Kenneth Neil CUKIER <firstname.lastname@example.org>
The development of electronic commerce outside the United States suffers because of the high cost and poor state of telecommunications infrastructure, which creates the situation whereby a large volume of global Internet traffic passes through the United States before returning to the region of origin. Despite new infrastructure, the situation is likely to endure. However, policy makers can pursue initiatives to reduce this problem by encouraging more efficient regional Internet infrastructure.
A revealing sign that the Internet's use for electronic commerce might pose a few challenges for companies and users outside the United States was demonstrated last year by a peculiar development in one Internet service provider's (ISP) business strategy.
In October 1998, FranceNet SA, a Paris-based firm that hosts corporate websites, announced it was opening up a powerful server center in California, leasing space from the U.S.-based company Frontier Global Center. Rafi Haladign, FranceNet's president, explained that hosting the sites from the United States would improve the performance not only for U.S. users, but for users in Asia and Europe as well. 
Although the company might be lauded for pursuing a global strategy, to those familiar with the Internet's current topography the news served as a potent symbol of defeat: Why should a company called FranceNet choose to reach European users from the United States? Why should it pay a U.S. company and rent server space to do so?
FranceNet's experience is not unique. Many companies around the world understand that despite the Internet's inherently meshed architecture, the network has a nucleus and it is the United States -- so companies host their sites there.
Moreover, ISPs outside the United States are in the uncomfortable position of leasing a disproportionately large amount of bandwidth to the United States to route regional traffic, since in many cases it costs less than leasing a high-speed capacity line to a neighboring country in their own region.
This situation -- commercially sound though architecturally inefficient -- is an important detriment to electronic commerce worldwide. A robust Internet can develop only if there is satisfactory regional telecommunications (telecom) infrastructure outside the United States, in terms of price and performance.
Certainly many other factors play a role in the development of international e-commerce, such as prices of computers, Internet penetration rates, high-speed access on local-loop infrastructure, skilled labor, consumer behavior, a competitive commercial climate that encourages entrepreneurship, and a regulatory framework that promotes confidence in electronic transactions.
However, sufficient international infrastructure and routing at the backbone level is indispensable, since these other factors must rely on the long-haul physical transmission network itself. It makes cyberspace a very tangible entity.
This paper examines the manifestations of the so-called "U.S.-centric Internet" by looking at its effects on regional Internet deployment. It then explains how this phenomenon historically developed by reviewing the construction of international telecom networks. It details the consequences for countries if this situation is not changed, and notes that new regional infrastructure outside the United States will not likely solve the problem, since it is dwarfed by the enormous amount of transoceanic links that will light up over the next three years. Finally, the paper concludes by offering a framework for coordinated government action to help reduce the problem.
Only if industry and government are aware of the situation and its causes -- and see common approaches to solve the problem internationally -- can a hearty e-commerce environment exist for the benefit of businesses and consumers of all nations.
FranceNet's experience was not specifically mentioned when Francis Lorentz, the head of the French government commission on e-commerce whose citation begins this paper, decried the U.S.-centric Internet. At a colloquium packed with over 200 government officials and senior industry executives at France's Ministry of Economics, Finance and Industry in February 1999, Lorentz noted that the volume of Internet traffic from France that passes through the United States is over half of the total French Net traffic. 
The high demand for U.S.-based content, the lack of regional infrastructure such as low-cost and high-speed international circuits, and the lack of dominant interconnection points to exchange traffic all serve as both causes and embodiments of the underlying problem: an Internet that is not global, but centered in the United States.
This leads to inefficient Internet traffic routing between countries outside the United States, and results in requiring ISPs that handle e-commerce traffic to pay U.S. carriers for access to the Internet.
Barry Castle, an executive at 3Com Corp. in the United Kingdom and a founder of the Internet Alliance for Europe, believes that Europe's high cost of bandwidth is retarding the development of the Internet in Europe and also is funneling money to U.S. carriers who handle the majority of Europe's traffic when European ISPs use the United States as their international routing hub. 
In Singapore, Liana Raveendran Greene, the board member of the Asia and Pacific Internet Association, decries the high-cost of regional bandwidth and the lack of cooperation among different national backbone projects that do not interconnect with one another in the Asia-Pacific, but only in California. 
Hard data on international Internet traffic volume are difficult to come by, since carriers are reluctant to share the information, citing competitive concerns. However, figures from September 1998 in a study by the Washington, DC-based telecom statistical think tank TeleGeography Inc. graphically show the magnitude of the matter (see backbone connectivity map for Europe below; figure 1). 
