(NB: I am moving a few 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.)
What's a Platform?
Let's try this first without any computers involved: If you live in North
America, take a minute and examine the nearest electrical appliance. Somewhere
on or by the power cord, you'll find printed or stamped something like the
following:
The plug itself will have two parallel metal blades, and may have a third cylindrical prong set midway and below the two. You know you can plug this into any wall outlet in North America, and it will work. The power you use there might have been generated by the local utility, or it may have been 'wheeled' across interconnected transmission lines from a generator several states away, owned by another company.
Within living memory, it's always been this way. But a century ago, there was a Power War to establish the standard for electricity transmission, just as virulent as today's struggle for definition of computer and Internet standards. The Microsoft and Netscape of that war were the General Electric Company, owner of Thomas Edison's patents on the direct current (DC) scheme of power generation and transmission, and Westinghouse, purveyor of Nikola Tesla's competing alternating current (AC) method. As you can tell from the logo you found, Westinghouse ultimately won that war: VAC stands for Volts, Alternating Current; Hz is Hertz, the frequency of alternation. It took a while for the connector specifications to stabilize (that third prong was added for safety reasons in the '60s) but the power transmission system we take for granted in the US, Canada and Mexico was determined by a platform war now little known beyond historians of technology, but which presages today's struggle.
With that as an example, here's a stab at defining 'platform':
A standardized specification of a service, allowing its providers and users and their products to interoperate without special arrangement.
And a few more examples:
- Standardized freight containers
- The Touchtone dialing system
- Railroad gauges
- SKU bar codes
- NTSC video encoding
Why Care?
The Internet and its constituent standards are a platform. That's why I can provide a service (this article) and you can read it, without any special arrangement. Even partial control over the definition of a platform can be of immense value to its purveyor: You'll notice that Westinghouse and General Electric are both still multibillion dollar concerns, 100 years after the Power Wars.Microsoft, Netscape, Sun and myriad others understand this, of course, and are competing frantically to grab effective control of parts of the definition of 'Internet.' Making the contest even more interesting is a theory called increasing returns economics' espoused by Prof. Brian Arthur, of Stanford University and The Santa Fe Institute.
Traditional economics has held that returns diminish with increased production, as (for instance) the pool of high quality resources is inherently limited. Each further unit of farmland put under the plow, or metal ore mined, is likely to be of diminishing quality. Arthur points out that this is often not the case with technological goods. The big investment comes up front, whether in designing the Pentium and its fab plant, or coding a browser. After that, each copy is cheap to make, and further amortizes the initial costs, a pattern of increasing returns.
Compounding this are 'network effects': a positive feedback cycle in which use of the product is likely to lead to further use. You may remember when a fax number on a business card was a rarity. Then suddenly they were cheap, everywhere, and it was a faux pas not to have one on the card. Sound familiar? Got your e-mail and home page addresses on there yet? That's a network effect.
Arthur's next step of reasoning is where things get really interesting. He claims that this interaction of increasing returns and network effects makes the definition and control of technical standards into a winner-take-all contest. Moreover, the winner of the contest is determined as much by chance events at the start of the battle as it is by the ultimate 'merit' of the winning solution. The technical standards in question can create a platform, by our definition. Arthur is in fact proposing that platform wars obey the mathematics of chaos theory, an approach in vogue at the Santa Fe Institute.
When taken into the legal domain, Arthur's theory amounted to a statement that a monopoly on a valuable platform can exist without regards to technical merit (and social benefit), due to the inherent leverage of the situation. Suddenly, Netscape became very interested in Brian Arthur. Its lead antitrust attorney, Gary Reback, brought Arthur's theory to the attention of the Justice Department, then investigating the proposed Microsoft/Intuit merger. Arthur himself encouraged the hostile application of his theory to Microsoft, saying "We end up with bad, cheap products that are priced higher than they should be." The gist of this argument was then picked up by other industry figures hostile to Microsoft.
