Cyberplace and Cyberspace: Two Approaches to Analyzing Digital Intercity Linkages

From the mid-1990s, there has been an enormous growth in the use and diversity of information and communications technologies (ICT), such as global positioning systems (GPS), virtual reality (VR), voice over Internet protocol (VoIP), short message service (SMS), and others. The use of webcam conversation, digital signature, real-time chat, e-mail, and several other digital communication applications are almost everyday practices and are indispensable in business transactions. Traditional activities such as shopping, banking, entertaining, and working are being progressively supplemented by numerous e-applications such as e-working, e-shopping, e-banking, and e-entertainment. Although the effect of this “e-volution” has to be put into perspective, most e-applications complement but do not substitute for traditional activities.

A number of e-products also generate completely new fields of application. For instance, in order to secure or send information, we all put information into a digital format. Castells has pointed out that the world annual production of information in different forms amounts to 1.5 billion gigabytes of which, in 1999, 93 percent was already produced in digital format. The availability of this huge volume of digital information enables us to communicate and derive information “any time, any place.”

The increasing use of the Internet and its applications is also perceptible in the global interrelationships between cities. Transactions of information, communication, and the facility of other intercity linkages are altered by new ICT development. Cities have an important long-term role in the process of globalization as centers of information exchange. The new digital technology strengthens this important central coordination and control function. The mix of central functions applied to societal networks and material and human resources, when combined with the ability to communicate or obtain information at high speed, makes cities into major players in the global economy.

Many research workers studying the various ways in which cities and their linkages can be understood have based their approaches on digital intercity networks.1 The aim of this paper is to gain a comprehensive view of the various methods available for the analysis of “virtual” intercity linkages, that is, a linkage based on ICT and not an undefined relationship. We identify two methods, the cyberplace (CP) approach and the cyberspace (CS) approach.

The CP concept investigates virtual intercity links as actual connections between physical points such as those within the network of Internet cables and their interconnections. This approach assumes that the physical infrastructure represents the digital relationship between places. Most research on digital intercity linkages is based on this approach. On the other hand, the CS concept has to our knowledge not hitherto been employed in the urban network context (empirical analyses such as those of Heimericks and Van Besselaar draw on the cyberspace approach to study the international connections of scientific research). This approach observes intercity linkages starting from the “invisible” structure of the virtual world such as Internet hyperlinks, the structure of search engines, and e-mail traffic. In the cyber world, one communicates in a new kind of space where rules other than those of geographical separation apply because of the anticipated contrast between physical and virtual distance. In this, we consider both approaches in the context of our review.

This paper is organized as follows. The first section presents a general introduction to the literature on transnational urban networks. We then present the analytical framework in which we deal with the CP and CS approaches. Thirdly, we discuss the analyses based on both concepts and discuss the outcome. The empirical analysis based on CP is an additional contribution to existing studies. The CS analysis, on the other hand, is new in urban network research. In the concluding section, we summarize our main findings and propose avenues for further research.2

The Position of the Internetwork and Its Applications within Research on Transnational Urban Networks

The contemporary literature on transnational urban networks can be traced back to two interrelated papers. J. Friedmann was the sole author of one, and with G. Wolff, was the co-author of another. Both texts framed the rise of a global urban system in the context of a major geographical transformation of the capitalist world economy. This restructuring, most commonly referred to as the “new international division of labor,” was premised on the internationalization of production and the ensuing complexity in the organizational structures of multinational enterprises (MNEs). This increased economic-geographical complexity, Friedmann argues, requires a limited number of control points in order to function, and a select group of highly interconnected metropolitan areas were deemed to be such points. The publication of Saskia Sassen's The Global City in 1991 marked a shift of attention to transnational inter-city flows resulting from the critical servicing of worldwide production rather than to its formal command through the corporate headquarters of MNEs. Sassen's approach focuses upon the attraction of advanced producer service firms (providing professional, financial, and creative services for businesses) to major cities with their knowledge-rich environments and specialist markets. In the 1980s and 1990s, many such service firms followed their global clients to become important MNEs in their own right. These advanced producer service firms thereupon created worldwide office networks covering major cities in most or all world regions, and it is exactly the myriad of interconnections between service complexes that, according to Sassen, made way for the formation of transnational urban networks.

Empirical research on transnational urban networks has long remained somewhat underdeveloped because of the lack of appropriate data, a problem which Short et al. referred to as “the dirty little secret of world cities research.” This empirical poverty can, for instance, clearly be seen in Castells' book, which is part of a trilogy that is above all an attempt to reformulate social studies for a global age in which “networks constitute the new social morphology of our societies.” However, when it comes to providing a basic cartography of this global network society, Castells' argument falls short of the conceptual shift he advances: the only actual evidence he comes up with in the chapter on the “space of flows” consists of some limited information on inter-city flows gathered from Federal Express. One can, therefore, only conclude, as Taylor has recently done, that “the evidence [Castells] marshals is mightily unimpressive.” This gap between theoretical sophistication and evidential poverty was, however, not a lacuna specific to Castells' book: it has been a structural feature of research on transnational urban networks because data for assessing such urban networks are in general insufficient or even totally absent.

