Digital Borders

In the 1990s, many (…) scholars believed that the Internet was eroding the authority of governments. The web’s salient features—instant and universal communication, geographical anonymity, and decentralized routing—made it easy for computer users inside a nation to get illegal information from computers outside the nation. American college students could download copyrighted songs from servers in the South Pacific and bet on digital blackjack tables on computers in Antigua. Saudi Arabians could access porn sites in Holland, and Italians could read banned books on web pages hosted in Australia. Nations seemed unable to stop violations of local laws via the Internet.

This conception of the Internet began to crumble in April 2000, when two French antiracism organizations sued Yahoo!, the American Internet portal, in France. The groups charged Yahoo! with hosting Nazi auction sites that were accessible in France and that violated French laws against trafficking in Nazi goods.

At the time, Yahoo! was the entrance point for more Internet users than any other website. Jerry Yang, Yahoo!’s billionaire cofounder, was confident and brash—he and David Filo had chosen the company’s name because, according to the company’s official history, they “liked the general definition of a yahoo: ‘rude, unsophisticated, uncouth.’ ” Obsessed with expanding the firm’s market share, Yang thought governments dumb and speech restrictions dumber still. When Yahoo! received a summons from Judge Jean-Jacques Gomez of Le Tribunal de Grande Instance de Paris, a French trial court, Yang shrugged. Reflecting conventional wisdom, he believed French officials had no authority over a computer in the United States.

And if France could do nothing to stop Yahoo! in the U.S., it also seemed hard for French officials to block access to the Nazi auction sites in their country. Too many Internet communications crossed France’s borders for the government to stop and screen each one. The Internet’s decentralized routing system carries messages from point to point, even if some connections along the way are blocked, damaged, or destroyed. “The Net treats censorship as a defect, and routes around it,” declared John Gilmore, the libertarian Internet activist who cofounded the Electronic Frontier Foundation. To keep out the Nazi pages, it appeared that France would have needed to shut down every single Internet access point within its borders—a seemingly impossible task.

Yahoo!’s arguments were premised on the 1990s vision of a borderless Internet. Half a decade later, this vision is fast being replaced by the reality of an Internet that is splitting apart and reflecting national borders. Far from flattening the world, the Internet is in many ways conforming to local conditions. The result is an Internet that is increasingly separated by walls of law, language, and filters. This bordered Internet reflects top-down pressures from governments like France that are imposing national laws on the Internet within their borders. But it also reflects bottom-up pressures from individuals in different places who demand an Internet that corresponds to their preferences, and from the web page operators and other content providers who shape their Internet experience to satisfy these demands.

THE ECONOMIST ANNOUNCED IN 1996 THAT “English may now be impregnably established as the world standard language: an intrinsic part of the global communications revolution.” In the late 1990s, 80 percent of all information on the web was in English, according to the United Nations’ annual Human Development Report. English seemed on its way to becoming the Net’s universal language, much as TCP/IP had become the universal protocol that made Net communications possible. Al Gore captured this view when he recounted an episode during a 1993 visit to central Asia. “[T]he President of Kyrgyzstan told me his eight-year-old son came to him and said, ‘Father, I have to learn English.’ ‘But why?’ President Akayev asked. ‘Because, father, the computer speaks English.’ ”

By the end of 2002, however, less than 50 percent of the pages on the World Wide Web were in English. A seemingly essential feature of the Net turned out to be a temporary blip based on the fact that English-speaking Americans, who dominated early Net use, created the Net in their image.

Mighty Microsoft learned this lesson when it tried to distribute an English version of Windows in Iceland in 1998. Redmond executives thought that 500,000 worldwide Icelandic speakers did not justify spending money to translate the software program into their language, especially since most Icelanders also speak English. But Icelanders felt that Microsoft’s plan would imperil their language, which has retained basically the same grammar, spelling, and vocabulary for more than a thousand years. “It’s a very big danger because schoolchildren need computers, and the language of computers soon becomes the language of the kitchen,” said Kristjan Arnason, a professor of Icelandic at the University of Iceland. After the Icelandic government threatened to mandate local use of different operating systems (Apple software came in Icelandic), Microsoft relented and wrote an Icelandic version.