Does the argument hold water? The remainder of this column, and the next two, are my attempt to examine the Arthur theory as applied to the Internet, in light of the architecture and evolution of computer systems, the theory of transaction costs, and the actual history of the struggle for computer platform dominance since the '70s. The subjects of inquiry are:
- How do a platform and its provider(s) create value for the market and society?
- Has the computer and network platform war been a game of chance, or of skill and strategy?
- What are the strategies and tactics used to conduct a platform war in the computing domain?
- Are software architectures sufficiently stable for chaos theory, and Arthur's analysis, to apply?
The Value of Platforms
Let's repeat that definition of platform:A standardized specification of a service, allowing its providers and users and their products to interoperate without special arrangement.
The first benefit of a platform is then to reduce specificity of goods and services. From my previous column, reducing specificity drops the overall costs of conducting transactions, in particular by removing some of the costs of matching up goods with buyers. Given the electrical power example, imagine the extra engineering, distribution, and sales costs if every US state had its own voltage and Hertz standard. (If you're reading this in Europe, you don't have to imagine it, you get to live it!) So if nothing else, a platform will have the beneficial effect of reducing the overhead on every transaction, with a great saving to the market and society. Note that this specificity benefit increases as the concentration of the market on a single platform among competitors grows, a support for both the reality and the social benefit obtained from Brian Arthur's network effects.
Transaction costs also influence the organization of economic effort. Value exchanges with high transaction costs tend to be conducted in hierarchies, generally firms with vertically integrated product lines. Lower transaction costs allow the exchange to be conducted in a market setting instead, with competition among suppliers. When markets can form, they tend to be more efficient at delivering a total solution than vertical integration. Hence, the defining of a platform can lead to the formation of a market around that standard where there was none before, again leading to an overall saving. This may sound very academic, but it's also very real: The emergence of the Web and HTML as an authoring platform for online content cut right through the vertically integrated products and business plans of Prodigy, CompuServe, and MSN, leaving their proprietary authoring tools at a tremendous competitive disadvantage to the market leverage of the Web.
Finally, a platform allows both its provider(s) and users to focus their efforts on products which obey the specification, and improve their efficiencies of production and use. Particularly in technological endeavors, decreases in costs of production are modelled by 'learning curves,' which are driven by cumulative units of production. When a dominant platform allows concentration of production experience on its peculiar technologies, these will tend to move along a learning curve faster, increasing the lead over a trailing platform. The best known learning curve is "Moore's Law" which predicts the doubling 'generation' time of semiconductor capacity. In the early '90s, it was known in the PC business that Intel, provider of the chips for the PC/Windows platform, had moved about a half generation ahead of Motorola, provider for the trailing Macintosh, at least partly due to superior volumes and hence learning and ability to generate capital for product improvement. (This gap was eventually erased by the IBM/Motorola PowerPC deal.) Again, we find that winners tend to stay ahead.
Transaction costs analysis then tends to support Arthur's thesis of network effects, from multiple points of view. However, it also supplies evidence for great economic benefits obtained by the definition of a platform, a point given short shrift in his analysis. From what we've seen so far, Bill Gates might be both a monopolist and a social benefactor.
Platform Business Models
There are a number of different ways to base a business on creating a platform. Here are some that will be significant when we start looking at the computer and Internet wars, in a rough order from most to least integrated:- Sell everything needed to operate the platform, including its technical definition, as a single service package. The original Bell Telephone built the phones, strung the lines, built and ran the switchboards, and sold the service.
- Sell components that embody the platform, but license and partner with others to create whole solutions. Often, this involves the creation of a 'downward' interface from the platform into which these partners or licensees must fit. MS-DOS and Windows are in this model. This is one of the competing definitions of 'open.'
- License the platform specification, let others do all the production. This is the Dolby Labs model.
- Give away the platform specification, be a leading provider of supporting products. Cisco and Sun partly follow this model. This is the other competing definition of 'open.'
- Give away everything. This is freeware, and not a business model per se, but has been historically important on the Internet. It does create social benefit, but has problems in competing with any model which has a positive feedback from market share into funding. [Update 2003: In the form of give away products, charge for support and services, this becoming a viable business model.]