The basic reason for this problem of evidence is that standard data sources are ill-suited for such analyses. To get an evidential handle on big issues, researchers normally rely on the statistics that are available, that is to say, already collected. But such collection is carried out usually by a state agency for the particular needs of government policy rather than for social science research. The result is that such data that are available have an attributional bias (measurements of administrative areas rather than between administrative areas) and are limited to national territories. Where official statistics extend beyond a state's boundaries, they will still use countries as the basic units (e.g., trade data). Thus, there is no official agency collecting data on, say, the myriad flows between London and New York. The major result has been that “few of the available data reveal anything about the flows and interdependencies” that are at the heart of this body of literature. This leads Alderson and Beckfield to note that in the past, relatively few of the empirical global-city network (GCN) studies “utilized the sorts of relational data necessary for firmly establishing such rankings empirically.”

These data problems have put researchers to work in recent years, and we have, therefore, witnessed a proliferation of empirical studies that explicitly seek to rectify this situation. Researchers have relied on a wide variety of data, but some information sources have come to dominate the empirical research, especially information on corporate organization (e.g., data on ownership links between firms across space) and information on infrastructure networks (e.g., data on the volume of air passenger flows across space). The success of both approaches can, of course, be traced to their commonsensical appeal: the corporate-organization approach acknowledges that well-connected cities derive their status in large part from the presence of key offices of important firms, while the infrastructure approach recognizes that well-connected cities are typified by the presence of vast enabling infrastructures. Put simply: the most important cities harbor the most important airports, while the extensive fiber backbone networks that support the Internet have equally been deployed within and between major cities, hence creating a vast planetary infrastructure network upon which the global economy has come to depend almost as much as it depends on physical transport networks.

Table 1 summarizes the approaches developed in the empirical literature on transnational urban networks through an overview of some key studies in this research domain. The table acknowledges that the basic bifurcation between corporate organization and infrastructure needs to be deepened on the basis of the exact types of firms and infrastructures, and equally shows that all this is in practice somewhat more complicated because of the presence of a limited number of studies that make use of other types of data (e.g., Taylor's analysis of non-governmental organizations) and/or that combine indicators from both approaches. In the next section, we focus on studies that use data on digital connections between cities for assessing transnational urban networks.

TABLE 1
A Taxonomy of Empirical Approaches in Empirical
Global-City Network Research

Indicators	Corporate Organization		Infrastructure		Other and/or a combination of indicators
	Global service firms	Multinational enterprises	Tele- communications	Air transportation
	Examples	Beaverstock et al. (2000)	Alderson and Beckfield (2004)	Malecki (2002)		Smith and Timberlake (2001)	Taylor (2005)
Derudder et al. (2003)		Rozenblat and Pumain (2007)	Rutherford et al. (2004)	Derudder and Witlox (2005)	Rimmer (1998)
Source: Derudder 2006

The Cyberplace (CP) and Cyberspace (CS) Approaches

The Cyberplace (CP) Approach

The cyberplace approach (CP) is an amalgamation of the vital linkages and essential connections within the virtual network among the existing infrastructures of satellite, telephone, computer, facsimile, credit card, fiber-optic cable, and other information and telecommunications systems. In general, it can be defined as the “Internet's physical fabric.” The CP approach makes use of the tangible infrastructure to analyze the “virtual” transnational linkage of cities.

Contemporary economic globalization is constituted by intensive transnational networks of exchange and transaction. Cities contain the junctions in the exchange of communication and information. It is, therefore, in a city's interest to have a well-developed communications network. Several studies have explored the fundamental connection between telecommunications growth and economic success. Sommers and Carlson, for example, document several examples of United States cities in which large companies demanded a developed telecommunications infrastructure before they chose an urban area as a site for operations. The extent and quality of digital telecommunications infrastructure, through which the virtual exchange and transactions occur, is consequently an important measure of digital intercity relationships. The access to new technology is of prime importance in large cities where the largest markets are found and is of secondary importance in smaller places. Most empirical studies on digital intercity flows base their analyses on the CP approach. Gorman and Malecki examine the virtual network of major United States cities, observing, for example, the most extensive fiber-optic cable networks. Grubesic and O'Kelly explore the fiber-optic backbone points of presence (PoPs) and note that these are generally located in major cities such as Atlanta, Boston, Chicago, Dallas–Fort Worth, Los Angeles, New York, Philadelphia, San Francisco and Washington, D.C.3

To implement the CP approach, the first step is to find appropriate data concerning the physical infrastructure that guarantees the digital accessibility of a place. The majority of CP studies start from observation of the cable network that supports intercity digital information flow. CP studies are mostly based on maps and data that are available on the World Wide Web. Global telecommunications companies such as the BT4Group, AT&T, or Verizon give ample treatment to promoting their networks of cables and satellites in order to demonstrate to potential customers how extensive and capable these are Malecki and Boush even state that during the last 10 years, telecommunications firms have spent more money on advertising than on new technology. The Internet architecture is, however, far more complex. Cable networks are not the only elements of the infrastructure. The goal of this analysis is to broaden the existing CP studies to show the importance of including the entire Internet infrastructure in such an analysis. Internet users experience this “network of networks” as a seamless, global, and ubiquitous communication medium directly connecting two points. However, behind the scenes lie many individual networks, owned and controlled by different corporate, institutional, and governmental entities, and joined to each other by various, less-known interconnection arrangements.

