Brace Yourself for the Segmented Internet

by Andy Oram
December 20, 2001

Nothing can stop the spread of the Internet. Even in Somalia, which lost all vestige of government and most of its economic activity during the 1990s, Internet service found a foothold in 1999.

As the influence of the Internet spreads, governments (sometimes reflecting popular opinion) are becoming increasingly dissatisfied with the de facto policies it brings. They will never cut off Internet access to the outside world, because they need it for commercial, academic, and other informational reasons. But they may be tempted to form their Internet users into a kind of local area network with particular policies, locally enforced, and with clearly demarcated gateways to the outside world. The result will be reminiscent of how private organizations segment their local area networks, so I call this possibility the segmented Internet.

Such a development would be very troubling to those of us who like to bring cultures closer together and foster a better understanding of shared values. Since I helped to create a document called One Planet, One Net for Computer Professionals for Social Responsibility, you can tell where my sympathies lie. But we have to understand why so many people around the world fear a borderless and policy-free communications medium, just as they fear so many other aspects of the trends lumped together as “globalization.” And notwithstanding John Gilmore’s famous claim, “The Net interprets censorship as damage and routes around it,” countries or regions may turn in desperation to technical measures that help them enforce policies within their boundaries.

Why Segment?

It’s surprising how many factors could tempt governments to create a demarcation between their networks and the rest of the world. I think it significant that the incident where someone came closest to segmenting the Internet, up to now, appeared not in China or Saudi Arabia but in France. Here, in a country with innumerable ties to the rest of Europe as well as to former colonies in Africa and elsewhere, a single judge called into question the undivided nature of the Internet.

The date was November 20, 2000, and the trigger was a lawsuit brought against Yahoo! by French anti-racist groups. Some of the auction sites using Yahoo! were offering Nazi material and memorabilia, which is illegal in France (and many other countries). Since the auctions are not illegal in the United States where the sites were located, the judge launched a special inquiry into geographic segmentation of the Internet. After consulting with technical experts, he declared that it was possible for Yahoo! to block French users from bidding on the Nazi material and ordered the French Yahoo! portal to do so.

The parent Yahoo! company declined to enter into this enormous technical experiment, and decided instead to ban all Nazi material. While the judge set an important precedent for restricting speech on the Internet, the threatened segmentation did not occur.

It’s a Hard, Hard Reign

Mainland China is well-known for banning speech, and for monitoring Internet traffic in order to spy on residents. To serve these needs, it licenses ISPs and places requirements on them to serve its desires for content control and surveillance.

Countries with relatively new and small user bases, and whose governments sense danger from outside influences, find this model appealing. For instance, the BBC reported on November 7 that “Iran’s Supreme Council of Cultural Revolution has ruled that the state alone can provide internet connections.”

But those objectionable practices are not the only possible motivations for segmenting and gatewaying. For instance, there’s also the bizarre domain-name controversy China had with mega-registrar Network Solutions (now a subsidiary of VeriSign).

In November 2000, the Chinese government decided that domain names with Chinese characters should be registered in a special manner that it considered more authentic than Network Solutions’s Chinese offering. The government further declared that only approved registrars in China could offer these names. The reasons for this ruling were the subject of much speculation, but the whole dispute suggested that nationalist controversies could be reflected by nationalist Internet policies. While there was widespread skepticism that such a rule could be enforced, China’s small number of interconnection points to the larger Internet would theoretically allow it to restrict all DNS queries to approved registrars. Meanwhile, some of us took a furtive pleasure in seeing a government that was willing to stand up to Network Solutions.

Wouldn’t trademark holders love a segmented Internet! Trademarks are associated to geographic regions by law. So if the resolution of a domain name depended on where the query came from, trademark holders could control names within the jurisdictions where they hold trademarks.

Don’t protest that this kind of conditional domain name resolution violates the whole concept of DNS. The trademark holders have been willing to do plenty of other things that violate its basic concepts, such as claim ownership over a second-level domain name across all top-level domains. But such conditional name resolution is probably unfeasible for technical reasons; more likely, governments will use a segmented Internet to enforce trademark laws within geographic jurisdictions.

The Soft Sell

Most countries would not emulate the Chinese government’s tight grip on Internet traffic, but they could be attracted to segmentation for other reasons. For instance, many African and Asian ISPs complain bitterly that they have to route traffic through the U.S. and back to their own continent because they have a lack of backbone infrastructure at home. The far-flung routing not only slows down traffic, but subjects traffic to fees from U.S. backbone providers and exposes sensitive data to snoopers outside the sender’s jurisdiction. I could see a country or a consortium of countries choosing to route traffic through a few gateways in order to encourage companies to lay backbones locally.