The Internet has been celebrated for allowing open, universal communication. “Information wants to be free,” said Stewart Brand, who cofounded one of the first online communities. But information does not want to be free. It wants to be labeled, organized, and filtered so that it can be searched, cross-referenced, and consumed. Information filtering is especially crucial to the Net, where it is so easy to publish, and where the danger of information overload is so great. In the late 1980s, the Internet already connected massive amounts of information that were notoriously hard to find. “It was akin to walking down each aisle of a library, scanning each book just to figure out what is there, but doing this all in the dark,” said Robert Reid, the author of Architects of the Web: 1,000 Days that Built the Future of Business. The Net became a revolutionary information technology only after Tim Berners-Lee developed the software and protocols of the World Wide Web, which made general browsing and hyperlinking possible, allowing users everywhere to easily organize, discover, exclude, and deliver content.

Geography turns out to be one of the most important ways to organize information on this medium that was supposed to destroy geography. Most people have preferences that cluster according to where they live. In addition to language, borders mark off differences in culture, currency, climate, consumer norms, and much more. People in Japan are interested in Tokyo’s weather and the value of the yen in Japanese, not the weather in New York or the price of American movie tickets in English. A Croatian who buys a Dell computer on the web wants not only instructions he can understand but also the address and phone number of a local repair service and a return address in the real world, in case problems arise.

JUDGE GOMEZ RULED PRELIMINARILY IN MAY 2000 that Yahoo!’s U.S. websites violated French law, and he ordered the company “to take all necessary measures to dissuade and make impossible” visits by French web surfers to the illegal Yahoo! Nazi auction sites on yahoo.com. Jerry Yang was dismissive. “We are not going to change the content of our sites in the United States just because someone in France is asking us to do so,” he said. “Asking us to filter access to our sites according to the nationality of web surfers is very naïve.”

Yang’s defiance reflected turn-of-the-century assumptions about the Internet’s architecture. Internet protocol addresses (each computer’s Internet ID), Internet domain names (such as mcdonalds.com or cnn.com), and e-mail addresses were not designed to indicate the geographical location of computers on the Net. These architectural “facts” meant that most users of 1990s Internet technology did not know where their e-mail messages and web pages were being viewed, and thus what laws in which nations they might be violating. Yahoo! said that it didn’t know where its users were, and which laws it should comply with.

Worse, if France could govern Yahoo! in America, every other nation could as well. And this raised the worrying possibility that Internet firms and users, confronted with a bevy of conflicting national laws, might begin to comply with the strictest among them in order to avoid legal jeopardy. “We now risk a race to the bottom,” predicted Alan Davidson of the Center for Democracy and Technology, in which “the most restrictive rules about Internet content—influenced by any country—could have an impact on people around the world.”

THE SPECTER OF CONFLICTING NATIONAL LAWS applying to every Internet transaction might have given Yahoo! the edge at trial, had it not been for the unlikely intervention of Cyril Houri, a Frenchman then working in New York’s Silicon Alley. On a trip home to Paris in 1999, Houri made a discovery that upended his career as a software engineer, not to mention conventional thinking about the Internet. Staying in his parents’ apartment, he turned on his laptop after dinner to check his e-mail. As the computer came on, Houri saw the portal he was used to seeing in New York. Blinking cheerfully at the top of his screen was a banner advertisement for an American flower delivery service, accompanied by a 1-800-flowers number usable only in the U.S.

In that moment, Houri realized that the Internet did not point inexorably toward the flattening of frontiers. He saw that, to the contrary, a borderless flower-delivery service made no sense at all. And he grasped that people would pay for software that took the boundaries of the real world and re-created them on the Internet, so that flower deliverers and a thousand other e-tailers would know where their customers were. There would be big money, he thought, in a technology that prevented people outside America’s borders from seeing the American ad, and that substituted a French ad for a French audience and a German ad for a German one. The same technology would allow news and entertainment sites to segment their content according to the whereabouts of their audiences. All it would take was a program to pinpoint the physical location of users. So Houri founded a dot-com, Infosplit, devoted to doing just that.