Internet users need interconnection arrangements to communicate with one another via computers both next door and on the other side of the globe. The Internet is a network of networks, owned and organized by different companies. Communication takes place in a manner similar to air travel, where, as a consequence of the airlines' organization, intercity passenger flows are restricted to specific locations for their connections. Airline passengers fly from A to B passing through a hub switching point C. In the telecommunications industry, information transmission is also limited to specific locations that structure the network. Interconnection is needed because no single network operator could possibly provide Internet access in every part of the world. Different Internet service providers (ISPs) regulate the network around the block and around the world. In order to provide end users with universal connectivity, ISPs have to interconnect with one another to exchange traffic destined for each other's end users.

The exchange of Internet traffic between networks is called “peering.” Peering is an interconnection of separate Internet networks for exchanging traffic between the customers of each network.5 A bit flowing from point A on the network of Internet Service Provider (ISP) A to point B on the network of ISP B has to change its ISP at an interconnection point C. The physical points that allow various ISPs to peer are called Internet eXchange Points (IXPs). ISPs that want to use the IXP to connect to other ISPs run one or more links from their own routers to the exchange point and connect them to the IXP routers. An impact of interconnection arrangements (IXPs) can be seen in Figure 1 below.

FIGURE 1 Interconnection Points in the Global Internetwork

The early development of the Internet and the deregulation of the long-distance telecommunications market in North America resulted in interconnection arrangements between North American ISPs and other providers that were strongly biased in favor of the former ISPs. Even now, for example, it is common practice for Internet users in Spain to communicate with users in Sweden or Finland over a path that is led through an IXP in the United States. (See Figure 1.) Although there exists a direct link, the time cost of connecting through an exchange point located on the other side of the world is significantly lower than the cost of a direct connection. These extreme examples are being slowly reduced by the growth of interconnection points spurred by the deregulation of the telecommunications industry.

Figure 2 highlights the establishment of new IXPs (i.e., interconnection points) per year in comparison with the total number of existing European IXPs.

FIGURE 2 The Establishment of New IXPs in Comparison to the Total Number of Existing European IXPs

Although the interest for urban scholars is “not the physical pathway but the endpoints of the connectivity package being delivered,” the examples above stress the importance of considering the entire Internet architecture. A number of cities have, in a manner similar to the hubs in the airline network, major advantages in terms of digital accessibility through their function as interconnection points. Direct interconnection through these points avoids data travel to other cities, and potentially to other continents, to get from one network to another, thus reducing latency. They present the ability to switch from a regional to a continental or global network without first travelling on different regional networks. Table 2 presents a number of these interconnection points. The CP approach discussed here is based on such IXPs (results are described in the empirical section below). The first goal of the analysis is to prove the importance of considering the entire Internet infrastructure in a CP approach. The second goal is to introduce IXPs as important CP components to gain an insight into digital intercity linkages.

TABLE 2
A Number of Internet Exchange Points

Abbreviation	Name	Location
AMS-IX	Amsterdam Internet Exchange	Amsterdam, Netherlands
LINX	London Internet Exchange	London, United Kingdom
DE-CIX	Deutscher Commercial Internet Exchange	Frankfurt am Main, Germany
JPNAP	Japan Network Access Point	Tokyo, Japan
Netnod	Netnod Internet Exchange i Sverige	Stockholm, Sweden
ESPANIX	Spain Internet Exchange	Madrid, Spain
JPIX	Japan Internet Exchange	Tokyo, Japan
HKIX	Hong Kong Internet eXchange	Hong Kong, China
BIX	Budapest Internet Exchange	Budapest, Hungary
KINX	Korea Internet Neutral eXchange	Seoul, Korea

Studies of telecommunications hubs have been performed from many different perspectives, including the technical architecture of exchange points, the business and economic models that underlie peering, transit agreements, and so on. Grubesic and O'Kelly analyze the geographical position of PoPs in the United States6in order to examine the digital accessibility of cities. PoPs give information about the location of the access/switching points per ISP. Thus, while the IXPs do serve for many ISPs, the PoPs7tell something about the switching location of a single ISP. Examining IXPs gives centralized information about the Internet traffic passing through an exchange point. Other researchers note IXPs as important digital points in city networks but do not give more detail.

The IXP data used here are drawn from the Euro-IX report. This was compiled by the European Internet Exchange Association (Euro-IX) and gives information about the amount of traffic being transferred in IXPs that are members of the European Internet Exchange Association.8 A substantial proportion of IXPs publish traffic statistics on their websites in order to promote their interconnection point to potential new ISP members. Amsterdam, ams-ix.net (AMS-IX) for instance, gives information as daily and yearly traffic load graphs and monthly reports about their traffic volume, total multicast, IPv6, and broadcast traffic, ISP members, technical and statistical information, topology, and so on. The Euro-IX report has accumulated these graphs for cities in Europe. The results of the empirical analysis based on these data are described below in the CP approach section of the empirical analysis.

The Cyberspace (CS) Approach

In addition to the traditional geographical space wherein the CP is embedded, we have witnessed the emergence of a new approach for analyzing digital flows: the CS approach. CS is the virtual world wherein people communicate with each other using computer systems. The CS concept makes use of this new space to analyze digital intercity linkages. Because the functioning of several communication and information technologies such as e-mail, web consulting, and video conferencing, are based on CS, such an approach can give a refined picture of existing virtual linkages between cities. To introduce this idea, we discuss the most important features of cyberspace.

