The Actually Existing Internet: Mapping Network Topology

Liam Mullally

31st January 2024

Figure 1: Visualization study of inbound traffic measured in billions of bytes on the NSFNET T1 backbone for the month of September 1991.


My previous series of blogs (see 1, 2, 3) for Autonomy dealt with the question of the rise of techno-pessimism on the left since the end of the 1990s. This second blog series, on network topology, turns towards a more overtly materialist analysis through a focus on infrastructure. Some recent work at Autonomy has been drawing on CAIDA’s topological data to build a clearer picture of the material contestation of the internet over time (thanks especially to Sean Greaves of the Autonomy Data Unit, who has been a great help on the data side); this blog series draws on some early findings from this work (along with some more conventional “desk” research) to start sketching what might be called the “actually existing internet” – the internet as a material occurrence in the present.

The goal here is to clarify how the material conditions of the past differ from those of the present, how this might have affected changing imaginaries and what position it leaves us in today. In short, it might begin a (fairly low resolution) materialist analysis of the production and contestation of the networks which make up the internet, with an eye towards if and how the left might be able to contest it in the future.

Studies of the shape of network architecture generally occupy one of three paradigms: logical topology, physical topology and topography. The first of these, logical topology, describes ideals or designs of network structure in abstract terms and without reference to real infrastructure. These are effectively schematics of modes of relation to be adopted within the network. Examples of logical topologies include mesh, ring and bus.

Figure 2: Some of the most common models of logical topology.

Physical topology, by contrast, is interested in mapping actual network systems, and does so from the abstract perspective of the network itself. Often, this is represented only as a series of nodes linked by vectors in space.

Figure 3: A graph of physical internet topology produced by CAIDA, showing total abstraction from the network’s articulation in geographical space.

Finally topography, a paradigm pioneered by materialist media scholars like Lisa Parks and Nicole Starosielski, describes the network from the external perspective of its physical infrastructures and routes as they contend with real world geography and politics. While such studies capture the actual articulation of networks in space, they are also inevitably local and partial in scope. They offer profound insights into the relation between architecture and the built environment, but this kind of granular analysis can never capture the network as a whole.

Figure 4: Excerpt of TeleGeography’s 2019 Submarine Cable Map, showing the routes of submarine cables around the UK.

These three registers encounter networks at different levels of abstraction, but all are required to understand its shape: logical topology is primarily a concept of design, a rubric for the production and interaction of agents in the network; physical topology engages with the actual shape of an existing network, but does so from a perspective internal to the logics and mechanisms of that network (and so a typical ‘map’ of physical topology will include no reference to geographic arrangement at all); and finally topography, which is intensely interested in the position of material infrastructures – cables, exchanges, routers and servers – but less interested in the protocols and logics which route information within them. For this reason topography tends to have more of an affinity with talking about basic network infrastructure, and topology with the flow of traffic over this infrastructure.

As in any domain of knowledge, none of these registers can be stably isolated. Geography produces the patterns of physical topology, even if its visualisations do not usually acknowledge this, and the topography of a network is determined as much by logical topology as it is geography. It would be wrong to lend any of these levels an ontological primacy or immediacy. We should not, for instance, suggest that topography is more real than topology; afterall, as we navigate the internet’s virtual space topology, not topography most immediately defines our experience. Topography, physical topology and logical topology exert pressure on one another. This would be a “trialectic”, to use an unwieldy term of Henri Lefevbre’s (it might be easier to just say an ecology). The job of analysis here is to identify the changing contours of this material ecology over its lifetime, to extract and clarify pressure points that might be used to steer it towards a more liberatory arrangement, towards some kind of horizon.

