, putting the last nail in the Mac's coffin. The Microsoft Network [Update 2006: NOT today's version] a consumer online service based on Microsoft's distributed technologies, was getting underway. For quite a while, it seems, Gates and his top lieutenants simply didn't believe that a freeware-based, lame-featured thing like the Web should be taken seriously. By the time they responded, it and Netscape had destroyed the MSN business plan and were threatening Microsoft's planned network platform.
Regardless of what one thinks of Microsoft, the agility of the response when it came has to be admired. In a similar situation, IBM had wasted its advantages over years. Microsoft killed off several major projects to liberate over 800 staffers, by some counts, in a space of weeks, reached outside to hire key people and buy technology components, and rewrote the MSN business plan and the architecture of the
next Windows release on the fly. Microsoft's turn will become a Harvard Business School case (along with the Netscape move).
By late fall 1995, Microsoft had the first version of Internet Explorer out the door, as a free download, and had the next two versions in simultanous production. When Internet Explorer 3.0 was released, Microsoft had arguably erased Netscape's early advantage in browsers. Internet standards groups were changed from collegial old boys' clubs into slugfests between Microsoft and the loose alliance around
Netscape. OLE/OCX had been renamed ActiveX and superimposed on the browser as a means of extensibility. Microsoft was getting ready to "migrate the browser into the operating system," accreting the breakaway functionality back into its own platform.
OOP OOP a DOOP
At Sun, the language project called Oak had been in the corporate wilderness for several years. It was an attempt to combine the advantages of object-oriented programming (OOP) with the cross-platform portability and compactness of interpreted code. For a while, Oak was part of an interactive TV venture, First Person, which Sun hoped to spin off. First
Person was stillborn along with the ITV market, but Sun salvaged and kept Oak, without a clear market in mind.
Sun was beginning to sell a lot of boxes to support the Web, but they were mostly running other companies' software. Attempts to influence the platform struggle through formal channels, such as the Object Management Group's CORBA, were clearly insufficient to match the pace being set by Microsoft and Netscape. Perhaps to address these problems, Oak was plucked from obscurity, renamed Java, and released to the Internet and licensed to Netscape. When Java was used to construct "applets" which could run within the browser, it became a head-to-head competitor with ActiveX. Microsoft also took a license to Java, wrote its own compiler, and began trying to position it as just another language within the Microsoft ActiveX programming framework.
Java's claims to cross-platform advantage on the desktop were limited: it didn't run on Windows 3.1, and the Macintosh battle was nearly over. But it did have a good deal of value to those wanting to hedge the UNIX-vs.-NT battle erupting on servers. And since it didn't need a fullscale PC operating system to support it, Java could serve as a vehicle for lower-cost computing devices for home or office. These would have the dual benefit (from Sun's point of view) of needing servers to support them, and not having any Microsoft systems in the platform. The network computer was born. [Update 2003: And died soon after. Most 'thin clients' are browsers running on PCs.]
Sun spun off Javasoft as a licensing entity for Java, chartered to drive it into new markets. The NC and Java picked up a collection of fellow travelers, each with an interest in seeing Microsoft stalemated:
IBM defending its server (mainframe) business, Apple looking for a new story, Oracle seeking markets for its media servers, Netscape (Navio) and everyone else looking for market opportunities in embedded devices.
In early April 1997, the JavaOne conference saw the rollout of a large set of APIs for adding functionality to Java. While this is tending to bloat Java's formerly sleek definition, subsets are being defined to run on sub-PC machinery. The intended effect of all this is to create a platform, at the programming-language and application level, that owes no allegiance to operating systems or implementing hardware. Java attempts to render Wintel irrelevant.
With Microsoft's inclusion of Internet technologies in Windows, and the Java alliance's addition of comprehensive APIs, the battle has turned into a head-to-head confrontation. The prize is now the platform standard for distributed object-oriented programming; the scope potentially includes everything from TV settops through PCs to arrays of transaction-processing servers, connected by a global TCP/IP Internet.
Looking Back, Looking Forward
This series of articles began as an inquiry into the practical economics of platform wars, and particularly as an investigation of Microsoft's role and whether any intervention in the struggle would be justified. Other than as an entertaining soap opera, my reason for elaborating this twenty-year history is to provide a background for my own conclusions, which I now offer:
Microsoft has won [Update 2006: Not so fast....], for the most part, due to its patience and the blunders
of its opponents. Very few companies would have had the doggedness or strategic insight to sustain ten years of dead losses on Windows before it began to pay. This is in marked contrast to many of its competitors which pursued quarterly results or "strategies du jour."
I find little to choose between Microsoft's attempts to exert the leverage of its market share and those of its current or erstwhile competitors. Having negotiated with both companies, I can say that Netscape could be just as tough in taking advantage of the lead position in browser share as Microsoft has been in exploiting its O/S share. There is no rationale for denying the tactic to either competitor.
True head-to-head competition is the exception rather than the rule in platform warfare, and it's only in this competitive mode that increasing-returns economics clearly applies. Transformative moves such as accretion, hollowing out, or conversation change are necessities for both defense of a position or attack of an incumbent. These are all strategic mechanisms to change the effective market definition of share, by juxtaposing differing visions of specificity reduction through platform definition. Denying such
moves to an incumbent, or forcing it to jump through regulatory hoops in order to use them, is tantamount to a death sentence, given the speed of the market. There is no precedent or justification for such a proceeding when the game is still in play.
Each of Microsoft's current competitors, the Rebel Alliance of Netscape, Sun, Oracle, IBM and Apple, is basically in a defensive posture, dealing with the encroachment of the Wintel PC into its core franchise. In this situation, attempts by each to expand its scope independently are likely to result in disaster; instead, each must contribute to the "Java story" playing from its own core strength. In short, the Alliance needs to stop whining and forge an integrated technical and market strategy that is credibly and steadfastly maintained. Reliance on extended legal arguments rather than on delivering value to customers will leave the allies where Apple ended up in the GUI war, while Microsoft gathers the fruits of its persistence. [Update 2003: Linux has emerged as a rallying point for the Rebel Alliance
and many others. The fight in server-land is now evenly matched.]
For Microsoft, on the other hand, an offensive position is imperative. The PC marketplace has slowed; Microsoft has about ten billion dollars for investment, and the stock market has anticipated its continued growth. Microsoft bears part of the responsibility for stagnation in PC penetration, due to the complexity of its platform, and needs to clean up its act. But it has no option but to expand into auxilliary markets, such as servers, transaction processors, and consumer devices and services, if it's to make its numbers. Formerly, Microsoft had years in which to make its initial mistakes and find its way to success. Now, it will have to fight multiple battles simultaneously against competitors with greater experience. Faster victories will be needed to sustain growth. As it moves further from its core expertise, Microsoft will need to maintain agility and insight in a larger and larger organization, if it's to avoid being drawn onto a killing ground itself. [Update 2003: The jury is still out. Microsoft is pouring billions into Xbox, tablets, and forays into telematics and portable phones. ROI is not yet clear.]
[Update 2006: Jury in. Microsoft has played too slowly, and with too much attention to incremental tactical advantage, at a time where the very definition of platform and the means of competition have been shifting. Who would have believed, when this original story was written, that MSFT would find itself in a three-way fight with Yahoo! (which did exist then) and something called Google, both of which give away most of their end-user products and services. In the next weeks and posts, I'm going to be revisiting the War for the Web (2.0 of course).]