The study notes that in Europe, the maximum link between any two countries was under 450 megabits per second (Mbps), while U.S.-bound bandwidth is over 3.5 gigabits per second (Gbps). In Asia, no two countries shared links of over 155 Mbps, while capacity to the United States is around 2 Gbps. Specifically, Singapore and Malaysia share a 45 Mbps link, yet have a combined bandwidth of 450 Mbps to the United States. This is surprising, and possibly caused by traffic going to the other country traveling first to the United States, since the two countries represent each other's most frequent international telephone call destination: 56.1% percent of all international calls from Malaysia go to Singapore, and 35.1 percent of Singapore's go to Malaysia. 
In Africa, the situation is more dramatic. The only country that has any continental interconnections is South Africa, which shares links with Namibia and Swaziland. While nearly all African nations have Internet connectivity, in every other case the connection is via an industrialized nation (usually its former colonial power). There are no interconnection points for pan-African traffic. Save for the two South African cases, traffic from any nation to another must first pass through a third country. For example, Benin and Burkina Faso border each other in West Africa, yet traffic between the two nations travel via France (for the former) and Canada (for the latter).
European and Asia ISPs state that up to 75 percent of traffic goes first through to the United States, a portion of which is routed back to the region. Bruno Havet, a French researcher, has noted that in one case, an independent French ISP called ISDNet exchanges traffic with incumbent carrier France Telecom in the United States. 
What these bandwidth allocation statistics mean practically for ISPs in those regions are multifold: First, traffic travels via a longer distance connection rather than a more direct route -- which is technically inefficient. According to the underlying concepts of the Internet's routing design, traffic should take the most direct path. Indeed, the architecture of transmission control protocol/Internet protocol (TCP/IP) allows for dynamic updating to find the most efficient path for the data flow. Yet the current topology is not truly distributed if there is a preponderance of bandwidth on certain routes (i.e., to the United States) based solely on economic causes (the lower cost and better availability of bandwidth) rather than engineering ones (the most direct route for traffic to travel).
Second, under these circumstances, the traffic travels on an international link, which means that a foreign carrier -- usually the United States -- receives income from the local ISP. In essence, the development of the Internet in a country outside the United States requires a payment to a U.S. carrier. For example, 71.6 percent of Singapore's Internet traffic travels between North America, carried by the Canadian/U.S. operator Teleglobe Inc. Much of this traffic can be assumed to reach other parts of the world, the Paris-based Organization for Economic Cooperation and Development (OECD) notes, since only 8.2 percent of Singapore's voice traffic goes to North America. 
The strange topology of the infrastructure for e-commerce is further demonstrated by the reasons that justify an ISP outside the United States interconnecting with other ISPs from its region in the United States. The OECD shows that Internet traffic travels between Singapore and another country in the region at 885 milliseconds on average, while the average time between Singapore and North America is 382 milliseconds.  Thus, there is not only a problem with today's routing topology, but since the North American-bound link is often less expensive, and permits the ISP to access U.S.-based content as well as interconnect with other national ISPs, there is a commercial incentive to maintain this situation, even though it is self-defeating in the long term for the foreign carrier, since it entrenches the position of the United States and U.S. carriers for e-commerce traffic.
Some industry executives and policy makers believe that bringing U.S.-based content closer to the source of the demand will ease the problem. But hosting U.S.-based Web content regionally does not help, as the experience of Stefan Westman, the Internet network manager of Swedish operator Telia, shows. After Telia began hosting a European version of the AltaVista search engine in Stockholm, the carrier found that it while it saved on the cost of U.S. bandwidth, it simply deferred the cost toward dedicating more intra-European bandwidth, which is more expensive, since traffic requests were going to Sweden rather than the United States. It cost more to host the site in Europe from an infrastructure point of view than to transit the Atlantic to meet users' page requests.
Asian and European multinational corporations and ISPs state that the region's high cost and insufficient telecom infrastructure holds back e-commerce. It is because of a lack of region-wide planning historically and their political geography that small and fragmented nations in some circumstance are reluctant to cooperate on pan-regional initiatives. Also, the differing languages thwarted major traffic volumes between the countries, and the different currencies made international development more complex. Finally, both regions are just now beginning to experience telecom liberalization.
