home page about us documents miscellaneous sitemap
home page about us documents miscellaneous sitemap

Topological Ordering in Cyberspace
  EASST '98 General Conference, Lissabon, September 30th - October 3, 1998

Jeanette Hofmann , 10/98

  jump-off point
1  Ordering telecommunicative spaces
2  Ordering cyberspace
3  III
4  IV


  The purpose of addresses is to localize and identify things and people. At the same time, however, addresses generate a spatial order. My talk revolves around the topological order created by Internet addresses. How are things located in the immaterial world of cyberspace? And related to this question: What is space? How do we imagine space?

"Could space be nothing more than the passive locus of social relations, the milieu in which their combination takes on body..." asked Henry Lefebvre as early as 1974 in his book, 'The production of space'. Of course, his answer was 'no'. Spaces are more than passive loci of social relations. Anthony Giddens seemed to continue this thought a couple of years later: "Most social analysts", he wrote, "treat time and space as mere environments of action and accept unthinkingly the conception of time, as mensurable clock time..." (Giddens 1984, 110)

As if Einstein had never existed, space and time used to be regarded as stable, uniform forces that constrain social development, to be sure, but remain beyond the reach and influence by society. Henry Lefebvre and Anthony Giddens were among the first social theorists to challenge what is today dubbed as the "container model" of space: An empty sphere to be filled with objects and actions without ever being changed by what is done within. Both authors argued that time and space should be treated as social products rather than independent, neutral categories. As Giddens put it:

"The phrase might seem bizarre, but human beings do 'make their own geography' as much as they 'make their own history'. That is to say, spatial configurations of social life are just as much a matter of basic importance to social theory as are the dimensions of temporality." (Giddens 1984, 363)
For Giddens, the dominant pattern of change of modern society can be described as "time-space distanciation"; a "stretching" of society over time and space, caused by the development of transport and communication technologies. Today, we are able to communicate with absent others. Thanks to telephone, TV and computer networks, we can be present without moving our bodies. Presence in time is decoupled from presence in space. The dissociation of physical places from social activities generates what Giddens calls the "emptying out" of space. The disappearance of bodies and things in space creates "a single world where none existed previously", (Giddens 1991, 27). It is a world not longer characterized by the "presence of others and things" but by their absence, as Peter Burgess writes.

In the meantime, Giddens' concept of social space has been criticized for

  • not being relative and constructivist enough,
  • for sticking to the traditionalist notion of one space instead of multiple spaces and so on.
However, he paved the way for the idea that communication does not take place within a pre-given space but constitutes various forms of space - spaces, that are not simply empty, but structured by so-called "spatial practices" or "topological operations".

Spaces generated by communication technologies have a topological order. This is true even for Giddens' emptied out, dislocated or virtual spaces. In my talk I'd like to illustrate this thought by presenting some of the basic ordering procedures that can be found in the organization of cyberspace. The Internet's communication technology has generated a global space, the survival of which crucially depends on the maintenance of precise, unambiguous topological orders. As it turns out, the constitution of the virtual electronic world doesn't confine itself to one space, but rests on several interacting spatial orders. The Internet engineers refer to the rules and results of spatial ordering as "architectures".

Before I get into the details of the Internet's ordering procedures, I'd like to take a step back and show - using zip codes and telephone numbers as an example - that and how communication technologies constitute spaces.

1 Ordering telecommunicative spaces
  Both the telephone and the mail system rely on their own infrastructure. In both cases, topological maps of the communication territory form an important part of that infrastructure. However, such maps should not be confused with the common geographical depictions of territory. The territories of the mail and telephone service do not consist of cities, villages, streets and rivers; instead, they consist of numbers - numbers that tell us about the identity of objects and where they are - numbers, in other words, that form addresses. Within telecommunicative spaces, all objects or agents are identified and located by numbers.

Numbers or addresses are the basic element of the topologies formed by telecommunication technologies. Each communication system generates its own spatial order. The topology of zip codes, for example, doesn't match that of the telephone numbering plan. The reason for the variety of topological orders to be found in Gidden's empty spaces has to do with the circumstances that brought them forth. Zip codes were invented to automate parts of the sorting of mail while telephone numbers were introduced to substitute human operators. The spatio-numerical orders of both systems reflect the administrational procedures and the locations linked to those procedures that they were meant to substitute.

Both, zip codes and telephone numbers generate hierarchical spatial matrices in order to arrange its objects in a suitable, manageable way. The resulting topologies are no natural phenomena, they do not mirror a given physical territory, but offer their own version of a communication landscape. Spatial orders of telecommunication systems are no pure inventions though. Rather, they present itself as variations or further developments of already existing topologies. Most prominent among such spatial models are geo- political entities such as the territorial state and its subdivisions like districts, counties and so on. Spatial orders created by postal and telephone services tend to reinforce such spatial configurations.

