You are currently viewing <span class='p-name'>Decentralize or perish: Making network architecture political</span>

Decentralize or perish: Making network architecture political

I originally wrote the following essay for an academic audience, so it isn't quite in my typical lighthearted, no-holds-barred style. Nonetheless, I believe it's well worth your time if you do anything at all on the Internet ever.


‘Lo’! This was the very first message sent across the nascent internet (Schwartz & Kleinrock, 2010, p. 32). It was not intentional. In order to verify that the connection worked, Leonard Kleinrock meant to transmit the word ‘LOGIN’, letter by letter. ‘L’ and ‘O’ went across. However, on ‘G’ the system crashed. Thus, it would forever be that the first utterance propagated across the network was ‘lo’, as if the circuitry itself was saying “Lo and behold, the future has arrived!”. This is rather appropriate for the grandeur with which it was proclaimed that the internet would change the face of humankind. It would become a virtual public square, where everyone had an equal voice in the global debate. It would belong to all and be under the control of none. It would engirdle the world, unite humanity and set free the dormant forces of innovation. Cyberspace declared independence from the past and welcomed all into the “new home of Mind” (Barlow, 1996).

This was a rather paradoxical vision from the beginning. However manifestly open, the internet was in fact being built behind closed doors. It was governed by decrees in the form of code whose ramifications were understood by few. It was no “magically levitating public sphere”, but firmly grounded in physical infrastructure which was privately owned (Chaves, 2010). More recently, with the consolidation of internet platforms offered by giant companies like Google and Amazon, speech has also started being invisibly shaped by these firms’ policies. The result has been that, for most people, online life is dominated by opaque, private organizations which take most major decisions.

Watching the stark contrast between the internet’s ideals and its realities, one cannot help but feel as if we have taken the wrong exit off the information superhighway and are now parked in the roadside dust, wondering what happened. This will be the underlying theme of the following pages, as I explore the way the internet is governed, why it never came to fulfill its romantic vision and how it may yet endeavour to do so.

At the heart of the problem is the idea of centralization. For the internet, this word takes two distinct, albeit closely intertwined meanings. Firstly, it refers to who has decision-making authority over public cyberspace. In this context, centralization means that this power gets concentrated into a smaller number of hands. Secondly, it is about ownership of platforms and infrastructure. In an analogy to Marx, I would call these means of computing. In this case, centralization refers to large numbers of people relying either on communications channels which have an increasingly small number of owners, or relying on an increasingly small number of privately owned communications channels. Examples are communities having fewer choices of internet service providers (ISPs), higher percentages of all websites being powered by dwindling numbers of large hosting providers, or individuals increasingly using only a handful of large social media platforms for their communications. Because ownership and authority so often go together, I use the generic terms ‘centralization’ and ‘decentralization’ from now on to refer to both meanings, unless the situation specifically demands more nuance.

This is all rather dense. To make it easier to understand, take the following analogy. Imagine that we all live in a village and that the Internet is whatever places people gather at to discuss. Imagine that at larger gatherings in the public square, it is the village elders who get to decide who is allowed to speak and in what manner. Or imagine that they get to pick the format, such as, for example, that everyone is allowed to speak about just one issue for one minute every week. These are examples of centralized decision-making authority. Now imagine that most people usually choose to gather at Bob’s house to discuss, because he has nice lawn chairs and makes barbecue for everyone. Sometimes, people also gather at Mary’s. You may try to start discussions outside of these two places, but you’d be swimming against the tide, since those two are just where everyone goes. On the Internet, platforms and servers are these gathering places. This is an example of the centralization of the means of computing. Thinking of it this way also makes it quite obvious why ownership is often equated with authority. Bob can kick out of his house anyone he doesn’t like, which severely limits that person’s contributions to the public discussion. Bob also gets to, for example, choose whether to place all the chairs in a big circle, or into several smaller ones. Such acts do not directly decide what is debated, but they certainly frame and shape speech. Obviously, decentralization is then the opposite: authority is split between more people; ownership of the meeting spaces is split between more people; there are more meeting spaces.

