December 1998

Regulating The Internet

BY BROUGH TURNER

The Internet under control - the very idea is so troubling that it inspires, directly and indirectly, an abundance, even an excess, of political debate. For example, arguments continue to rage over who assigns domain names, those groups of letters, like nmss.com, that act as addresses for our e-mail and locations on the Web.

Another control issue (and enduring source of controversy) is the fear of consolidation. Earlier this year, when Worldcom proposed to acquire MCI, many people protested that the combined company would be able to dominate and control the Internet backbone. And indeed, the Worldcom-MCI merger was blocked until the principals agreed to sell MCI's Internet backbone to Cable and Wireless.

Personally, I'm not worried that anyone will be able to gain "control" of the Internet. Consequently, I believe efforts to prevent such control are unnecessary. I've even avoided commenting on control issues - until now, that is.

I started to doubt the wisdom of withholding comment when I read the October issue of Boardwatch magazine. In this issue, Jack Rickard, whose views I've always respected, wrote an editorial calling for FCC regulation of "peering arrangements." Whoa Jack, bad idea! - and from a guy who really should know better.

I wondered how many people might share Jack's opinion. So, I asked around, and what I learned surprised me. I discovered an amazing lack of knowledge of the real issues, even among high-tech people. I decided that some information - and some comment on why we needn't worry about too little regulation - might be in order.

PEERING RECONSIDERED
In just a few years, the Internet has evolved from a government-sponsored, academic-run network to the commercial Internet of today. If you're not familiar with how today's Internet works, briefly, it's an interconnection of many networks - networks owned both by private corporations and by public service providers. (See also "On The Horizon," April 1998.)

Interconnection is hierarchical. At the top of the hierarchy are a few very major Internet backbone providers - companies like Worldcom (UUNET and ANS), GTE (formerly BBN), and AT&T (the former Cerfnet). Typically, these top-level backbone providers exchange traffic between each other in peering relationships where the traffic in each direction is roughly symmetric and no one charges anybody anything - at least that's the theory.

Originally, top-level backbone providers exchanged traffic at 10 or 11 public exchange points called NAPs (Network Access Points), MAEs (Metropolitan Area Ethernets), or FIXs (Federal Internet eXchanges). As the Internet grew, the number of backbone carriers exploded (there are over 40 in the U.S. today). And, as traffic grew, the public exchange points saturated, so major backbone providers began entering into private peering agreements.

These agreements are typically one-to-one between companies that are otherwise competitors, so the terms are typically confidential. The result is that, today, no one really knows the details of how much traffic there is on the Internet, or how the principle peering arrangements work.

In the early (and more public) stages of peering, no backbone provider wanted to carry traffic to destinations not on their own network, so something called "hot-potato routing" became the rule. Hot-potato routing means you hand off packets to the destination network as quickly as possible, and the destination network then has to carry them across the country or wherever else they must go.

Of course, with any set structure, someone will find new ways of making money. One result of hot potato routing is a new type of backbone "network" - one that hosts major web sites, but doesn't actually carry its own packets very far. This is possible because worldwide web traffic is very asymmetric. A request packet of less than 100 bytes can easily generate 20 kB or 100 kB of return data. With hot potato routing, the web hosting network carries the small packets while the "peer" backbone carries the heavy flow.

Now, no one wants to provide a subsidy to a competitor, so earlier this year, UUNET gave notice to several of the networks they peer with that they would no longer provide "free" peering. The criteria for being cut off appear to involve asymmetric traffic flow. Since almost all peering agreements are made under nondisclosure agreements, there's rampant speculation about exactly who was subject to these cut-off notices. And, with the speculation, came cries for government regulation.

But no regulation is needed. Market arrangements are already evolving. As the web is increasingly used for commercial purposes, backbones are becoming a commodity. The critical issue today is access to individual subscribers - subscribers who buy goods and services. What counts is who has the best access to the most subscribers. And here we find thousands of independent Internet Service Providers (ISPs). Despite repeated forecasts of future ISP consolidation, the number of ISPs keeps growing. And, just when UUNET is trying to cut back on their "free" peering agreements, a new class of backbone provider is emerging that will render the current discussion moot.

Today, if I want to advertise or sell on the Internet, I want my web site to be rapidly accessible by the largest number of prospective customers. What's emerging to help is a group of "tier one aggregators" - companies like Frontier and AboveNet who host web sites and purchase direct trunks from their web sites to those ISPs that give them one-hop access to the maximum number of consumers. Companies with products to sell reap real advantages and so will pay extra to have their web site be rapidly accessible by the average consumer. If UUNET can't provide one- or two-hop access to a substantial group of consumers, then who cares what their peering policy is?