2 Ordering cyberspace
  Like empty space produced by traditional communication systems, that of the Internet needs some sort of spatial order. Network objects, the so-called nodes, need to be arranged systematically so that they can be located and routes be established between nodes intending to communicate. Indeed, only objects that possess a meaningful, syntactically correct address can be reached on the Internet. And Internet addresses are meaningful and correct only within a system that relates objects to each other in a topology-generating manner.

However, the constitution of cyberspace differs from that of traditional telecommunicative spaces in two respects. First, it is not, or at least far less bound to physical territories. Cyberspace as generated by the Internet is designed as a global realm without any notion of national boundaries or geographical distance. On the contrary, due to the speed of data transmission, spatial distance seems to be transcended in cyberspace. No matter where on the planet we log in, the distance to other netizens seems to be always the same. Second, (and directly related to the first point), the spatial constitution of cyberspace has to manage without (most of the) authoritative resources available to and involved in the topological ordering of traditional telecommunication services. There is no global government or administration that could be asked to impose on the inhabitants of cyberspace a universal, comprehensive addressing system like that of the telco or postal world. While the latter have always drawn on state authority to develop and successfully establish supranational orders of space, the decentralized Internet proves to be fairly immune to traditional forms of political authority.

What then does the spatial order of cyberspace look like? How is it organized?

  In the early days of the Internet, its developers didn't even think about topological ordering. No address allocation plan was made, and no attention paid to the immense scaling problems of a topology with global extension. Instead, Internet addresses were treated as names only, that is, as a set of symbols or bits to identify a network, but not to locate it. Accordingly, Internet addresses were assigned sequentially. Their allocation followed a chronological order: lowest numbers were given to those who came first. This is why until a couple of years ago, Internet addresses did not convey any information about the actual geographical location of a network. (If the network you are hooked up to got its Internet address before 1994, it still gives no hint about where it is.)

Imagine your telephone number or your zip code wouldn't reflect the physical location of your telephone or mailbox, but the date when you got your phone or mailbox! Telephone numbers would be assigned worldwide without any relation to continents, countries and regions. This is exactly how Internet addresses were allocated until 1994. As a consequence, the identification and the actual location of nodes formed two entirely independent topologies. On the one hand, there was a logical space consisting of several million addresses that served as identifiers or names of Internet nodes. On the other hand, there was a second space consisting of physical addresses that point to a concrete wire to which networks and nodes are attached.

No direct relationship existed between these two spaces. In other words: If you knew the name or identity of a node, you still had no idea where it is located and vice versa: it was impossible to conclude the identity from the physical location of a node. (If you think once again about the telephone numbers system, you always find a combination of locating and identifying properties: while the first digits point to the location, the last digits form the name of the object.) In order to allow Internet nodes to communicate with each other, a third space is needed - a space able to bridge the topologies of physical and logical objects.

This third space is the routing space and by far the most interesting. It shows the disposition of things in cyberspace from the perspective of routers. The routers' job is to direct the data flows from the source to the destination address. Therefore, routers need an idea of what is where.

  • Where in cyberspace means: how can it be reached.
  • How means: through which nodes.
The information about what is where is stored in tables maintained by all routers. Routing tables show cyberspace from each single router's point of view. These tables are, by the way, the only maps available of the topological arrangement of networks and other things in cyberspace. General maps that would provide a kind of bird's eye view of networks and traffic routes in cyberspace are completely impossible to draw. The problem of portraying the spatial ordering on the Internet is the ephemeral nature of the object. How network objects can be reached, in other words: where they are, is a transient matter, that can change every minute. A broken router or network, for example, makes paths invalid and provokes changes that can affect large parts of the routing topology. Topology in cyberspace consists of a flexible, permanently changing order. Information about this order (which ususally takes on the form of standardized messages between neighboring routers) is exchanged among neighboring computers every other minute or so. Thus, a fugitive phenomenon such as the topology of cyberspace could not be depicted by something like a road map.

Routing tables inform routers about their neighbors, for example. Neighbors are nodes with a direct connection to the router in question. Remarkably, it is not the physical distance that determines whether or not an object is regarded to be a neighbor. In cyberspace, it is the number of so-called hops that define neighborhood, nearness and distance.1 Hops are the basic unit of measurement and evaluation in cyberspace. Even "time to live", the standard measure used to determine the survival time of data packets during the transmission process, is counted in hops. (After "time to live" of 30 hops or so, data packets are to be dropped and sent anew, because it can be assumed that they got caught in a loop and are unable to reach their destination.)