Ever since the Net became commercialized in the mid-1990s, Internet firms had tried, with varying degrees of success, to discover the geographical identity of their customers. The web’s omnipresent “choose a country” links are one way. Another is to ask users to type in an area code or send geographical identification (such as a driver’s license) by fax or mail before allowing access to a page. Yet another is to check the address associated with a credit card as proof of geographical identification. But these techniques are sometimes unreliable and, worse, they’re time-consuming. “The entire point of the Web is to bring you information simply and quickly,” thought Sanjay Parekh, the founder of the geo-ID firm Digital Envoy, during an “a-ha!” experience similar to Houri’s. “Why do I have to scroll through dozens of countries before accessing the site? Surely there has to be a way for [the site] to recognize where I am.”

In the past decade, Infosplit, Digital Envoy, and half a dozen other firms set out to make geographical identification on the Net easy, reliable, and invisible. Instead of requiring Net users to take steps to reveal or prove their location, they devised a way to identify a user’s location using the very features of Internet architecture that supposedly defied geography.

IP addresses (like “192.168.0.55”) don’t readily reveal a computer user’s physical location. But a savvy user can determine that location by sending “tracing” packets over the Internet. These packets report a list of computers through which they travel, much as a car driving along a network of highways collects a receipt at each toll. Just as a car’s origin can be determined by looking at these receipts, computers can examine the path of these packets to figure out the computers closest to the originator and recipient of any communication of the Net. This information can then be cross-checked against other IP databases that offer different clues about the geographical location of almost every computer connected to the Internet. When the databases are cross-referenced and analyzed, the location of Internet users can be determined with over 99 percent accuracy at the country level.

Internet geo-identification services are still nascent, but they are starting to have effects on e-commerce. Online identity theft in the U.S. causes firms and consumers to lose billions of dollars each year. Geo-ID is helping to solve this problem by identifying when stolen credit card numbers are used on the Net from locations like Russia, the home of many such scams. It is also improving Internet advertising, as Houri and Parekh envisioned, by making it easier to display ads geared to local conditions. And it is speeding the delivery of electronic data, allowing firms to deliver content from the closest “cache” website without having to ask the consumer where he is.

Finally, and potentially most important, these technologies are starting to enable the geographical zoning of entertainment. An important hurdle to the distribution of entertainment on the Net has been that certain material cannot lawfully be viewed in certain places. Geo-ID technologies can help to solve this problem by ensuring that online movies, web gambling sites, software programs, and other digital products do not enter countries where they are illegal. In other words, the software designed to respond to the local demands of consumers can also be used to help ensure compliance with different laws in different places.

FOLLOWING JUDGE GOMEZ’S MAY 2000 PRELIMINARY RULING, Cyril Houri contacted the plaintiff’s lawyer, Stephane Lilti, and told him that his software could identify and screen Internet content on the basis of its geographical source. Houri was invited to Paris where he showed Lilti how his software worked. When the plaintiff’s legal team saw what it reported, they were astonished. Yahoo!’s servers, which the firm had claimed were protected by the First Amendment to the U.S. Constitution, were actually located on a website in Stockholm. Yahoo! had placed a constantly updated “mirror” copy of its U.S. site in Sweden to speed access to the site in Europe.

When the trial resumed in July 2000, Yahoo!’s lawyers reiterated that it was impossible to identify and filter out French visitors to the firm’s U.S.-based websites. Lilti responded by explaining how Houri’s geo-location technology showed that Yahoo! auctions in France were not coming from servers in the U.S. Suddenly, the assumption that every web page was equally accessible to every computer user everywhere in the world seemed wrong. If Yahoo! could direct content to French users from Swedish servers, it could potentially identify users by geography and, if it liked, it could screen them out.