And other, new network paradigms are emerging. Skycache is offering a satellite-based distribution scheme that allows local ISPs to dramatically reduce the amount of bandwidth they must buy from the traditional backbone providers, while at the same time improving download performance for their customers. In this scheme, Skycache aggregates the statistics of all the web page requests of all their customers' customers. Based on these statistics, they stream copies of the most popular pages via satellite to the ISP's site.

At the local ISP, conventional web caching technology stores these popular pages until a customer makes a request. Since statistics are aggregated from a very large pool of users, the chances the desired web page will be in the local cache are very high. Again, the backbone provider faces competition from a new source. If Worldcom and MCI had merged their Internet backbones, it's not clear it would have mattered. With individual subscribers spread amongst over 5,000 ISPs, new backbones and new backbone paradigms will emerge no matter what the peering policies of today's backbone providers.

These are just two creative ways to bypass traditional backbones and let the free market take its course. The current backbone providers can fight over their peering agreements, but if they don't provide rapid access to a large pool of consumers, they are all going to be bypassed in the next twelve months - making regulation of peering irrelevant.

LEARNING FROM THE PAST
The best justification for government regulation is in the case of a "natural monopoly." If only we could agree what that is. The public right-of-way, in urban areas, seems to be a natural monopoly. But, contrary to 19th century railroading practice, long-distance rights-of-way don't seem to be a natural monopoly. Increasingly, long-distance rights-of-way are private - railroads, pipelines, electric utility corridors, microwave and fiber corridors, and even some highways - and this works just fine. Indeed, when the Bell System was broken up in the early 1980s, the deregulated half - the long distance market - blossomed with new services running on alternate long-distance rights-of-way.

But if the goal of the Bell System breakup in 1983 was to take maximum advantage of market forces, we made a mistake. In breaking up, or partitioning, the Bell system, we assumed local telephone service was the natural monopoly. With the advent of xDSL technology, it has become clear the natural monopoly is not the provisioning of local telephone service. In reality, the only pieces of the infrastructure that are in some sense a natural monopoly are the wires on the telephone poles and in the conduits in the street. The services provisioned on those wires would serve us better today if they were subject to market forces, not monopoly and regulation.

For example, there are competitive local exchange carriers (CLECs), companies like COVAD and Northpoint, who are focused on data connectivity. But they are complaining of difficulty gaining access to copper wires. Apparently, the Telecommunications Act of 1996 has not yet managed to make existing copper wire readily available to non-Bell companies. In fact, in the past two years, some incumbent Local Exchange Carriers (ILECs), US West, for example, have largely succeeded in eliminating all of the historic tariffs under which it was possible to rent bare copper wire - wire which could have been used for xDSL service.

So here's a place where it might make sense to restructure the monopoly. Should we now break up the ILECs into regulated suppliers of local copper or fiber cables on the one hand, and unregulated providers of service over those cables? A radical idea, perhaps, but it's more relevant than regulating "peering."

WHAT COUNTS FOR THE FUTURE
If one were to propose regulation for any part of the Internet infrastructure, the "last mile" - that is, the local connections to homes and businesses - seems the best place to focus. After all, if Internet subscribers were forced, by access considerations, to move from today's independent ISPs to their local telephone company, it would be a major setback for competition. But even here, I'm not overly worried about a lack of regulation. There are enough players in the market to allow competition.

In the last mile, the legacy Bell companies, whether called ILECs or RBOCs, control only the twisted-pair copper infrastructure. A separate "monopoly" exists in the local cable TV companies that control coax cable access to the home. For Internet services, cable and telephone companies are head-to-head competitors. There are also the electric utilities. They may not know how to leverage their power distribution networks for IP access just yet, but there are many companies hoping to teach them.

And, finally, there is wireless IP access, the best long-term solution for local connectivity (see "On the Horizon," June 1998). While most of our radio spectrum is currently bound up by extensive regulation, there is no reason why wireless technology can't provide broadband access to both home and business. Prices are still high, but Moore's law is reducing wireless access costs more than those of landline access, so we'll see an economic cross-over in the next few years. Companies like Rooftop Communications and Metricom are already deploying wireless IP solutions today.

So, even though deployment of local broadband access is being slowed by the existing monopolies, there is competition. Broadband access in the "last mile" will not be blocked. Regulation of local access might help, but it's not essential.

In the past, we've misjudged exactly what "natural monopolies" were appropriate. With today's hectic pace of technology evolution, we should be careful in any attempts to regulate, working instead to limit or reduce the scope of those monopolies that already exist.

Brough Turner is senior vice president of technology at Natural MicroSystems, a leading provider of hardware and software technologies for developers of high-value telecommunications solutions. For more information, call Natural MicroSystems at 508-620-9300 or visit the company's web site at www.nmss.com. E-mail to the author ([email protected]) is also welcome.