The early architects of the Internet were aware that decentralization had to be one of the core properties of the network they were building. The reasons weren’t just lofty ideals, but also pragmatism and Cold War anxiety. As Taplin (2017) neatly puts it, the designers of the Internet weren’t just “geeks on acid, dreaming of the future.” but also “financed by the military-industrial complex.” (p. 43). They had to be planning for scalability (making it easy to grow the network), redundancy (there had to be no single point of failure) and resilience (if whatever part of the network was knocked out by, say, a nuclear bomb, the remaining bits of the network would still each be able to function on their own). Accordingly, the core design tenet was the end-to-end principle (van Schewick, 2010). In a nutshell, this meant that the network itself would be as simple as possible, with the applications working at the edge, on the end host (i.e. your computer, or a website’s server). The implication would be that the Internet would be neutral and blind, neither encouraging nor discouraging any particular way of using it. This would also mean that there could be no central authority that could dictate how the network would be used, since the network would be little more than a loose collection of practices and wires. In fact, owing to fortuitous regulation, in the US even the telephone wires themselves, though owned privately, could not discriminate against any Internet traffic anyone wished to use them for (Lessig, 2001). As Lessig (2001) points out, “[The] commons was built into the very architecture of the original network. […] Neutrality existed at the physical and code layer of the Internet.”. In other words, it was utterly permissionless.

Permissionless is not the same as lawless, though. Cyberspace does have rules and norms, which are borne out of the code which underpins it (DeNardis, 2013). Network protocols are not merely technical standards and formats. As van Schewick (2010) points out, “since in most cases, a system’s functional requirements determine the system’s architecture only in part, system architects have considerable latitude […] to consider other goals too.” (p. 3). Indeed, it is in this space between necessity and possibility that a seldom-acknowledged phenomenon happens. This phenomenon has variously been called code as “politics by other means” (Latour, 1988, as cited in Musiani, 2016, p. 83), “code as law” (Lessig, 1999) and “ideology as expressed in […] technology” (Wu, 2010, p. 183). What it means is that coders are human. As such, when the technical architects of our public cyberspaces define the features of those spaces, it frequently happens that their preferences, biases and ideologies shine through. Their own beliefs mould our platforms and algorithms, the virtual environments which we inhabit. They influence us, as well as shape our discourse (Chaves, 2010).

The effect of the coders’ ideologies on cyberspace’s features is sometimes unintentional, but it can also be, and often is, quite deliberate. In Protocol – How control exists after decentralization, Galloway (2004) argues that technical specifications matter philosophically and politically. More than being tools or instructions, protocols become in his view a form of soft social control. They are a way of managing distributed, unhierarchical systems of humans like the Internet. Galloway (2004) explains how this form of policy works using the following analogy:

On an empty town street where drivers are prone to speed, the local residents elect to have speed bumps installed. […] In another part of town is an identical street also plagued by speeding. But here the neighborhood decides against speed bumps. Residents pass laws to reduce the legal speed limit. They install speed limit signs and increase police radar surveillance. Both solutions succeed in reducing the number of speeding cars. But which solution is protocological? […] It is the former. (p. 241)

The author explains that signs and police are effective in coercing drivers to slow down, but they place the burden of deciding whether to abide by these rules on the driver. Speed bumps, on the other hand, create an environment where it becomes beneficial to drive slowly. By reframing the conditions, they encourage a certain type of behaviour. They subtly enforce a certain norm. In a similar way, discourse is shaped online by technical architecture. In Galloway’s (2004) words, for a platform or network, “the question “how does it work?” is also the question “whom does it work for?” ” (p. XII).

Later in the book, the author shows how politics have quietly gone from being a side issue of system design to becoming an integral part of it. For example, he mentions a quote from a technical manual, which affirms that “IP uses an anarchic and highly distributed model[…].” (Hall, 2000, as cited in Galloway, 2004, p. 8), with Galloway pointing out that “That a technical manual glowingly uses the term “anarchic” is but one symptom of today’s strange new world!” (Galloway, 2004, p. 8). Galloway does not plead that it is wrong to make code political. On the contrary, he points out that protocol is a form of control with the implication that care should be taken in what values it abides by. It is inherently political, and therefore a matter of public debate.

