Brewster Kahle unlocks the cellar door of a wooden building in San Francisco’s Presidio Park. He steps inside, turns on the fluorescent lights to reveal a solid black wall of humming computers, and throws out his arm theatrically. “This”, he says, “is the web.” It is a seductive idea, but the web isn’t really housed in a single San Francisco basement. Mr Kahle’s racks of computers merely store archived copies of many of its pages which Alexa, his company, analyses to spot trends in usage. The real Internet, in contrast, is widely perceived as being everywhere, yet nowhere in particular. It is often likened to a cloud.
This perception has prompted much talk of the Internet’s ability to cross borders, break down barriers and destroy distance. On the face of it, the Internet appears to make geography obsolete. But the reality is rather more complicated. If you want a high-speed digital-subscriber line (DSL) connection, for example, geographical proximity to a telephone exchange is vital, because DSL only works over relatively short distances. Similarly, go to retrieve a large software update from an online file library, and you will probably be presented with a choice of countries from which to download it; choosing a nearby country will usually result in a faster transfer. And while running an e-business from a mountain-top sounds great, it is impractical without a fast connection or a reliable source of electricity. The supposedly seamless Internet is, in other words, constrained by the realities of geography. According to Martin Dodge of University College London, who is an expert on Internet geography, “the idea that the Internet liberates you from geography is a myth”.
What’s more, just as there are situations where the Internet’s physical geography is all too visible when it ought to be invisible, the opposite is also true. There is growing demand for the ability to determine the geographical locations of individual Internet users, in order to enforce the laws of a particular jurisdiction, target advertising, or ensure that a website pops up in the right language. These two separate challenges have spawned the development of clever tricks to obscure the physical location of data, and to determine the physical location of users — neither of which would be needed if the Internet truly meant the end of the tyranny of geography.
Down on the Farm
To see just how little the Internet resembles a cloud, it is worth taking a look at where the Internet actually is. The answer, in short, is in cities. This is partly a historical accident, says Anthony Townsend, an urban planner at the Taub Urban Research Centre at New York University. He points out that the Internet’s fibre-optic cables often piggyback on old infrastructure where a right-of-way has already been established: they are laid alongside railways and roads, or inside sewers. (Engineers installing fibre-optic cables in a New York building recently unearthed a set of pneumatic tubes, along which telegrams and mail used to be sent in the 19th century.) Building the Internet on top of existing infrastructure in this way merely reinforces real-world geography. Just as cities are often railway and shipping hubs, they are also the logical places to put network hubs and servers, the powerful computers that store and distribute data.
This has led to the rise of “server farms”, also known as data centres or web hotels — vast warehouses that provide floorspace, power and network connectivity for large numbers of computers, and which are located predominantly in urban areas. A typical example can be found in Santa Clara, just off California’s Highway 101. It is run by Exodus Communications, a web-hosting firm which has nine server farms in Silicon Valley and another 35 around the world. From the outside, the farm is a deliberately nondescript building. A sophisticated security system, with hand scanners and video cameras, keeps out unauthorised visitors. Inside, the building resembles a jail, rather than a farm: it is packed with row upon row of computers in locked metal cages, their fans whirring and lights flashing. The air is filled with the deafening hum of air-conditioning. There are no windows and few people, and the lights are triggered by motion sensors, keeping unvisited parts of the farm in darkness. Exodus’s customers house their computers inside the metal cages, which are supplied with power and network connections. Most of the world’s biggest websites live in buildings like this; Exodus hosts 49 of the top 100.
As if to emphasise how physical constraints apply even to virtual spaces, server farms are still rented by the good old-fashioned square foot. According to figures published in April by Salomon Smith Barney, worldwide server-farm capacity is growing by 50% annually, and will reach 22m square feet by the end of 2001, despite the demise of the dotcoms. Cage space turns out to have other uses, too: boastful corporate logos hang from many cages, and some firms have posted job advertisements in the hope of poaching technical staff from rivals.
The signs are that the storage of information is going to become even more physically concentrated. One reason is the growth of “managed hosting” where, instead of renting space on a farm for their own servers, firms rent the computing capacity along with the power and network connectivity. In short, they simply hand over their data, and leave running the servers to the hosting company. As a result, there is no longer any need for customers to visit farms, so they need not be located in metropolitan areas, where space is limited and expensive. They can be anywhere, provided enough power and bandwidth are available.