Although it is often referred to, cyberspace is quite a difficult entity to grasp and define. It has been variously described as follows: “a consensual hallucination experienced daily by billions of legitimate operators, in every nation, by children being taught mathematical concepts”; “unthinkable complexity”; a “parallel universe”; a “multi-media skein of digital networks which is infusing rapidly into social, cultural and economic life”; “a space that is difficult to comprehend and mentally visualize … in which it is easy to get lost and confused.”9

A more concise definition of CS is given by Leadbetter, who describes this space as “an immaterial world of computers and communications, in which we can work at the touch of a button.” This definition highlights the three most important features of cyberspace:

•	It is all about computers and virtual communication. CS is the virtual network wherein computers communicate with each other.
•	CS is an immaterial world in that the physical aspects of the material world involving distance do not apply in this new space.
•	In this immaterial world of computers and communication, we work “at the touch of a button.”

Distance has a diminished importance in the cyber world. A place in CS is close by when it is just “one click away” or “at the touch of a button.” The need to click a lot to get on to a website that is actually physically nearby is perceived as its being more remote and less accessible than a site that, just one click away, is physically more than thousands of miles away. The time cost is so small that virtual access is practically invariant with physical distance. Access in virtual space, therefore, follows logical links rather than physical paths. The CS approach uses these features to analyze the virtual linkages of cities.

We discern two types of analysis applicable to the CS approach: a content-based analysis (CBA) and a structure-based analysis (SBA). The SBA is based on the CS structure seen as the hierarchy of websites, connections via hyperlinks,10 and so forth. The CBA examines intercity linkages using the information available on the pages of the World Wide Web.

The most examined SBA, hyperlink analysis, is based on the logical links that connect websites. Park and Thelwall (61) promote this type of SBA in the belief that with the increasing importance of the Web for an ever-broader spectrum of human activities, the structure will reflect more and more the existing relationships between people, cities, institutions, and so forth. A hyperlink analysis (i.e., an SBA) is an analytical method for studying the networked (or connected) structures on the World Wide Web. A hyperlink “transmits” the information or contents of a web page at the touch of a button. It is a technological capability that enables one website to connect seamlessly with another and, therefore, functionally brings the two sites closer together. Putting a hyperlink on your website implies the possibility of a jump from one location/actor to the other and indicates the implicit presence of other locations/actors such as firms, people, or information in the immediate vicinity. The use of hyperlinks in the creation of a web structure belongs to the designer, who, therefore, controls the potential ways a user can move through the Web.

Commercial firms are supposed to be cautious in placing hyperlinks. Any link to an external site is an extra exit route for visitors. A natural reaction may be to avoid hyperlinks, thus creating an isolated self-contained structure. Thelwall shows, however, that 72 out of 232 sites (or 31 percent) of commercial firms were found to have hyperlinks based on affiliated business relations. For service reasons, commercial sites offer links to complementary sites. Their intent is to build a trust-based customer relationship. Thus, while website creators have complete freedom, hyperlink structures are designed, sustained, or modified to reflect relationship choices. Park and Thelwall note that from these communicative choices and agenda, we can discern the fingerprints of social relationships between the system components (i.e., people, private companies, public organizations, cities, nation-states, and so on). An example of a hyperlink analysis is given by Heimeriks and Van den Besselaar. They analyze hyperlink networks on the scientific web in order to study the development of research fields and the relationship between research organizations and the relevant institutions in their environments. Before the digital age, joint papers in journals and academic contacts at conferences were important measurement tools. Nowadays, the Web gives us additional academic network information.

Hyperlinks between websites represent not only direct relationships in the “offline” world but also other and desired relationships. Hyperlink analyses can be drawn up for several layers. The underlying belief for the CS approach is that collaboration and information exchanges between cities are reflected in the hyperlink networks and/or in the number of hyperlinks to urban components. Whatever the importance of hyperlinks for cities or urban components, we should nevertheless consider that “there are some types of areas of the web for which links are important and others where they are not, both accounting for a significant proportion of the web.”11

To implement the cyberspace approach in this paper, we have opted to base the empirical analysis on web information, i.e., a content-based analysis.12 The World Wide Web (WWW) is the most up-to-date and most global information source. The relationships between cities must be traceable in this large database. To trace these relationships, we make use of web search engines that draw on robots or “spiders” to index the huge amount of information. Spiders are computer systems that continuously download web pages and go through the published information. Depending on the search engine, all or a part of the published information, such as key words, links, and so on, is saved in a large database. The spiders continuously pursue the search drawing on the saved links. The final search result matching the query (i.e., number of presented pages and rank order) depends on the index decisions. More than 150 criteria exist to determine the relevancy of pages. PageRank and link popularity are the most important techniques applied by Google. PageRank is a technique that, according to the web search engine Google.com, relies on the “uniquely democratic nature of the web by using its vast link structure as an indicator of an individual page's value.” In essence, Google interprets a hyperlink on a webpage as a vote, by this page, for the linked page. When the page that casts the vote is itself “important,” it weighs more heavily. Using these and other factors, Google provides its views on a page's relative importance. So far, we may conclude that the most trafficked parts of the Web are indexed by search engines. Querying the number of joint appearances of two city names will give us the relative proportions of actual intercity linkages according to the web search engine. For this empirical analysis (results are described in the second part of the next section), we base our search on the currently most popular search engine, Google.13

Taken together, of the two possible methods, CP and CS, that can be applied to analyze the virtual intercity linkages, one might say that the CP approach analyzes the virtual intercity linkages via the real world, whereas the CS approach observes linkages in a “parallel” world. Although both approaches may seen similar at first sight, they start from completely different observations of virtual urban linkages and must, therefore, be analyzed separately.