According to a February 1999 report by the International Telecommunications Users Group , prices for international circuits are as much as six times the price for equivalent national circuits in different European countries. Moreover, the study notes that although the prices are actually increasing: in 1997, the prices were only two to five times higher. For instance, a 34 Mbps international half circuit in Spain has a list price of 42.75 million pesetas a month, more than six times the 7.04 million pesetas for a national leased line of the same distance. 
These prices are vastly higher than similar infrastructure in the United States. For example, a 300-kilometer, 2 Mbps circuit in a Europe country costs up to four times more than in the United States, and up to 17 times more if it is an international link in Europe. 
European regulators are aware of the problem. Hubert Ungerer, the head of the competition policy unit at the European Commission in Brussels, laid out the stakes, emphasizing the "requirement to re-balance market power in the Internet between Europe and the United States."  "The major cost to the European user is therefore not the ISP, but the price paid to the telco [for leased lines]," he said, decrying the "high costs in Europe for the backbone capacity."
In Asia, the Asia-Pacific Economic Council's working group on telecom is currently preparing a report on the effects of the high cost of bandwidth regionally, and the economic implications of needing to link with the United States. The International Telecommunication Union, the United Nations-affiliate agency in Geneva, also said during its ministerial meeting in Minneapolis, Minnesota, in October 1998 that it will study the matter.
Essentially, the policy makers will find that the incentives to hub Internet traffic via the United States -- despite holding back regional e-commerce development -- is powerful. According to Jim Dixon, the director of the EuroISPA telecom working group, it is often less expensive to purchase capacity from any European capital to the United States than from one capital to another.  He said that the price of a national circuit from London to a provincial city is 20 times more than a circuit for the same distance in the United States, making it cost-effective to lease international bandwidth to the United States to reach other parts of the United Kingdom. "All the technologies are behind in Europe compared with the U.S. due to the high cost of bandwidth," Dixon says. 
In a filing to the Asia Pacific Economic Council (APEC), U.S. carrier AT&T explained the matter with weighted data from three real, yet anonymous, nations in the Asia-Pacific region (see figure 2). The carrier argues that "The competitive, open market environment in the U.S. has driven the price of the U.S. half-channel to levels significantly lower than half-channel prices in other APEC economies." 
Figure 2: Relative International Facilities Prices (U.S. Half-Channel Equals 1) _________________________________________________________ To: Economy 2 | Economy 3 | USA From: _________________________________________________________ Economy 1 1.3 + 1.3 = 1.4 + 1.4 = 1 + 1.4 = 2.6 2.8 2.4 _________________________________________________________ Economy 2 X 1.51 + 1.4 = 1 + 1.51 = 2.91 2.51 _________________________________________________________ Economy 3 X X 1 + 1.4 = 2.4 _________________________________________________________ (Source: AT&T filing to APEC.)
AT&T's calculations show that it costs more to lease intra-Asian circuits than it does to purchase circuits from U.S. operators to the United States. In this example, all half-circuit charges to a second destination in Asia cost between 0.3 and 0.51 more than they do to the United States via AT&T (1.0). As a result, there is a financial incentive for ISPs from two different Asia-Pacific countries to carry their traffic to the United States and interconnect there rather than route the traffic directly to the country of destination. The U.S. carrier concludes that: "the real key to the development of the Asia-Pacific Internet Infrastructure is the opening up of markets to the competitive provision of facilities-based telecom services." 
FranceNet faces a big problem, whose roots stretch back a century when sub-sea cables were laid for international voice telephone calls. Also, the Paris-based ISP will have a tough time turning back the hands of time to remedy the initial architecture and routing design established 30 years ago when the Internet's predecessor, ARPANET, was created.
The United States is the center of the Internet for two important reasons: First, the U.S. backbone, composed of the tier-one ISPs like MCI WorldCom's UUNet, Sprint, and GTE Internetworking, controls the vast majority of the Internet's routing table. That means they are able to originate or terminate traffic on their own network without interconnecting. But what this means for other ISPs is that they need to interconnect with one or more of those tier-one ISPs to reach what can well be considered "the global Internet." A lack of such an interconnection would not enable the other ISP to reach all Internet users. And since the largest number of users, as well as the world's most popular Internet content, is based in the United States and hosted by these ISPs directly or via down-stream ISPs, ISPs outside the United States must connect with the U.S. backbone else they are effectively not on the Net.