In practice the constraint is power. A single server farm can consume as much power as a small airport, or four large hospitals. As a result, says Jon Feiber of Mohr Davidow Ventures, a venture-capital firm, the logical thing to do is to build out-of-town server farms with their own power stations. Such farms, he suggests, could be very large indeed: perhaps a dozen would be enough for the whole of the United States. Just such a facility, with a 24MW gas-fired power station, is being built just outside London by iXguardian, a British computer-services firm. It will be the largest server farm in Europe.
The combination of managed hosting and dedicated power stations means that data will be increasingly concentrated in large farms. The rise of wireless devices will drive this trend too: instead of storing data internally, such devices will store information on the network and access it when needed. But users wishing to access their data will still be spread out around the world. So centralisation will drive demand for technology that can smooth out the Internet’s geographical lumpiness and speed the delivery of data; in short, technology to obscure the physical location of Internet content from its users.
First, Hide the Data
One way to do this is to store copies of popular lumps of content in data caches sprinkled around the world. The leader in this field, with over 11,000 caching servers in 62 countries, is Akamai, a firm based in Cambridge, Massachusetts. The geographical distribution of Akamai’s infrastructure is strikingly different from that of Exodus. Broadly speaking, Akamai needs servers near the consumers of content, whereas Exodus puts its farms near the suppliers of content. Accordingly, Exodus has farms in North America, Europe, Australia and Japan, but not in Africa or South America. Akamai, on the other hand, has servers pretty much everywhere.
Akamai’s customers, which include CNN and Yahoo!, are content providers who are prepared to pay to ensure that users around the world are able to access their sites smoothly and quickly. Normally, when you visit a web server, a description of the page you have requested is delivered across the network. This consists of the page’s text, plus references to any graphics (or sound or film clips) associated with it. These items are then requested by your web browser and delivered across the network. Finally, the browser assembles all the components and displays the page. The problem is that while the text can be delivered quickly, the “heavy” items (such as graphics and video) are much larger and take longer to arrive. It is these items which Akamai can help to deliver more quickly.
It works like this. You request a web page in the usual way, and the page description is delivered. But the references to the page’s “heavy” items are modified to fool your web browser into requesting those items from Akamai, rather than from the original web server. Taking account of your location on the network, and given the prevailing traffic conditions, Akamai then delivers the heavy items from the nearest available cache, and the page pops up much more quickly. By monitoring the demand for each item, and making more copies available in its caches when demand rises, and fewer when demand falls, Akamai’s network can help to smooth out huge fluctuations in traffic. A further benefit is that the customer’s web server does not have to deliver the heavy items, which reduces the load on it dramatically and makes it less likely to collapse when faced with a sudden surge of visitors.
A number of firms have followed in Akamai’s footsteps by moving content to the “edges” of the Internet. But there are several other ways to speed up content delivery. One alternative approach is being taken by the Content Bridge Alliance, a group led by a California software firm called Inktomi, whose other members include AOL and Exodus. Rather than setting up a network of thousands of caches, as Akamai has done, the Content Bridge Alliance’s plan is to connect existing networks and farms together more efficiently in order to speed the flow of traffic. Yet another approach is being taken by Kontiki, a firm launched this week by veterans of Netscape. It is one of several start-ups that plan to combine Akamai’s approach with that of Napster, the infamous music-swapping service. Essentially, users’ own computers will be used as caches, so that recently accessed content can be delivered quickly when needed to other users nearby on the network.
Now, Find the Users
In parallel with all this effort to obscure the physical location of data on the Internet, there is growing interest in determining the location of its users. Laws and tax regimes are based on geography, not network topology; online merchants, for example, may be allowed to sell some products in some countries but not others. The growth in interest in “geolocation” services, which attempt to pinpoint Internet users’ locations based on their network addresses, also signals the realisation that traditional marketing techniques, based on geography, can be applied online too. Marie Alexander of Quova, a Silicon Valley geolocation firm, points out that goods and services exist in physical locations, and marketing is traditionally done on a geographical basis. Rather than messing around with fiddly (and privacy-invading) one-to-one marketing, she says, many firms are instead sticking with the old geographical approach, but taking it online. Thus different visitors to a website may be offered different products or special offers, depending on what is available nearby.
Quova’s geolocation service, called GeoPoint, is based on a continually updated database that links Internet Protocol (IP) addresses to countries, cities and even postcodes. If you visit a website that is equipped with GeoPoint software, your IP address is relayed to Quova’s servers, which look up your geographical location. This information is then used by the website to modify the page’s content based on your physical location. Quova claims to be able to identify web users’ country of origin with 98% accuracy, and their city of origin (at least for users in the United States) 85% of the time. Other firms, including Akamai, Digital Envoy, InfoSplit and NetGeo, offer similar services.