Empirical Analyses

The Cyberplace (CP) Approach

Figure 3a shows the upward trend in peak aggregated IXP traffic (in Gigabits per second, Gbps) per European city (based on the Euro-IX report). We compare this graph with Figure 3b, which presents the estimated bandwidth (in Gbps) per city.14 The digital bandwidth capacity data used in Figure 3b is based on the European Terrestrial Networks map.15

FIGURE 3a The Peak Aggregated IXP Traffic (in Gigabits per second) per European City

FIGURE 3b The Estimated Bandwidth (in Gigabits per second) per European City

Comparing both graphs gives an idea about the importance of European cities in the digital city network.16 One major feature is the divergent trend of the graphs. Whereas the bandwidth increases gradually per city, the exchange of Internet traffic is far more skewed towards a limited number of metropolitan areas, i.e., Amsterdam, London, Frankfurt, Madrid, and Stockholm. These cities are responsible for 72.26 percent of the European Internet exchange traffic. Amsterdam is the overall number one with 28.41 percent of total European IXP traffic. These cities function as key points in the digital intercity network. They significantly expand the reach of the virtual network.

A large amount of exchange traffic in a city points to its important switching function between regional, national, and global information communication networks. One would therefore expect a17large amount of IXP traffic to correspond to abundant bandwidth availability. An important switching point has to transfer its traffic to other places such as Frankfurt, London, or Amsterdam. A city such as Budapest exchanges much traffic but is less important in terms of bandwidth availability. This probably results from a high regional exchange of traffic and a lesser global traffic requirement.18 The IX point located in Budapest boosts the surrounding region in terms of Internet traffic possibilities. To consider this phenomenon in more detail, we look at Figure 4. This presents the relationships that cities have in terms of equal service providers.19 The relationship between cities is seen as the number of ISPs that peer in both cities. This means that service providers are active in both the cities shown and are consequently more global in reach. The volume of the nodes is based on the number of ISPs peered at the IXP.20 The most important exchange cities also control the most jointly peered ISPs. The “big three” of Amsterdam, London, and Frankfurt, with London first, hold almost all the ISPs that peer in the European Internetwork. These exchange cities function as global switching points. Cities such as Lyon, Florence, Budapest, and Ljubljana on the other hand are more regional exchange points. A small number of global ISPs link these IXPs with a more global exchange point such as Lyon–Paris, Ljubljana–Vienna, and Florence–Milan. The difference between more global exchange points such as Amsterdam, Paris, Zurich, and Stockholm and the regional IXPs such as Ljubljana, Lisbon, and Florence can be observed by the number of shared ISPs as compared with their individual ISPs.

FIGURE 4 The Relationships of Cities in Terms of Equal Service Providers

The global exchange point of Amsterdam has 80 percent of its ISPs peering at other ISPs (Stockholm, 83 percent; Zurich, 82 percent; Paris, 90 percent), whereas there are fewer ISPs peering at other IXPs for Ljubljana (17 percent), Lisbon (16 percent), and Florence (50 percent). Budapest, as mentioned above, is a more regional exchange point. Eighty-four percent of the ISPs in Budapest peer only at that city. This difference can also be observed from the position of the IXPs in Figure 4. Two cities situated on the edge of this figure are more regional exchange points linked with one or two more global exchange cities. There are, furthermore, internal differences between the regional IXPs. Warsaw and Athens are, for example, less linked to the global ISP network than Florence or Ljubljana. The three shared ISPs for Athens are the global service networks of AT&T Global Network Services, Viatel Global Communications B.V., and Verizon Business, whereas the shared ISPs for Florence and Ljubljana are more regional ISPs such as Florence–Milan with Telecom Italy. It is interesting to note that Brussels has less IXP traffic but profits from its geographical position between the global exchange points of London, Paris, Frankfurt, and Amsterdam. Brussels is situated at the crossroads of European and global service providers AT&T, Deutsche Telekom (T-Online), France Telecom/Opentransit, United Global Com, and KPN Nederland. Building an IXP in such a city boosts its virtual network reach.

The geographical position of Brussels results also in a high amount of bandwidth availability. (See Figure 3b.) The more regional exchange cities such as Budapest have low intercity bandwidth linkages. A further comparison between bandwidth capacity and exchange power can be observed in Figure 5. In this figure, we compare the logarithm of exchange traffic per city with the estimated digital bandwidth capacity.21 Four groups of cities can be discerned.22 The first group, with the most bandwidth and IXPs, are Amsterdam, Frankfurt, London, Brussels, Düsseldorf, and Paris. These centers receive a large amount of traffic from their feeding cities and transfer it to the global network (and vice versa). A second group, ranked slightly below, consists of Zurich, Munich, Madrid, Stockholm, Vienna, and Nurenberg. A third group—Prague, Malmo, Oslo, Torino, Bratislava, Dublin, Manchester, Barcelona, and Bilbao—is important in terms of bandwidth but less important in terms of exchange traffic. A fourth group—Budapest, Helsinki, Gothenburg, Warsaw, Rome, Tallinn, Lisbon and Bucharest—is more important in terms of IXP traffic than the available bandwidth capacity. These cities are regional IXP cities.

