this spring, David Holub, the president of Whole Earth Networks, a mid-sized San Francisco Internet service provider, challenged one of the handful of companies that now controls the Internet's central arteries. In return, on the last weekend of April, he was fired.
At the time, Holub said he was sacked because of his "strong stance" in a series of negotiations between Whole Earth Networks and UUNet, one of the world's biggest providers of Internet connectivity.
According to Holub, UUNet had, for its own benefit, unilaterally changed the rules by which the two companies exchanged Internet traffic. These rules are technically arcane but hugely important: They determine exactly how your e-mail gets where its going, or how easy it is for you to read the Web.
Holub lost his job because his board of directors refused to back his position. That in itself is nothing shocking -- your typical boardroom dispute, turned ugly. But far more was at stake here than just a simple disagreement over business tactics. As Holub wrote in a posting to the North American Network Operators Group mailing list, the real issue was "the accumulation of all the economic power over the Internet into a very small number of hands."
This seemingly obscure conflict over network traffic-handling was in fact a watershed moment for the Net. Basic questions about the Net's fate are being decided in boardrooms right now: Will the Internet service business remain a competitive market teeming with new players, with the public benefiting from widespread low prices? Or will an oligopoly of big companies start to call the tune, dooming the one thing about the Net the general public loves most -- cheap, unlimited, flat-rate access?
Holub was fired almost exactly two years to the day after the National Science Foundation, which sponsored the Net's early development, completed the privatization of the Internet. Today, five corporations control about 80 percent of the Net's infrastructure, according to Internet analyst Gordon Cook. Some of the corporations have familiar names, like MCI and Sprint. Others are less well-known -- BBN, now owned by GTE; ANS, now owned by America Online; and UUNet, a division of the fast growing Worldcom, the fourth largest long-distance phone company in the U.S. If anything unites these companies, it is their desire to reap greater profits from the business of Internet access.
How did we get here? Not too long ago, the Net had a relatively simple structure. In the United States, in the late '80s and early '90s, an end-user generally logged on to the Net via a local ISP -- usually through a university or corporate employee account. That ISP then passed any locally generated Internet traffic -- e-mail, Usenet newsgroup postings, file transfer requests -- up to a much larger network, a regional entity that typically comprised several states. The regional networks, in turn, exchanged traffic with other regional networks at any number of main interconnection sites on the old NSF "backbone" -- the network of high speed telecommunication lines and routers that carried the bulk of Internet traffic.
Today, the situation is infinitely more complicated: There is no such thing as "the backbone" -- there are many backbones. MCI has a backbone. So does Sprint, and so does UUNet. And more than 4,000 ISPs of all sizes, from one-person living-room operations to global monsters like AOL, intersect with these backbones at numerous points. The end result is a virtually unmappable mass of networked spaghetti that is constantly growing and changing.
The most important interconnection points in this sprawling labyrinth are the "network access points" or NAPs -- the Grand Central Station switching platforms of the Net, nodal points where huge flows of Internet traffic are routed between networks. At any given NAP, telecommunication equipment owned by many different companies is physically "co-located," creating a nexus point for both machinery and digital information. The NAPs aren't the only interconnection points, but they're the fastest and most efficient spots for an Internet service provider to connect to the big Net.
As part of the privatization process, in 1994-95, the NSF awarded contracts to four telecom companies, granting them the right to operate specific NAPs. Sprint took over New York. Washington and San Jose went to a company named MFS DataNet (which merged with UUNet in August 1996, and then was bought by Worldcom in December). Two Baby Bells also got into the act: Pacific Bell in San Francisco and Ameritech in Chicago.
Whole Earth Networks had a deal with UUNet allowing it to connect directly to the San Jose NAP and freely share Internet traffic with UUNet. This practice is called "peering." When two networks peer together, they agree to accept each others load of network traffic and pass it on to the appropriate destination. Free peering has long been a common practice at various key Internet interconnection points, dating back to the early days of the NSF backbone.
The problem, according to UUNet, was that Net economics had changed. (UUNet officials did not return Salon's phone calls, but their press releases are here.) In the old days, the regional networks that exchanged information with each other at the NAPs were usually the same size, which made peering a sensible strategy. But now the Internet service market had differentiated. Big players, the Sprints and UUNets, had built huge backbones and carried significant portions of the Net's traffic. Smaller enterprises , like Whole Earth Networks, hadn't. In UUNet's view, they were no longer peers: Whole Earth was getting access to the entire UUNet network, while UUNet, in return, received access to the comparatively tiny Whole Earth Network.
In March, UUNet informed Holub and a group of other Internet service providers that it would suspend the practice of "free peering" with them within 90 days. From now on, UUNet planned to share traffic only with those ISPs who were as big and as well-connected as it was. If Whole Earth Networks and other small ISPs wanted access to the rest of the Net, they could buy it directly from UUNet or another backbone operator.
Why does this matter? To smaller ISPs like Whole Earth, it's literally the difference between being a "peer" (and potential competitor) to big rivals like UUNet -- or simply being their customer. Under UUNet's new rules, notes Dave Hughes, an ISP operator and vocal Net activist in Colorado Springs, Colo., "You're no longer an independent provider -- you're a downstream provider."
"If you want to relegate yourself to being a customer," says Holub, "instead of deciding that you want to compete with them, well, God bless you. But when only a handful of companies own the market, then there is only the illusion of competition."
Furthermore, says Holub, UUNet also refused to tell him just how big Whole Earth Networks had to be to qualify for peering privileges. And the corporation demanded that any negotiations between individual ISPs and UUNet, Holub says, had to be conducted under non-disclosure agreements that prevented the smaller ISPs from joining forces and sharing information.