Chaves (2010) made a similar point when she argued for the recognition of the “political nature of space”, suggesting that the virtual arenas of our discourse shape and are shaped by politics. She also went further by asking who should decide how these arenas work. She contended that a fair and neutral digital participatory public sphere, one which does not by its design favour the “interests of some social forces and ideas over others”, is unattainable as long as Internet governance remains as opaque as it is today.

She has a point. There is a fundamental contradiction between the idea of an Internet as a fair public space for everyone, on the one hand, and a method of governing it that involves the vast majority having no say in the policy which affects the network, on the other. Yet that is the situation we find ourselves in today.

That contradiction did not kill the dream of the Internet utopia on its own. There are several reasons why it was never attained, including the fact that the potential of the Internet might have been overestimated. But I would argue that there is one key avoidable aspect in which the Internet has gone the wrong way, one which makes a big difference: it has defied its initial vision and has become increasingly centralized.

To be fair, there have always been elements of centralization in the organization of the Internet. Some have been less controversial and generally considered harmless, like the ICANN’s[1] monopoly on controlling and distributing IP address numbers or the centralization implicit in the client-server model of the Web, which entails that a website’s data usually lives on a single central server from which the “clients” may access it (van Schewick, 2010). A point of centralization considered somewhat problematic, though, has long been the Domain Name System (DNS) (Chaves, 2010; Malcolm, Cohn & O’brien, 2017). The DNS is the central ledger of the Internet, which keeps track of which IP address number corresponds to each domain name (e.g. corresponds to Because it represents a choke point through which almost all traffic flows, the worry has been that it may be used as a tool for censorship, with Berners-Lee calling it the “Achilles’ heel by which [the Web] can all be brought down or controlled.” (Berners-Lee, 1999, as cited in Galloway, 2004, p. 10).

Even so, I would argue that there are six recent areas of increasing centralization which are far more consequential. First, Internet cables are owned in every country by small numbers of large ISPs. In the US, for example, following progressive mergers, only two giant companies remain to serve over three quarters of all broadband subscribers (Statista, 2019). Second, organizations and individuals increasingly rely on a few large cloud hosting providers to host their websites or systems (Andriole, 2015). Amazon Web Services dominates that market by a wide margin (Synergy Research Group, 2018). Third, the Web browser market is highly centralized, with almost two thirds of all people using Google’s Chrome (Statcounter, 2019a). Fourth, on mobile, which has become the top way of accessing the Internet, almost all applications come either from the Apple-controlled App Store, or the Google-controlled Play Store (Statcounter, 2019b). Fifth, internet search, which is one of the main ways people access information, is absolutely dominated in the West by Google (Haucap & Heimeshoff, 2013, p. 8). Lastly and most visibly, the advent of social media has meant that much of online public discourse has moved from personal websites, forums and the press to a tiny number of very large centralized social media services: Facebook, YouTube, Twitter, Reddit.

As already mentioned, ownership is often equated with control. Balkin (2004) highlights this fact in an essay on free speech on the Internet when he says that “The capitalist theory identifies the right to free speech with ownership of distribution networks” (p. 19). He goes on to argue that

Just as in the case of intellectual property, businesses that control telecommunications networks will seek to limit forms of participation and cultural innovation that are inconsistent with their economic interests. Once again, the goal is not necessarily censorship of unpopular ideas but rather diversion and co-optation of audience attention. (p. 22)

In an era when the Internet has become so important a venue for public debate that it is being called the “Fifth Estate” (Margetts, 2013), it should then be a matter of concern that a decreasing, already small number of private actors hence become arbiters of speech, especially if they get to shape discourse using the subtle tools offered by protocols and opaque algorithms. Indeed, for this reason and others, there appears to be a near consensus among scholars and technologists that a centralized Internet is a negative development for society (Berners-Lee, 2010; Kermarrec, 2013; Musiani, 2015; Poblet, 2018; van Schewick, 2010; Zittrain, 2018). As van Schewick (2010) puts it, “increasing the controllability of the network removes the very features of the network that make the Internet valuable.”.