FIGURE 5 The (Logarithm of) Exchange Traffic in Comparison with the Estimated Digital Bandwidth Capacity per European City

Increasing the number of clusters to 10 puts the relative importance of IXP traffic versus bandwidth availability into the picture. The first group is, for example, split up in three groups: (1) Paris and London, which are the leaders in terms of exchange traffic and digital bandwidth availability; (2) Amsterdam and Frankfurt, which are foremost in terms of exchange traffic but less important in terms of bandwidth; and (3) Brussels and Düsseldorf, which are important in terms of bandwidth but less important in terms of exchange traffic. The other groups can be split up similarly.23

The empirical analyses indicate the importance of increasing the number of structural elements for a CP approach. Combining IXP data with information about the bandwidth capacity tells us a lot more than just the cable network information: it gives a more general picture of the network of intercity flows.

The Cyberspace (CS) Approach

The content-based analysis was set up in January 2008. We assembled the city-pair relationships between 40 European cities.24 For example, searching for (Google) web pages that jointly mention “Amsterdam” and “Kiev” resulted in 596,000 different web pages with data on, for example, the transport, business, history, and governmental relationships between both cities. Figure 6 presents the most important of these relationships, with more than 1 million results in the 40 × 40 intercity matrix.25

FIGURE 6 The Digital Intercity Linkages According to Google.com (January 9, 2008)

The search was made without any preferences or specific language choice in Google (http://www.google.com/advanced_search). City names are spelled in English, which is an underestimation of regional city linkages. English is, however, not only the dominant language on the World Wide Web in and by itself, but especially for linking. Most web pages have next to their “national language” page an international page in English.

The size of the nodes varies according to the total number of results. Paris, London, and Berlin are the most important cities on the Web. They have a high number of joint appearances on the Web and maintain thereupon their important relationships with almost all European cities. London represents the most dominant place on the Web. We can also observe some prominent national connections between cities; for example, Hamburg–Berlin, Birmingham–Manchester, Madrid–Barcelona. Furthermore, it is interesting to note the linkages of cities such as Bucharest, Düsseldorf, and Bratislava. These places are highly linked with secondary centers such as Copenhagen, Dublin, and Lyon. A strong link between these cities and London does not exist. They are located outside the strong central network of London, Paris, and Berlin and can be seen to be less digitally accessible.

In order to compare the differences between two time periods/search engines, we did two other Google search queries in April and November 2007, and a Google and AltaVista search query in November 2007. The continuous update of search engine databases results in a higher number of related documents for all city linkages queried in November 2007 with 8.4 percent more linkages found. There existed also a minimal rank difference with correlation between both searches being 86 percent. To compare the difference between web search engines, we performed the same analysis based on the AltaVista web search engine. Here, we also observed a minimal difference.

The CS analysis is the first investigation using the content-based analysis method. Using a so-called meta-search engine such as MetaCrawler, MetaFind, or SurfWax or a topic-based search engine such as, for example, business.com, news.google.com, or newslink.org, it may be possible to map more detailed digital intercity linkages. A meta-search engine transmits the search query simultaneously to several individual search engines such as Google, AltaVista, or Yahoo and their databases of web pages. Within a few seconds, the results from all the search engines queried are presented. A CBA based on this kind of search engine may, therefore, give more general results. Topic-based search engines, in contrast to the meta-search engines, may give us more accurate and relevant information.26

Discussion and Main Conclusions

This article has provided an overview of the dominant approaches for understanding intercity links through digital flows. Most analyses to date on digital intercity relationships are based on what we have dubbed the CP or cyberplace approach. The World Wide Web that we have called the CS or cyberspace approach, however, equally transcribes digital relationships. With the growing importance of the Web for an ever-broadening spectrum of human activity, the World Wide Web increasingly contains digital linkages between people, cities, institutions, and other entities. Examination of the CP and the CS approaches together points out the dissimilarity between these methods. Although both approaches deal with digital intercity relationships, the CP and CS “worlds” or layers are derived from completely independent observations and must, therefore, be discussed separately.

The CP approach is based on physical layer observations of cables, servers, interconnection points, and related apparatus. Although this approach is discussed in the literature at great length, a comprehensive analysis of the physical digital intercity flows is indispensable. The analysis presented here points out the lack of investigation into structural CP elements. The flows of bits are structured by different corporations and follow routes other than the direct, predictable connections between end users. In order to comprehend the cyberplace, it is important to study the entire Internet architecture. Airline hubs or gateways influence the way airline companies build up their networks, yet digital flows are structured by gateways or interconnection points. For instance, maps showing the network of undersea cables tell us something about the speed at which digital information is transmitted overseas. Being located at an appropriate place in this network means that you have the ability to download and upload information rapidly and to maintain strong digital communications across oceans. The existence of exchange points forces the importance of these linkages. Digital gateway cities are able to transfer regional traffic to other interconnection points overseas and to distribute global traffic in their own region.

