Virtual History of the Internet

"Arguing in My Spare Time," No. 3.01

Arnold Kling

January 1, 2000

May not be redistributed commercially without the author's permission.


Niall Ferguson recently edited a book called "Virtual History: Alternatives and Counterfactuals." It gives alternative histories assuming that the American Revolution never took place, or that England did not support France in the first World War, etc.

In that spirit, this essay looks at three ways in which the development of the Internet might have been affected by alternative scenarios that were plausible at the time.

1970: routing tables are controlled by mainframes

The issue was control vs. overhead. The directors of the computer centers at the five universities that were going to participate in the original ARPANET were trying to decide where to put the crucial routing tables, which directed traffic. One approach was to embody these tables in their mainframe systems; the other approach was to build separate, stand-alone devices to be called routers.

The argument for using stand-alone devices was that it would insulate the computer centers from the overhead costs of running the routing tables. Given budget constraints, this was an attractive option. In addition, there were a few visionaries who argued that having separate routers would make for a more scalable system that someday would allow small computers and other devices to connect. However, this was dismissed as a crackpot view—after all, the personal computer had not yet been invented.

Instead, the argument that won the day was the case for control. As one computer center director put it, "I can’t have some little router machine outside my domain sending instructions into my mainframe. If something is going to be putting demands on the resources of my system, I need to be in charge of it."

As a result, the decision was made to store and process the routing tables of the ARPANET on the participating mainframe computers. The decentralized Internet never developed. The computer center directors controlled the growth of the Net. The original separation between communications and content gradually became compromised, so that the computer center directors ultimately exercised control over content as well.

When the personal computer was developed, some people advocated trying to develop an Internet that would connect these devices, the way that mainframes were connected. However, this idea never took hold. The researchers who used the mainframe Internet were not sufficiently interested in a PC Internet to undertake the effort to resolve all the technical issues. Without the academic community or government support, business could not form a coalition capable of developing and implementing standards and infrastructure needed for a parallel Internet.

1993: AOL Spurns the Web

Late in 1993, America Online faced a dilemma. People still considered it the Number 3 online service, behind Prodigy and CompuServe. Its membership was stuck at around one million. Some executives proposed to use the Internet as a competitive wedge. While all three services were open to Internet email, these executives advocated that AOL provide its users with access to Gopher, with access to the World Wide Web to follow in August of 1995.

Other AOL executives objected. They pointed out that Gopher and the Web had the potential to foster disintermediation that would undermine AOL’s ability to charge businesses for hosting their content on its proprietary service. As one executive put it, "If we let our customers view Gopher and Web sites, then instead of coming to us, corporations are going to put their content on the Internet. We would be handing businesses our customer base without our being able to charge them for it."

This argument was accepted, and AOL decided against offering Internet content to its customer base. Although AOL continued to compete successfully against its rivals by sending out free disks and promoting its popular chat rooms, these rivals, too, decided that it would be suicidal to provide Internet content to their customers. Thus, all three major online services remained proprietary. All were able to achieve profitability, but their membership growth was only about 10 percent per year.

Because of AOL’s decision, and those of its rivals, the demographics of the Internet remained poor from a commercial perspective. As one analyst put it, "What can you sell to a male college student, other than porn? If you have a mainstream product, you’re much better off setting up shop on one of the Big Three."

Lacking a mainstream consumer clientele, the commercial World Wide Web slowly evolved into a Red Light District, consisting of pornography, gambling, pirated software, and other fringe merchandise. Internet usage grew only a little faster than that of the online services, and the Internet’s growth was due mainly to business-to-business commerce applications.

"Your typical family is like a deer caught in the headlights," the analyst explained. "One parent wants CompuServe, for business. The other parent wants Prodigy, for shopping. The younger kids want AOL, for chat. And the older kids want the Internet, for help with schoolwork. The result is deadlock. They don’t sign up for anything."

Microsoft, AT&T, and other companies also offered proprietary services, further muddying the issue. By 1997, there were eleven national online services, none

of which offered access to Internet content. For businesses, the fragmentation of the online market was frustrating for many years.

1995: Netscape stays private

Following a long dispute that divided its board, Netscape Communications chose not to move forward with its initial public offering. "We decided that before we sell stock to the public, we want to be an established, profitable company," a spokesman explained. However, some board members did not agree with this strategy, including John Doerr, who resigned in protest.

Doerr wanted Netscape to use its IPO to raise funds and visibility for an all-out war with Microsoft. Instead, Netscape decided to focus on its core web server business. Even there, Netscape chose to concede the Windows NT market to Microsoft. "We want to build a brand name that stands for reliability," the spokesman said. "You’re never going to be able to do that on NT, so we want to focus our developers on the Unix platform."

Because Netscape took the view that its web server market was the most critical part of its business, it made a decision to force the developers of its own web site to drop their resistance to using the critical components of Netscape’s server technology. Their webmasters became part of a powerful feedback loop that Netscape used to hone the reliability and ease-of-use of its server product.

In 1996, the open-source "Apache" project folded. "Netscape owns the Unix market," one participant lamented. "Businesses want to buy a brand-name product. And Netscape somehow convinced them that having a graphical interface for installation is a big deal."

As the Web exploded and Netscape’s share of the server market continued to increase, revenues grew to $2 billion in 1996 and $4 billion in 1997, when Netscape finally made its long-awaited IPO. The company soon had a market capitalization of over $100 billion.

Meanwhile, other Internet companies that tried to sell stock to the public before achieving profitability were given a chilly reception. "Everyone’s looking at Netscape and saying that it’s smarter to wait," explained one market analyst. "People want to buy the companies that take their time and build their business. They don’t want to throw money at start-ups with no track record of earnings."