While this only explains why a non-U.S.-based ISP needs to link to the United States, it also reveals the incentive to hub traffic in the United States to be re-routed elsewhere: All ISPs outside the United States share this market dynamic, and so nearly all have direct connections to the United States. As a result, they find it more convenient and less expensive to exchange traffic in that one central hub, rather than dedicate infrastructure to every country where they need to reach ISPs to swap traffic. Finally, as noted above, the U.S.'s low telecom costs for high-speed bandwidth create an incentive to hub traffic there even if the traffic returns to the same region.
This latter point -- the need for ISPs to exchange traffic -- is essential. While Europe and Asia have built out interconnection points over the past two years, they are mainly for national ISPs to swap traffic. In Europe, there is no dominant exchange point, although London and Amsterdam both serve as large entrepot for international operators. In Asia, the five backbone networks -- the A-Bone in Japan, Hongkong Telecom's Net Plus, Kokusai Denshin Denwa's Internet KDD, Singapore Telecom's Singtel, and Telstra's Big Pond -- do not interconnect regionally with each other.
Thus, the U.S.'s content, users, low-priced high-speed circuits, and "routing table," along with the lack of viable central interconnection points in other regions, leads to the U.S.'s role as the center of the Net. Yet there is still another reason why the situation cannot be resolved easily: the history of international telecom development.
Today's physical telecom network was built based on assumptions about international call traffic patterns. For example, Americans trade with Britons, share a common language, and are two of the richest nations. Unsurprisingly, the capacity between the two countries is among the best in the world. And since it has traditionally been a lucrative telecom route, an ever-increasing number of competitors have an incentive to construct new capacity between the two nations since the likelihood of recouping the cost is greatest. By the same token, the telecom infrastructure between countries that lack these characteristics are poor; for example, between Vietnam and Indonesia.
When the Internet was first developed, and rose in popularity in the early 1990s, it rode atop of the existing telecom infrastructure that was constructed for the purpose of long-distance voice calls. For this reason, the best prices and best-performing links -- as well as the best availability -- were on U.S. routes, the most popular incoming and outgoing calling destination, and so where the most infrastructure was built. U.S. carriers own the vast majority of that capacity.
For example, in terms of minutes of voice traffic, Europe originates half as much traffic to North America as the latter to Europe. And Asia accounts for only one-third as much traffic to the North America as vice versa. In 1997 Europe and Asia swapped 1.7 billion minutes between themselves, while North American communications to both regions tallied 13.3 billion minutes. 
Most of today's international Internet traffic still gets carried over the fiber optic cable and satellites originally intended for voice traffic, and therefore follows the most dominant route: to the United States. New infrastructure builds, intended for data traffic, remain on transoceanic routes, due to a smaller barrier to entry. The lack of telecom liberalization outside the United States, along with the lack of rights-of-way to lay physical fiber, has meant a slower telecom infrastructure build in other regions. New satellite facilities for data may change the pattern of international Internet traffic, but these projects are far from completion.
FranceNet's business customers seeking to enter the e-commerce marketplace and reach a global audience are probably unaware of these factors -- the high cost and lack of telecom infrastructure outside the United States and the historical development of international voice circuits -- that have influenced the Internet's global topology. Nor do they probably know where the money flows, which is in the same direction as Internet traffic itself: toward the United States by way of U.S. carriers. All they see is what is on the surface: their monthly bill, which is higher than in other parts of the world.
The high price of an ISP's bandwidth, which leads to the U.S.-centric network phenomena, imposes a heavy cost on the deployment of e-commerce in regions outside the United States.
Corporate websites have an incentive to host their operations from the United States rather than their country of origin. This creates a capital flight away from the country and to the United States in terms of telecommunication fees. Moreover, it also may mean that the U.S. subsidiary receives the operational revenues, as well as new jobs created by the business. Politically, national governments outside the United States may also lose out on potential taxation revenue if the transaction occurs in the United States.
Finally, hosting an e-commerce website from the United States to improve performance for a global audience also makes commercial sense to reach customers in the United States, where the majority of Internet users are based. To apply the traditional economic notion of network externalities, a greater number of users have the potential to be interconnected to the e-commerce application when it is hosted in the United States as opposed to other regions, since one factor that determines whether a site is accessed is network performance, which is better via the United States. Thus the site's value, in terms of network externalities, is increased when it is hosted from the United States. Yet in effect, U.S. customers gain disproportionately greater performance than do users in other regions, even though it is a non-U.S. company's site that is hosted there, aimed at a global market. 
AT&T, in its filing to APEC, states: "[L]imitations on the sources for underlying telecom infrastructure restrict the ability of ISPs to bring new applications to their customers." 