This process of centralization has emerged naturally because of economic, legal and social factors, but it is not inevitable. However, conventional policy measures against private control over the flow of information on the network, like ‘net neutrality’ regulations, have proven contentious (Candeub, 2012, p. 674). Given this, and considering the nature of the Internet, I would suggest that a better solution would be code in the public interest. We ought to recognize the politics implicit in the architectures of the systems that act as our digital public squares and virtual printing presses. Moreover, we should have them designed in the full awareness of the kinds of behaviour they encourage or enable, so that these cyberspaces embody our values. I would argue that such spaces ought to be, at the very least: resistant to censorship; transparently governed; unskewed toward any particular set of social forces, excepting social forces opposed to these basic values[2]; maintaining personal privacy, the lack of which translates into less liberty and “kills experimentation” (Schneier, 2018); preserving individual sovereignity, in the sense of not infringing, nor enabling infringement, on one’s individual freedoms.

That is a hefty list of requirements, but there is a growing number of technologists who, aware of the need for code in the public interest, are developing projects in an attempt to tackle the problems created by a centralizing Internet. Many of them fall under the label of the “decentralized web”.

A first example is represented by WebBox and Solid, two projects proposing that every person physically host their own data on personal servers dubbed WebBoxes and, respectively, pods (van Kleek, Smith, Shadbolt & Schraefel, 2012; Mansour et al., 2016). The WebBoxes and pods would then be able to interface with, and selectively give access to data to online platforms. The premise is that physical ownership of one’s own data amounts to digital sovereignity and facilitates privacy.

Another example, founded on a similar assumption, is provided by the Fediverse, a decentralized network of social media services where all the servers are owned by the users, instead of a central authority (Gehl, 2015; Zignani, Gaito & Rossi, 2018). The underlying code is open source, so that it may be transparently checked or contributed to by members of the community.

Finally, there is the example of peer-to-peer networks. In this model, central hosting is replaced by a distributed scheme, where the computers of many individuals collectively store the information of a website. Accessing it also means becoming a host of at least some of it. Hence, a peer-to-peer network is upheld only by the users’ cumulative contributions. Musiani (2015) argues that adopting this technology for everyday use is both practical and means a reallocation of power between the “Giants” and the “Dwarfs” of the Web. In a paper on this technology’s use by the Catalan independence movement of 2017, Poblet (2018) also shows that it is resistant to censorship, as well as that it amounts to a new kind of digital participation, one of collectively upholding the very (virtual) arena where the debating public gathers.

“Decentralized web” technology is at its start, however. It presents no guarantees. Yet I would argue that the idea behind it is correct: the Internet’s social value is much higher if control and the means of computing are decentralized. The most effective approach to achieving this policy goal is likely intelligent technical design. Indeed, code is political and we should behave as such. And, though for many that idea is new, it is also a return to an old Internet adage:


We reject: kings, presidents and voting.
We believe in: rough consensus and running code.

                                                                                            - David D. Clark, 1992


[1] Internet Corporation for Assigned Names and Numbers

[2] Issue is related to the “paradox of tolerance”. See Popper, K. (1945/1971). Ch. 7, note 4.


Andriole, S.J. (2015). Who owns IT?. Communications of the Association for Computing Machinery, 58(3), 50-57. doi: 10.1145/2660765

Balkin, J.M. (2004). Digital Speech and Democratic Culture: A Theory of Freedom of Expression for the Information Society. New York University Law Review, 79(1), 1-58.

Barlow, J.P. (1996). A Declaration of the Independence of Cyberspace. Retrieved from

Berners-Lee, T. (2010). Long Live the Web. Scientific American, 303(6), 80-85.

Candeub, A. (2012). An End to End-to-End? A Review Essay of Barbara van Schewick’s Internet Architecture and Innovation. Federal Communications Law Journal, 64, 661-674.

Chaves, E. (2010). The Internet as Global Platform? Grounding the Magically Levitating Public Sphere. New Political Science, 32(1), 23-41. doi:10.1080/07393140903492100

DeNardis, L. (2013). The Emerging Field of Internet Governance. In Dutton, W.H. (Ed.), The Oxford Handbook of Internet Studies (pp. 543-564). Oxford, UK: Oxford University Press.