“A brief history of the internet”

The last thing you should do if you want to construct an understanding of the web today is to consult so-called “brief histories” (e.g. here, here or here). Most of these, you will find, end somewhere between the W3C conference in 1993, and the development of the Mosaic and Netscape browsers in the following years. If you are lucky, you might be offered a brief account of the popularisation of open-source development in the early 2000s; some will go further to the emergence of “Web 2.0” and the growth of a platform architecture in web design which prioritised user content, engagement and interaction (although user content was also a feature in earlier internet architectures, like USENET). These final markers in the internet’s “history” seem to suggest changes in software and business models, not to hardware or basic protocols – concerns which are relegated to the 1980s and earlier.

Of course, more time has passed between the early 1990s and today than between 1969, when the first computers were connected to the ARPANET, and the 1990s. Within both these periods there have been many different articulations of what the internet is: from ARPANET and MILNET, the NSFNET backbone, satellite and submarine connections between nations, from USENET to the early World Wide Web, to the web of today. While some of the routes and infrastructures have remained unchanged, it is very reasonable to characterise the various internets of the last 50 years or so as distinct network technologies which embody different kinds of relations. This may be more obvious in the move between different basic protocols (different logical topologies) as in the arrival of something like TCP/IP or the World Wide Web, but it remains true within the specific architecture of the network itself. In aspects of its fundamental topography and even its topologies, the internet of today has become something else: proprietariness, cloud architecture, even the massive expansion of phones as an object in the network, should all have affected its infrastructure.

While some of the routes and infrastructures have remained unchanged, it is very reasonable to characterise the various internets of the last 50 years or so as distinct network technologies which embody different kinds of relations

There are different ways to map the internet and who owns it: fibre-optic cables, exchange points, server farms, autonomous systems, submarine cables and DNS servers all provide potential glimpses into its organisation. The actual infrastructure of the internet is layered on top of what came before: the routes of the earliest deep sea telegraph cable, built between the US and the UK in the 1850s remain in use, for instance. This is also true of the internet of today, which was built on top of the internet of the past. Indeed, one observation that will come out of a closer development of the arrangements that the internet has held over its lifetime, is the contested nature of structures which endure. Crucially, the internet does contain what might be called “left infrastructures”, as well as extractive, exploitative, controlling and accumulative ones.

Between this contestation, there exists something like ‘the actually existing internet’. This is not simply one unified thing, but an accumulation of network infrastructure, cable routes, satellites, exchanges, protocols, software and information. The actually existing internet has continually been reinvented since the 1990s and especially since the 1960s. This blog is an attempt to trace the emergence of the actually existing internet, to see what constitutes the internet today and how it differs from that of thirty or fifty years ago.

Perhaps it has become difficult to form horizons today, not because of a failure of our imaginations (as Jameson argued in relation to a more general crisis), but because of failures in the realm of politics and power, a failure to grasp at important pressure points while they were available.

Through these changes in infrastructure, rather than those of the ideas themselves, we might reach a clearer analysis of the shrinking digital horizons of the contemporary British Left that I discussed in the first blog of my previous series. Perhaps it has become difficult to form horizons today, not because of a failure of our imaginations (as Jameson argued in relation to a more general crisis), but because of failures in the realm of politics and power, a failure to grasp at important pressure points while they were available.

The following three blogs will track the development of the internet over time, starting with its development from the 1960s, then focusing on its privatisation from the 1990s and finally looking at the changing distribution of internet resources from 2010 to today.

Selected bibliography

  • Lisa Parks, Cultures in Orbit: Satellites and the Televisual (Durham: Duke University Press, 2005)
  • Nicole Starosielski, the undersea network (Durham: Duke University Press, 2015)
  • Henri Lefebvre, Rhythmanalysis: Space, TIme and Everyday Life (London: Continuum, 2004)
  • Frederick Jameson, The Seeds of Time (New York: Columbia University Press, 1994)

Liam Mullally is a CHASE-funded PhD candidate in Cultural Studies at Goldsmiths, University of London. His research interests include digital culture, the history of computing, information theory, glitch studies, and the politics and philosophy of noise. Previously he has worked as a copywriter and tech journalist. He is working on several projects with Autonomy, from skills commissions to policy strategy.