The United States accounts for 58.1 percent of all Internet hosts,  and of the top 100 websites, only 6 are based outside the United States.  The OECD notes the United States has over 6 e-commerce websites per 100,000 inhabitants, compared with the OECD average of 2 sites. 
One example of the U.S.-centric Internet's impact on e-commerce deployment outside the United States is the case of Disney Online. Its website for the European market, which was hosted from Amsterdam and London, was relocated in 1998 to Hawaii, where it is hosted on the Digital Island Internet backbone. Company executives say that the reason for the move was the better network performance for European users by hosting the content in Hawaii rather than in Europe itself, even though this is where the traffic is aimed. 
Unless the situation changes, Europe and Asia may lose a significant portion of potential e-commerce trade and revenue. However, the key factor to re-draw the balance for network traffic -- new regional infrastructure -- is not enough.
Massive initiatives to build more bandwidth capacity are being undertaken around the world. Europe's liberalized telecom market has sparked new network construction by traditional telecom operators, public utilities, auto routes, and new entrants. In Asia, national network development projects are ongoing, yet regional projects have been delayed because of the economic crises. In Africa, recent telecom initiatives are aimed at improving teledensity by installing new local lines; plans in 1995 for a sub-sea fiber cable around Africa to link countries to each other have been scuppered; and a regional satellite consortium is a decade behind schedule.
This leaves Europe, for the purpose of this paper, as the sole credible region where the U.S.-centric Internet's dominance may be reduced by recent industry trends toward building bandwidth and lowering its cost. However, an examination of new projects reveals that despite a dramatic increase in pan-European infrastructure, an even greater proportion of capacity is being built on transoceanic routes, which will likely lead to a continuation of the gross disparity between the cost of regional links and connections to the United States. In short, the U.S.-centric Internet may well be preserved.
During the next two years, over 10 major network construction projects will create a boom in European bandwidth, yet it may still fall short of the capacity being built on U.S. routes (see figure 3).
Figure 3. New Network Construction European Carrier | Planned Capacity _________________________________________________ BT/AT&T 200 Gbps Cable & Wireless 300 Mbps Carrier One 34 Mbps EBN * 2.5 Gbps Esprit 2.5 Gbps Flute 2.5 Gbps Pan European Crossing 960 Gbps + Hermes 180 Gbps KPNQwest 9.9 Gbps Level 3 N.A. Viatel 320 Gbps MCI WorldCom 40 Gbps Europe-U.S. Projects | Planned Capacity _________________________________________________ Atlantic Crossing 960 Gbps + Oxygen # 1,280 Gbps TAT-14 640 Gbps Asia - U.S. Projects | Planned Capacity _________________________________________________ China-U.S. cable 80 Gbps G.-P. Cable 5 Gbps Japan-U.S. cable 80 Gbps Southern Cross 40 Gbps PC1 cable 80 Gbps Figures in megabits or gigabits per second (Sources: Yankee Group, TeleGeography Inc., company reports.) * European Backbone Network (France Telecom and Deutsche Telekom). + Total aggregated capacity worldwide; European and Atlantic bandwidth not known. # The project is speculative; capacity is worldwide.
Relative to the new bandwidth in Europe, the amount of capacity on U.S.-bound routes will likely be greater. One major trans-Atlantic project, Gemini, recently began operations at 30 Gbps. Together, these factors will likely continue the current price and performance incentives to route Internet traffic via the United States.
Despite these conditions, the situation is not entirely bleak for Internet industry executives and government policy makers outside the United States who wish to implant an e-commerce culture where companies in their regions can operate without the need to become customers of U.S. firms. The sorry straits of FranceNet need not be permanent. However, it will take imagination and bold actions to reduce the current pattern.
The way to remedy the situation is both to attack the core problem and to create the broader commercial landscape for the Internet industry to further develop; this requires more regional bandwidth, lower prices, robust competition, and international cooperation to create central regional hubs for routing Internet traffic.
The following initiatives serve as possible steps that can be pursued in a cooperative effort between the private sector, government, and academia to reduce the U.S.-centric Internet:
Whither our friends at FranceNet? Is this a case of "bandwidth colonialism" as the title of this paper suggests? Does the U.S.-centric Internet lead to other regions serving under the yoke of foreign masters? Regrettably, the answer is probably yes, though few might permit the image in polite society. Today's Internet topology reflects the unequal balance of power, in technical terms, among regions. It is influenced by history because of the way the Internet first evolved. Despite the meshed architecture of the Net, it resembles a star with the United States at its center today.