Once the user’s location is known, existing demographic databases, which have been honed over the years to reveal what kinds of people live where, can be brought into play. But although targeted advertising is the most obvious application for geolocation, it has many other uses. It can, for example, be used to determine the right language in which to present a multilingual website. E-commerce vendors and auction houses can use geolocation to prevent the sale of goods that are illegal in certain countries; online casinos can prevent users from countries where online gambling has been outlawed from gaining access; rights-management policies for music or video broadcasts, which tend to be based on geographical territories, can also be enforced. The pharmaceutical and financial-services industries, says Ms Alexander, which are subject to strict national regulation, can be confident that by offering goods and services for sale online they are staying within the law. Borders, she notes, are returning to the Internet.
Interest in geolocation soared after last November’s ruling by a French judge requiring Yahoo!, an Internet portal, to ban the auction and sale of Nazi memorabilia over the Internet to users in France. The ruling was significant because it covered sales to French users even from Yahoo!’s websites located in other countries. The implication is that to avoid breaking French law, websites around the world where such items are sold must prevent French users from gaining access — and geolocation technology allows them to do just that. Of course, the technology is far from perfect; a panel of experts, including Vinton Cerf, the networking guru who is known as the “father of the Internet”, advised the judge that determining an individual user’s country of origin was unlikely to be possible more than 90% of the time. But all borders are slightly porous, and the French judge decided that 90% was good enough.
Rather than adopt geolocation technology, Yahoo! responded by banning the auction of Nazi items across all of its sites, and says it has no plans to reinstate them. But it is challenging the ruling in order to avoid having other such restrictions placed on its content by other jurisdictions. The company, which is based in America, has asked a federal court in San Jose to declare the French ruling unenforceable in the United States. (Ironically, Yahoo! said last month that it would begin using Akamai’s geolocation technology to target advertising and other content.)
Critics of the French ruling agree that it would set a dangerous precedent, by allowing one country to interfere with freedom of speech across the entire Internet. “If every jurisdiction in the world insisted on some form of filtering for its particular geographic territory, the web would stop functioning,” Mr Cerf declared. Stanton McCandlish of the Electronic Frontier Foundation, a pressure group, says he expects other governments to adopt geolocation and other similar techniques to balkanise the Internet in coming years. But he notes that geolocation is merely the latest example in a growing trend to impose local controls on the Internet. China, for example, already filters all Internet traffic flowing into and out of the country in order to prevent its citizens from accessing particular websites.
At the same time, the French ruling is regarded in some quarters as a logical and pragmatic way forward for Internet regulation; in the real world, after all, multinational firms are used to operating under different laws in different countries. According to Lawrence Lessig, a Stanford law professor, “the notion that governments can’t regulate hangs upon a particular architecture of the Net.” As the Internet’s architecture changes and becomes more complex, with the addition of services like filtering and geolocation, the idea that the Internet is beyond the reach of local laws and government regulation looks less and less tenable.
The Revenge of Geography
So much for the death of geography. And determining the location of Internet users seems likely to become even more commonplace, and even more accurate, with the rise of wireless Internet devices such as smart phones. Already, the first “location-based services” have been launched, capable of sending text messages to mobile-phone users in particular network cells. More accurate positioning will be possible in future using a number of other techniques, such as the satellite-based Global Positioning System. Advertisers are rubbing their hands at the prospect of being able to send precisely targeted offers to people near particular shops, or inside a sports arena, though privacy concerns may yet scupper their plans. Less annoyingly, users of smart phones may choose to call up location-specific information, such as maps or traffic updates, or to locate a nearby restaurant. According to a recent estimate from Analysys, a telecoms consultancy, global revenues from location-based services will reach $18 billion by 2006 — a figure that is regarded as conservative by many in the industry.
Mr Townsend notes that cities are, in a sense, vast information storage and retrieval systems, in which different districts and neighbourhoods are organised by activity or social group. A mobile Internet device, he suggests, will thus become a convenient way to probe local information and services. Location will, in effect, be used as a search parameter, to narrow down the information presented to the user. Mobile devices, he says, “reassert geography on the Internet.”
At the moment, Internet users navigate a largely placeless datasphere. But in future they will want location-specific information and access to their personal data, wherever they are — and wherever it is. This will be tricky to pull off, and impossible without taking geography explicitly into account. It is undoubtedly true that the Internet means that the distance between two points on the network is no longer terribly important. But where those points are still matters very much. Distance is dying; but geography, it seems, is still alive and kicking.
Copyright © 2001 The Economist Newspaper and The Economist Group. All rights reserved.