Galloway, A. (2004). Protocol: How Control Exists after Decentralization. Massachusetts, USA: The MIT Press.

Gehl, R. (2015). Building a Better Twitter: A Study of the Twitter Alternatives GNU social, Quitter,, and Twister. Fibreculture Journal, 190(26). Retrieved from

Haucap, J. & Heimeshoff, U. (2013). Google, Facebook, Amazon, eBay: Is the internet driving competition or market monopolization? . DICE Discussion Paper, 83. Retrieved from

Kermarrec, A.-M. 2013. Towards a personalized Internet: a case for a full decentralization. Philosophical Transactions of the Royal Society A, 371:20120380. Retrieved from

Lessig, L. (1999). Code: And Other Laws Of Cyberspace. New York, USA: Basic.

Lessig, L. (2001). The Internet under Siege. Foreign Policy, 127, 56-65. doi:10.2307/3183294

Malcolm, J., Cohn, C. & O’brien, D. (2017). Fighting Neo-Nazis and the Future of Free Expression. Retrieved from

Mansour, E., Sambra, A.V., Hawke, S., Zereba, M., Capadisli, S., Ghanem, A., … & Berners-Lee, T. (2016, April). A demonstration of the solid platform for social web applications. In Proceedings of the 25th International Conference Companion on World Wide Web (pp. 223-226). International World Wide Web Conferences Steering Committee. Retrieved from

Margetts, H. (2013). The Internet and Democracy. In Dutton, W.H. (Ed.), The Oxford Handbook of Internet Studies (pp. 543-564). Oxford, UK: Oxford University Press.

Musiani, F. (2015). Giants, Dwarfs and Decentralized Alternatives to Internet-based Services: An Issue of Internet Governance. Westminster Papers in Communication and Culture, 10(1), 81-94.

Poblet, M. (2018). Distributed, privacy-enhancing technologies in the 2017 Catalan referendum on independence: New tactics and models of participatory democracy. First Monday, 23(12). Retrieved from

Popper, K. (1971). The Open Society and Its Enemies. 1, The spell of Plato. Princeton, NJ, USA: Princeton University Press. (Original work published in 1945)

Schneier, B. (2018). Surveillance Kills Freedom By Killing Experimentation. Wired. Retrieved from

Schwartz, M. & Kleinrock, L. (2010). An early history of the internet. History of Communications. IEEE Communications Magazine, 48, 26-36. doi:10.1109/MCOM.2010/5534584

Statista. (2019). Number of broadband internet subscribers in the United States from 2011 to 2018, by cable provider (in 1,000s). Retrieved from

Synergy Research Group. (2018). Cloud Growth Rate Increased Again in Q1; Amazon Maintains Market Share Dominance. Retrieved from

Statcounter. (2019a). Browser Market Share Worldwide Jan 2018 – Jan 2019. Retrieved from

Statcounter. (2019b). Desktop vs. Mobile vs. Tablet Market Share Worldwide Jan 2018 – Jan 2019. Retrieved from

Taplin, J. (2017). Move Fast and Break Things: How Facebook, Google and Amazon Cornered Culture and Undermined Democracy. New York, USA: Little, Brown & Co.

van Kleek, M., Smith, D., Shadbolt, N. & Schraefel, M. (2012). A decentralized architecture for consolidating personal information ecosystems: The WebBox. (Conference paper, University of Southampton, Southampton, UK). Retrieved from

van Schewick, B. (2010). Internet Architecture and Innovation. Massachusetts, USA: The MIT Press.  

Wu, T. (2010). Is Internet Exceptionalism Dead? . In B. Szoka & A. Marcus (Eds.), The Next Digital Decade – Essays on the Future of the Internet (pp. 179-188). Washington, D.C., USA: TechFreedom.

Zignani, M., Gaito, S. & Rossi, G. (2018, June). Follow the “Mastodon”: Structure and Evolution of a Decentralized Online Social Network. Paper presented at the AAAI Conference on the Web and Social Media, USA. Retrieved from

Zittrain, J. (2018). Fixing the Internet. Science, 362(6417), 871.

The Ferridge

Wry writer and profligate pixelsmith

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.