After receiving additional expert testimony about the feasibility of geographical screening, Gomez issued a final decision in November 2000, reaffirming that Yahoo! had violated French law by allowing Nazi goods to appear for sale on web pages in France. Gomez determined that the French court had power over Yahoo! and its servers because the company had taken conscious steps to direct the prohibited Nazi auction pages into France. He pointed out that Yahoo! greeted French visitors to its U.S. website with French-language advertisements. This showed that Yahoo! was tailoring content for France and that, to some extent, it could identify and screen users by geography. Acknowledging that 100 percent blocking was impossible, the court ordered Yahoo! to make a reasonable “best effort” to block French users.

At first, Yahoo! threatened to ignore Gomez’s decision. But the company had a problem: its assets in France, including income from its French subsidiary, were at risk of seizure. In January 2001, Yahoo! abruptly surrendered. It pulled all Nazi materials from its auction sites, announcing that it would “no longer allow items that are associated with groups which promote or glorify hatred and violence to be listed on any of Yahoo!’s commerce properties.” The company claimed that it was motivated by bad publicity from the Nazi auctions and not the French ruling. “Society as a whole has rejected such groups,” said a Yahoo! spokesperson. But the timing and threat of French sanctions suggested that Yahoo!’s will had been broken.

YAHOO!’S CAPITULATION SHOWS WHY YANG AND SO MANY OTHERS were wrong to believe that nations could not control the local effects of Internet communications originating from outside their borders. Using powers of coercion similar to those France wielded against Yahoo!, nations can exercise control over the Internet experience within their borders. They do so by threatening the local people, equipment, and services that enable local Internet users to consume illegal communications. Government action against such local intermediaries makes it harder for users to obtain content from, or transact with, the law-evading content providers abroad.

Underneath the mists and magic of the Internet lies an ugly physical transport infrastructure: copper wires, fiber-optic cables, and the specialized routers and switches that direct information from place to place. The physical network is a local asset owned by phone companies, cable companies, and other Internet service providers that are already some of the most regulated companies on earth. This makes ISPs the most important and obvious focal points of Internet control. For example, Germany, France, and England require local ISPs to screen out unwanted content from abroad once they are notified of its existence. But the true champions of information-transport control are found in China, which from the beginning maintained rigid control of every element in the Internet transport pipeline.

Information intermediaries are another ripe target of government control. Google frequently complies with requests to remove specified pages from its search results, usually because of alleged copyright or trademark infringements. Many of these pages are located on servers outside the United States, beyond the direct control of U.S. law. But the government, or those invoking its laws, can block the offshore content provider by going after the local search engine instead. Other countries are much more aggressive than the U.S. in using search engines to block unwanted context.

Financial intermediaries are yet another way that governments control unwanted offshore Internet flows. In response to the rise of web gambling services in the Caribbean, U.S. enforcement officials focused their attention on local financial intermediaries that help Americans ante up. In 2002, New York State Attorney General Eliot Spitzer used threats of prosecution to convince every major American credit card provider and online payment system (like PayPal) to stop honoring web gambling transactions. “With this agreement, we will cut off an enormous line of credit that was a jackpot for illegal offshore casinos,” Spitzer proclaimed. And the technique seemed to work pretty well, driving half of Antiguan web gambling firms out of business, and, in the words of the Antiguan prime minister, leaving a “significant, negative impact upon the [Antiguan] economy.”

Nations are using these and many other techniques of local coercion to control the Internet, including communications on it from abroad. These techniques are not perfect. Some determined web gamblers, for example, are wiring money from local banks to offshore banks—a strategy that will set off a new round of government responses. But such regulatory adjustments are as old as law itself. The law need not be completely effective to be adequately effective. All the law aims to do is to raise the costs of the activity in order to limit that activity to acceptable levels. Government regulation works by cost and bother, not by hermetic seal.