People in many countries decry the lack of local Internet content in local languages. Segmentation would not automatically lead to the creation of such content, but combined with differences in speed or pricing, it might encourage people to patronize sites in their own country.

Indeed, soft goals like building backbones and local content are more likely to succeed than hard goals like censorship and surveillance. People who really want to get around restrictions (usually to access erotic content, not political content) boast that they know how to do so, through such tactics as tunneling, spoofing addresses, and using steganography. Unless states succeed in clamping down on non-standard uses of the Internet, they are better served by giving a gentle boost to achieve less stringent goals.

What’s Good for Business

The most compelling reason I’ve found for segmentation lies in commercial policies. Governments have been exploring the poor fit between such policies and the Internet for many years, although the public hardly ever takes notice.

Here’s a typical issue. A consumer in Poland buys a product from a Web site in Spain. It arrives in a condition that the consumer finds unsatisfactory, but the Spanish company insists that it’s fine. In what country can the two parties arbitrate the dispute? If it comes to a court case, in which jurisdiction should it be brought?

The European Union and an even broader grouping, the Organization for Economic Co-operation and Development, have been arguing a host of such commerce issues. Each country has its own contract laws, its own liability laws, its own consumer protection laws. European Union ministers, for instance, have recently proposed a prohibition on unsolicited email.

And all these countries want the economic boost they expect to receive from Internet commerce, which forces them to decide what law applies. The recently concluded Cybercrime treaty, which has received some publicity, deals with a few commercial issues. Heaps of other issues remain.

A glance at the daily newspaper can tell you that international trade agreements are not making very much progress. The last few meetings of the World Trade Organization got almost nowhere, although the most recent Ministerial Conferences in Qatar showed a small advance because the developed countries gave in to demands from underdeveloped countries. Even the highly successful European Union is punctuated with numerous exceptions and agreements that cover only a subset of member countries.

So I don’t hold much hope that countries around the world can harmonize their laws related to ecommerce within a few years. If governments fail to come up with global policies, they may choose to go in the direction of segmentation.

Is It Possible?

Segmentation is hard to carry out, luckily. It would be relatively easy if IP addresses were assigned on a geographic basis. Then a government could apply a net mask to identify a set of IP addresses that start with the same string of bits, just as organizational administrators do in order to segment their LANs. But address assignment is poorly associated with geographic and national boundaries.

Both currently and under proposed IPv6 rules, only the topmost level of address allocation is done geographically. There are three Regional Internet Registries that hand out addresses in three regions of the world that correspond very roughly to the three superpowers in George Orwell’s 1984.

Each RIR allocates chunks of the space to large Internet providers that run backbones. Small providers connect to one or more backbones, and buy chunks of addresses from the backbone providers. These chunks of addresses in turn get subdivided further as other organizations buy connectivity to ISPs. Thus, ISPs that serve many countries will hand out similar addresses in each country. A company served by that ISP may also subdivide its addresses in different geographical regions. In a world with potentially billions of addresses, the strict relationship of addresses to routers will be critical to preventing routers from being overloaded with too many routes.

Some mild correspondence exists between geography and IP address (and some countries have a relatively homogeneous address space), but experts have told me the correspondence is too weak to make net masks a viable way to determine where a user is geographically. Countries may therefore resort to some sort of ISP regulation, as China has. It is also possible to maintain mappings that pinpoint most users through their IP addresses, as is proven by the existence of many services for businesses that want to serve geographically targeted ads on their Web sites.

Precedents exist in the U.S. In 1984, the military created MILNET, a network that was connected to ARPANET through gateways and that eventually turned into a class A network on the Internet. After the September 11 attacks, the new Advisor for Cyberspace Security proposed a similar network called GOVNET that would cover the entire federal government; the proposal is probably too grandiose to succeed.

What will Internet users do if matters such as consumer protection and legal liability are solved on a geographic basis? Can we still do business across borders? Yes, by forming webs of trust, as I suggested in my article Forget the Global Marketplace—Trade with Someone You Know.

One danger in the imposition of national policies is that the Internet could become another battleground for existing political debates. For instance, those who want to emphasize national culture (by providing incentives for people to stay within national borders and to provide local content in local languages) will have to deal with the reality that culture both crosses national lines and is divergent within them. This problem has been plaguing Quebec for decades. Similarly, there are right-wingers in Northern Italy who feel more cultural affinity with Austria and Bavaria than they feel for Southern Italy. In short, governments that experience Internet access as a threat to traditional policies may find that the challenge comes from within their borders as much as from the outside.


Andy Oram is an editor at O’Reilly Media. This article represents his views only. It was originally published in the online magazine Web Review.