A comparison of digital bandwidth capacity with IXP traffic has led so far to a more refined picture of intercity linkages. We observed that in the European context, Amsterdam, Frankfurt, London, Brussels, Düsseldorf, and Paris share the highest digital bandwidth capacity. The digital weight of cities such as Paris, Brussels, Zurich, Munich, and Düsseldorf is, however, overestimated when only bandwidth capacity is observed. Amsterdam, London, and Frankfurt are far more important in their gateway functions than Paris, Brussels, and Düsseldorf. A combination of both CP data sources shows at a glance that Amsterdam, London, and Frankfurt are the top digital cities in Europe. It is likely that this hierarchy in digital importance will continue to shift, categorically and geographically, in the future. Continued infrastructure investment in cities such as Budapest will increase the digital possibilities of those regions and create significant regional gateways. Global gateways such as Amsterdam, Frankfurt, and London will increasingly serve at the global level.

It is important to note that CP analyses rely on recent data. The market for interconnection is changing rapidly. In less than five years, the number of European interconnection points has more than doubled. The Euro-IX (23) report is the first account available online concerning European IXPs, and more research about these data is required. The Internet infrastructure is constantly growing, and although the boom in cable infrastructure is already past, some regions are just beginning to develop cable networks based on modern technology. Examination of the entire Internet architecture will broaden our knowledge of digital intercity linkages.

The CS approach provides insights into intangible digital flows, which must be observed in the cyberlayer. Measurement of hyperlinks, e-mail contacts, or search results gives a notion of the existing relationships via the digital web. The CS exercise herein is a first investigation of a content-based analysis on intercity linkages.27The measured relationships engage airline connections, governmental agreements, family ties, firm locations, and so on, in several different formats, for example PDF, TXT, and URL. It is one of the most up-to-date existing data queries. Although the World Wide Web is constantly changing, search engines provide daily updates. At the end of January 2005, the indexed web contained more than 11.5 billion pages. Querying this huge amount of information by web search engines is the most general search on digital linkages that can be made. It provides information that is not accessible by a traditional metrical approach.

The content-based analysis (CBA) here shows that Paris, London, and Berlin collectively have the most important digital linkages on the Web. The dominance of North-west Europe on the Web is obvious. Amsterdam, Rome, and Frankfurt are secondary cities. Looking at the intercity relationships, we observe a number of strong regional linkages such as Lyon–Paris, Rotterdam–Amsterdam, and Madrid–Barcelona. However, we have to bear in mind that the CBA presented also has some shortcomings:

First, in addition to the surface World Wide Web there is also an immeasurable private web accessible only to banks, firms, and confidential organizations and a deep web, publicly accessible but not reachable by search engines.28 Search engines index only a proportion of current web pages.

Second, the applied search engine in a CBA has a great influence on the results. Using meta-search engines or topic-based search engines, for instance, would give more detailed information.

Third, we have to be aware that the CBA herein actually measures non-digital relationships saved in digital format. Undertaking a structure-based analysis is a next step in this context. Hyperlinks represent desired, existing, or obligatory relationships between website owners that are institutions, organizations, or firms. The SBA has already been widely examined in bibliometric studies. The investigation of structural intercity relationships on the Web is an avenue for further research.

This paper shows that two approaches can be used to measure digital intercity linkages. Both methods analyze different relationships digitally. The broader exploration of both empirical analyses will probably cause us to arrive at more refined methods for the study of digital intercity linkages. In our future research, we intend to explore new techniques of applying the CS approach, and we will gather other CP components for this approach.

Acknowledgements

We would like to thank the editor and the anonymous referees for their useful comments on an earlier version of this paper. The usual disclaimers apply. This research work is funded by the Research Foundation–Flanders.

Notes

Mokhtarian et al. Weltevreden

Castells 2001

Townsend 2001

Sassen 2002

¹ A good overview of the literature on “cybercities” is compiled by Graham (2004) .

Choi et al. 2006 Gorman and Malecki 2002 Moss and Townsend Rutherford et al.

² The neglect of the cyberspace approach compared to the cyberplace approach for the empirical analyses of digital urban linkages could in all probability be attributed to the fairly intangible character of the CS medium in contrast to the physical CP. Hillis even attributes this invisibility as the “key to grasping geography's relative disinterest in communications.” Communication flows are less studied than other flows in geography (Hillis, Kellerman). The confidentiality cited by private telecoms contributes to this lack of data and information.

Sassen 1991 Sassen 2001

Castells 1996

Taylor 2004

Taylor 2004

Knox

Derudder

Rutherford et al.

Taylor 2005

Beaverstock et al.

Gorman and Makecki 2002

Hanley

Graham and Marvin Hackler Moss Walcott and Wheeler

Malecki 1999

Choi et al. 2001 Choi et al. 2006 Gorman and Malecki 2000 Grubesic and O'Kelly Malecki and Gorman Malecki and Boush Moss and Townsend Rutherford Rutherford et al. Townsend 2000 Townsend 2001 Wheeler and O'Kelly

Gorman and Malecki 2002

³ A Point of Presence (PoP) is an access point to the Internetwork.

⁴ These are also used by “Internetwork” engineers to monitor and control traffic flows and network performance. (Dodge and Kitchen 2002 ). Detailed network monitoring maps and tools are, however, generally not made public for reasons of security and commercial confidentiality.

Gorman and Malecki 2000

Derudder et al. 2007

⁵ ARPAnet was, for instance, at the outset in 1969, not an “Internet” but a network of computers. At the time, the U.S. Defense Department contractors permitted the use of the ARPAnet by other U.S. Government agencies to develop similar networks, and they had to interconnect and thereby developed the real Internetwork. (Chapin and Owens)

Kellerman

Akamai Technologies

European Internet Exchange Association (Euro-IX)

Townsend 2001

Kende Norton

⁶ PoPs are housed in an IXP or in data centers that are used to house the server, storage, and network infrastructure of ISPs.