Just as a footnote: While we are used to measuring geographic distance by the time it takes to traverse it, the constitution of cyberspace seems to suggest the opposite convention: transmission time has to be translated into something else like hops - a quite arbitrary quantity - not least because of differing time zones and computer clocks which are not adjustable globally.) Unlike in the real world, a uniform time-of-day cannot be assumed in cyberspace.

Every network address known to a router and stored in its routing table has attached to it the paths through which it can be reached and the number of hops it takes to get there. If several paths are known, it is the number of hops that tell which path is the best. However, the number of hops is no objective figure. If a path through cyberspace is considered to be bad - say, because it crosses the wrong networks, too busy gateways or hostile territories - the number of the path's hops can simply be raised. An easy way of ruling out a path is therefore achieved by setting the number of hops higher than that of its alternatives. Thus, hops on the one hand serve as a quantity to measure distance, on the other they are used as a tool to express specific preferences in the selection of data paths. Apparently, the ordering of virtual spaces can do without the common distinction between subjective perception of space and its objectified counterpart, a context-independent, universally valid measuring of space.

Routing space binds the logical space of names to the physical space of wires and machines. It thereby generates its own topology, which allocates a virtual place to things on the net - a place whose topological position, neighborhood and traffic connections may change from one minute to the next.

4 IV
  The chronological order of the address space began to place strain upon the routers around the year 1990. Routing tables became longer and longer, and consequently more and more time was needed to compute the routes to the data packets' destination addresses. As mentioned before, the numerical address space of the Internet had only a chronological order. No topological structure existed, which would have allowed network addresses to be aggregated in a hierarchical manner. Therefore, all routers with direct access to the net needed a complete picture of all networks to guarantee full connectivity throughout the Internet. Consequently, routers became a bottleneck in the Internet's traffic flow. To avoid a complete breakdown in the near future, something had to be done to stop the accelerating growth of routing tables.

Long discussions took place among the Internet engineers as to how the address space could be designed to be more routing-friendly. A new order was needed, an order that would allow to group Internet addresses like telephone numbers under a few prefixes (like country codes, for example). Two proposals were presented. One is called the geographical addressing approach. It suggests that logical addresses should follow pyhsical addresses. According to this model, Internet address space would conform to geopolitical ordering conventions. The structure of Internet addresses would more or less reflect politically defined territories. However, the geographical approach to ordering the address space has several problems, the most important of which is the lack of a central power to enforce a centrally organized address plan. The address system of the telephone had, as Andeen & King (1997, 214) put it "developed in a monopolistic, paternal, top-down manner". In other words: spaces generated by the traditional communication technologies reflect political circumstances, which are neither reproducible nor desirable for cyberspace.

The second solution is referred to as provider-based addressing. This approach intends to design the address space around organizations instead of politically defined territories. A good example for this model of spatial ordering is the address plan of the mobile phone networks. The structure of mobile phone numbers doesn't refer to the physical location of its owner but to the company that provides the communication service. According to this model, the topology of the Internet's address space would reflect the organizational configuration of all Internet Service Providers world-wide. Where one is located would then depend on what provider one subscribes to.

While it remains unclear whether or not geographic addressing has at all a chance to succeed, provider-based addressing is well under way to reorganize the ordering conventions of cyberspace. Internet addresses that once belonged to the networks themselves are currently transformed into commercial goods to be controlled by competing Internet providers, who rent them out to their customers. The spatial consequences of this regulatory change poses all manner of questions: For example:

  • Will the provider-based ownership of addresses become a leverage for accumulating power in cyberspace?
  • What could the addresses' commodification mean for the conventions of identifying and locating objects on the net - and what for the computing of paths and distances in cypberspace?
  • Will the shortest path with the fewest hops be the one, that costs the least?
All in all: The stretching of society over time and space, as described by Giddens, brings about new forms of spatial orders. Right now, one can only speculate as to what form they will be take in future. More can be said, however, about the circumstances under which new virtual topologies develop. The ordering of things in cyberspace won't be a monopolistic, state-governed matter but the transitory result of competing commercial organizations. At the same time, public power, the traditional source of regulating telecommunicative spaces seems to fall into decay. Thus, new types of space constituted by a new type of communication technology meets with changing spatial ordering conventions. It will be interesting to watch to which extent traditional, state-bound spatial orders become challenged and transformed by this developments.

1 Hops are steps or jumps from one router to the next. Since the Internet's transmission technique doesn't reserve extra lines for the data exchange between two hosts, single data packets get to their destination by jumping from router to router.


home page about us documents miscellaneous sitemap
home page about us documents miscellaneous sitemap

Projektgruppe "Kulturraum Internet". c/o Wissenschaftszentrum Berlin für Sozialforschung (WZB)
Reichpietschufer 50, 10785 Berlin. Telefon: (030) 254 91 - 207; Fax: (030) 254 91 - 209;
; http://duplox.wzb.eu.