Whole Earth Networks was already paying for the right to place its machines at the NAP, Holub points out. It wasn't as if the ISP was trying to rip UUNet off. "No one was looking for a free ride," he says. "We just wanted to know exactly what were the criteria for peering. But they didn't have anything to say. They called it proprietary, confidential information."
Holub decided to fight. In fact, he viewed the situation as so dire that he committed what many self-identified netizens consider digital treason: In an e-mail barrage across cyberspace, Holub called for government intervention.
"I am a proponent of some level of regulation in the Internet as soon as possible," Holub wrote to the Network Operators mailing list. "I urge some minimal level of immediate regulation such that the application of accepted telecommunications law and practices are upheld."
Holub's stance suddenly placed him in a political and ideological minefield. On the Net, the words "regulation" and "Internet" are like matter and anti-matter: Put them together and you hear a loud boom, as libertarians fly in every direction, squawking furiously. The Net's phenomenal growth rate and obvious vigor, we are told, again and again, are best nurtured by unfettered competition with no government interference. Never mind the niggling little fact that the Net as we know it today owes its origin to federal government dollars. That's ancient history. Let the market decide, rings the battle cry, and the devil take the hindmost.
But just what does the phrase "unfettered competition" mean? Do four or five companies make for a competitive market?
To Holub, it means ensuring that all companies, large and small, interested in providing Internet access can compete on a level playing field. But without someone enforcing any rules, the end result of an unrestricted environment threatens to be the exact opposite -- a markedly uneven playing field in which a few large corporations like MCI and Worldcom dictate terms to a host of small fry. Or even worse, an oligopoly of big players cooperating to set rules and rates for the rest of the industry and the public.
So far, it's been easy to dismiss such fears. A cursory look at the status quo suggests that the libertarians have been correct. More ISPs exist now than ever before -- 4,500 and growing. Cheap access abounds.
"Since the NSF got out of this business, the number of users of the Internet has skyrocketed, and prices have come down, which shows you how effective competition is at providing wide access," says Hal Varian, Dean of the School of Information Management and Systems at the University of California at Berkeley.
But what of the future? The Whole Earth Network-UUNet showdown undermines the libertarian vision of a free-market utopia. Those who have the big sticks will use them.
"Libertarians always discount greed, conspiracy, collusion and cartel behavior," says Dave Hughes, the Colorado ISP operator. "Big companies scratch each other's backs in a mutuality of interest and screw everybody else. That's just the way they behave, that's the topology of corporations ... That's the nature of the beast."
"Internet connectivity ONLY through the big five is to be feared because the long term survival of an open and competitive and affordably priced Internet is at stake," wrote telecom analyst Gordon Cook in a posting widely circulated across the Net in the aftermath of Holub's firing. "The national press needs to understand the seriousness of the problem for, with the Internet in the US, which is currently unregulated and therefore subject to the whims of the big players, what room is left for the little guy may disappear very rapidly."
The national press didn't ignore the Holub affair. But they painted the dispute as part of a period of normal and necessary shakeout rather than a sign of consolidation. Forty-five hundred ISPs is too many, argued various economists and pundits -- best to get rid of some.
Several observers even welcomed the Holub incident as evidence that the days of cheap access to the Net were numbered. The flood of competing ISPs had artificially lowered the price of access to the Net beyond any rational economic justification, they argued. Now, finally, more sensible pricing schemes could be deployed, as the shakeout cleaned up the Internet mess.
The phone companies, in particular, would love to see cheap, flat-rate Internet access accounts go the way of the dodo. For months they have been complaining that Internet users who log on for days at a time are clogging up a phone network never designed for such usage patterns. But their concern focuses on the phone lines that are the end points of the network, not the Net's main arteries; it's an entirely separate issue from whether the economics of Internet access are out of whack.
It's not necessarily true that 4,500 hundred ISPs offering cheap Net access can't survive. Given true competition, say veterans of ISP trade, it is still possible to make a profit and offer cheap access.
"Frankly, flat rate is not a money loser," says Matt Dillon, a self-described "technical guru" at Best.com, a medium-sized California ISP. "It reduces the burden on the accounting department, for one thing. We make our money off of value-added services like hosting Web pages."
And it's always possible that new technological breakthroughs will subvert the established paradigm. Dave Hughes, for example, is investing heavily in wireless technology. Berkeley's Hal Varian notes that massive satellite communications systems will also come online in the not-too-distant future. High-speed cable access is still a possibility.
Then again, the same large corporations busy piecing together control of the Net's backbones today can also be expected to buy control of any new technologies that challenge their dominance. In an era of supposed deregulation, the massive restructuring occurring in the telecommunications industry -- the merger of the Baby Bells Nynex and Bell Atlantic; British Telecom's purchase of MCI; Microsoft's stake in cable, satellite broadcasting and Net access (in 1995 it purchased a 13 percent stake in UUNet) -- suggests that if left to themselves, the number of truly competitive players will keep dropping.
"It happened with Big Steel, it happened with Big Railroads," says Hughes. "And now it's happening with the Net."
That's the true threat: not congestion, but consolidation.
Even though David Holub lost his job, he was victorious on at least some levels. On May 12, UUNet published a press release in which it set forth specific criteria for satisfying peering requirements, thus addressing one of Holub's major complaints.
Still, the text of the release can hardly be read as a backing away from plans to raise prices or seek increased market leverage. The ultimate impact of the Holub affair may be to sound a public alarm about the concentration of power over the Net into ever fewer hands.
"The most important thing is to educate people about what's going on," says Holub. "The light of day and public attention seems to have very positive effect on these matters."
Shares