THE ENFORCEMENT OF NATIONAL LAWS in cases like Yahoo! Inc. and Yahoo France, and increasing consumer demand for Internet products tailored to local conditions, mean that what we once called a global network is becoming a collection of nation-state networks—networks linked by the Internet protocol, but for many purposes separate.

The bordered Internet is widely viewed to be a dreadful development that undermines the great network’s promise. But the Net’s promise was not fulfilled by the 1990s vision of an Internet dominated by the English language and the idiosyncratic values of the American First Amendment. People who use the Internet in different places read and speak different languages, and they have different interests and values that content providers want to satisfy. An Internet that accommodates these differences is a more effective and useful communication tool than one that does not.

People in different places, for example, disagree about what types of information they deem harmful. These differences are reflected in different national laws, and government officials charged with enforcing national values must enforce these laws, as cases like Yahoo! make clear. “Every jurisdiction controls access to some speech . . . but what that speech is differs from jurisdiction to jurisdiction,” explained Lawrence Lessig and Paul Resnick in a 1999 law review article. “What constitutes ‘obscene’ speech in Tennessee is permitted in Holland; what constitutes porn in Japan is child porn in the United States; what is ‘harmful to minors’ in Bavaria is Disney in New York.”

France’s ban on pro-Nazi speech is a reaction to its occupation by and flirtation with Nazi Germany during World War II, and its related belief that a person’s right to be free from threatening and degrading speech trumps the right to voice one’s political ideas, however harmful. The U.S., influenced by a very different history and tradition, takes a different view. These dramatically different attitudes toward proper speech among mature democracies reflect important differences among the peoples that populate these countries—differences in culture, history, and tastes that are legitimately reflected in national laws.

To understand the virtues of a bordered Internet, consider the opposite: an Internet dominated by a single global law. When you choose a single rule for six billion people, odds are that several billion, or more, will be unhappy with it. Is the American approach to Nazi speech right, or is the French variant? To what degree should gambling and pornography be allowed? Should data privacy be unregulated, modestly regulated, or heavily regulated? A single answer to these and thousands of other questions would leave the world divided and discontented.

SOON AFTER YAHOO!’S LEGAL DEFEAT IN FRANCE, the Chinese government insisted, as a condition of access to Chinese markets, that Yahoo! filter materials deemed harmful or threatening to the Communist Party’s rule. Yahoo! agreed to China’s demands, and by 2005, the company that was recently the darling of the Internet free-speech movement had become an important agent of thought control for the Chinese government. Yahoo! today provides Chinese citizens with a full suite of censored products. Its Chinese search engines do not return full results, but block sites deemed threatening to the public order by the Chinese authorities. Yahoo!’s popular chat rooms feature software filters designed to catch banned phrases like “multi-party elections” or “Taiwanese independence.” The company also uses a team of human censors who monitor chat room conversations and report the most flagrant offenders to the Chinese authorities.

China and other authoritarian nations represent the downside of the bordered Internet. But technologies of control in China are essentially the same technologies designed to satisfy consumer demand for geographically tailored Internet products. And as the Yahoo! case shows, governments in democratic states are using these technologies to respond to entirely appropriate constituent demands for protection from unwanted Internet harms. Technologies of control designed to serve legitimate and desired ends can rarely be limited to those ends, and will often be co-opted for illegitimate purposes. The more important lesson is that the Internet is not, as many in the 1990s believed, an unstoppable technological juggernaut that will overrun the old and outdated determinants of human organization. To the contrary, the Internet itself is taking on the characteristics—good and bad—of the governments and people beneath it in different parts of the world.

What does Jerry Yang think of these developments? “To be doing business in China, or anywhere else in the world, we have to comply with local law,” explained the one-time champion of Internet freedom. “I do not like the outcome of what happens with these things,” Yang said, “but we have to follow the law.”

Jack Goldsmith is a professor at Harvard Law School and Timothy Wu is a professor at Columbia Law School. They are co-authors of Who Controls the Internet?: Illusions of a Borderless World, from which this article is adapted.