But just as the architecture of the Internet changed when the NSFNET backbone was privatized in 1995 and new backbones emerged, as the Internet develops internationally the topology will also change. In time, the U.S.-centric Internet may be fully eradicated and seen as a historic anomaly of its early development. Electronic commerce can truly flourish in an environment where companies have adequate resources -- bandwidth and technology -- to innovate and reach consumers.
As with colonialism, a financially and technologically superior entity is stripping away resources from the weaker entities. If information wants to be free, than bits want the best bargain -- and they route around obstacles like inferior, high-cost infrastructure. Exported are skills that could be honed, technologies that could be developed, and companies that could gain revenue by addressing customer needs.
But the means to reverse the situation is in the power of government and industry if they rise to the challenge quickly. As the world moves into an Internet era, the stakes merit the dedication by business leaders and government officials. Otherwise the information society will simply make some countries mere customers of others, and preserve -- or increase -- the inequality of economic opportunity afforded to businesses and citizens. FranceNet might just get to earn its name after all.
 Press conference on 6 October 1998; interview with author
 At conference "La nouvelle donne du commerce electronique," at Bercy in Paris, France, 4 February 1999. The figure does not indicate the volume that is sent back to Europe, although Lorentz's point is that the amount is large.
 Publicly stated at ISP Forum, Cannes, France, October 1998
 Presentation at the OECD Internet Governance workshop in Osaka, Japan, June 1998
 See TeleGeography 1999, a yearbook by TGI. While the map shown here does not account for new bandwidth allocation, it nevertheless demonstrates the discrepancy between regional and U.S.-bound bandwidth. The data are based on research from Cohen Communications Group. This author wrote an essay in TeleGeography 1999 for which this and other graphs were used. I am aware of no other data source on international bandwidth allocation by nation. A 1998 study by IDATE, a Montpellier, France-based market research firm, provides data on Internet bandwidth by country, yet does not breakdown where the connections are nation-to-nation.
 TeleGeography 1999
 CommunicationsWeek International, "Time to break US grip on Internet says OECD"; 20 April 1998.
 Presented at ISP Peering and Interconnection conference, London, U.K., December 1998.
 "Internet Traffic Exchange: Developments and Policy"; OECD, May 1998.
 INTUG Leased Line Survey 1999; Montrichard, France
 See CommunicationsWeek International, "Users mount assault on 'high-cost' leased lines"; 15 February 1999.
 Reuters - Marcel Corminboeuf, purchasing manager, Reuters EMEA Networked Economy conference, Paris, France, 14-15 May 1998.
 European Internet Service Provider Association (EuroISPA) conference, 4 June 1998, Brussels, Belgium.
 OECD Internet governance workshop; Osaka, Japan, June 1998.
 Interview with author, February 1999.
 AT&T, "AT&T Position Regarding Internet Charging Mechanisms and the Development of the Asia-Pacific Internet Infrastructure" presented to the telecommunications working group of the Asia Pacific Economic Council, February 1999; for the International Charging Arrangements for Internet Services (ICAIS) study.
 TeleGeography 1999.
 For a fuller discussion of network externalities related to the U.S.-centric Internet phenomena, see Rob Frieden; "International Charging Arrangements For Internet Services, Module 1" (ICAIS), for the APEC Telecommunications Working Group, February 1999. <http://www.pecc.org>
 Network Wizards, July 1998
 Web21 statistics, 1998. Some sites may have developed overseas operations since the time these statistics were compiled.
 OECD, Internet Infrastructure Indicators, November 1998.
 Interviews with author May 1998 and January 1999
While this paper is based on numerous conversations with cherished people, I wish to single out Sam Paltridge of the Organization for Economic Cooperation and Development, whose truly pioneering work is the muse for my own examinations. Also, to George Sadowsky of the Internet Society, who encouraged its creation and aided its realization. Finally, I thank TeleGeography Inc. for permission to use the map of Internet backbone connectivity, and my employer CommunicationsWeek International for the opportunity to report on these themes.
Kenneth Neil Cukier is a senior editor and Paris correspondent for Communications Week International, covering the technology, economics, and public policy of the Internet. From 1992 to 1996 he worked at The International Herald Tribune in Paris.
Kenneth Neil Cukier <email@example.com>
86 rue de Varenne
75007 Paris, FRANCE
Tel: (33) 1 45 55 00 55
Fax: (33) 1 45 55 03 55