⁷ For an overview of the geographical position of switching facilities in the United States (2001), see Malecki and McIntee, and in Europe (2001), see Kellerman.

Drewe Gorman and Malecki 2002 Townsend 2001

⁸ On October 2006, Euro-IX had 36 member IXPs from 21 European countries, 7 IXPs from Japan and the United States, and four patrons from the switch vendor community. Although the Euro-IX report attempts to list all known IXPs in Europe, it is expected that a small number of IXPs may have been left out. The IXP data do not take into account the traffic exchanged by privately interconnected participants whose traffic does not pass over the IXP switching fabric (European Internet Exchange Association).

Gibson

Benedikt

Graham 1998

Dodge and Kitchen 2002

⁹ This elusiveness has resulted in a particular wording. Virtually all cyberterms (even the notion “cyberspace”) are described by standard spatial and territorial metaphors: e.g., web site, chat room, mail box, portal site, teleport, and so on (Adams, Smith and Timberlake). Kellerman pointed out that “the geographical language has become a major tool for the structuring, organization, and use of cyberspace.”

Dodge

Mitchell

¹⁰ There are in fact two approaches to hyperlink analysis: webometrics (e.g., counting hyperlinks to websites, see for example Smith and Thelwall, Thelwall and Vaughan) and hyperlink network analyses (e.g., social network analysis applied on hyperlink networks; see, for example, Rogers and Marres, Krebs, Park and Thelwall). Here we consider both analyses in common.

Pirolli and Card

Jackson

Siegel

Grossman

Matheison

Thelwall

¹¹ According to Park and Thelwall, little is known about the validity or correct interpretation of hyperlink analyses results except from academic domains; i.e., bibliometrics (Tang and Thelwall).

¹² The above-mentioned structure-based analysis (SBA) is not worked out in this paper. An empirical analysis based on the SBA-method calls for an extensive, paper-length discussion, and hence falls outside the scope of the present paper.

Brin and Page Mostafa

Google

Austin Seopage.nl

Marketshare.hitslink.com

¹³ We used www.googlebattle.com/godin.php, which enabled us to query several city pairs at a time.

¹⁴ Information on the amount of digital traffic through these infrastructures is generally unavailable (Kellerman).

Telegeography.com

¹⁵ The European Terrestrial Networks map illustrates the total lit bandwidth traversing through 145 European cities (above 0.5 Gbps). The 145 cities exist as the top 60 European cities, as ranked by access to lit bandwidth, and 85 other important localities (major city, capital city, etc.). (Telegeography.com)

¹⁶ We have to bear in mind that we compare capacity with real traffic where the former represents long haul links and the latter local/national/international exchange traffic.

¹⁷ We expect that a large amount of bandwidth availability is related to a high amount of exchange traffic. The long haul links are market-based decisions and probably correspond the need to exchange a high amount of traffic (and vice versa).

¹⁸ As mentioned above, the IX point exchanges also traffic on the city/country level and prevent the use of expensive long haul links to other cities to switch from local ISP

¹⁹ This figure is based on the peer-matrix of the Euro-IX report. Only the relationships of three or more equal-peered ISPs are drawn.

²⁰ Figure 4 is constructed using the lay-out technique “spring-embedding” (UCINET (Borgatti et al.)).This technique is based on the concept of a system in which N cities are connected. According to the volume and the number of flows of a city, the intercity links have different strengths. From the initial arrangement of the city locations, the system oscillates until it stabilizes at a minimum-energy arrangement resulting in a dynamically balanced system. (Castillo and Sim)

²¹ Note that there is a time difference between the two different CP datasets. Telegeography's data are from 2002 while the IXPs data refer to 2006. However, since there is a serious lack of data in the cyberplace literature, we argue that using (and comparing) these data sources outweighs the deficits due to the time difference.

²² The clusters are based on the hierarchical Ward and K-means method.

²³ The variation in IXP-traffic decreases at the tail of the distribution (see Figure 3a ). This makes it difficult to make solid conclusions for cities such as Tallinn, Lisbon, and Bucharest. The expansion of IXPs in the future will probably lead to more transparency.

²⁴ The 40 cities rely on the “world city” list of the Globalization and World Cities (GaWC) Study Group and Network.

²⁵ Figure 6 is constructed using the lay-out technique “spring-embedding.” (Borgatti et al.)

Internetworldstatistics.com

²⁶ Additional difficulties of meta-search engines are the use of complex search terms based on Boolean logic (U.C. Berkeley library) and the information required (i.e., number of related documents). These are challenges for further research.

Park and Thelwall

²⁷ “Indexed” means the part of the Web that is searchable through search engines. Estimating the size of the whole Web is quite difficult, because of its dynamic nature. Three sources for tracking the growth of the Internet are “Search Engine Showdown,” “Search Engine Watch,” and World Wide Web size.

Gulli and Signorini

²⁸ Private networks account for around 14 percent of Internet traffic (McClelland). Besides the inaccessible private networks, there are also web pages that are publicly accessible but not reachable by search engines (i.e., the “deep” or invisible web). This part of the web consists of web pages with no hypertext links to their content, pages you need to be registered for and logged-on to access, documents with a content that is dynamically generated (based on constantly changing databases) and other pages difficult to index by search engines.