Balázs Bodó, University of Amsterdam
Jaya Klara Brekke, Durham University
Jaap-Henk Hoepman, Radboud University

The rapidly evolving blockchain technology space has put decentralisation back into the focus of the design of techno-social systems, and the role of decentralised technological infrastructures in achieving particular social, economic, or political goals. In this entry we address how blockchains and distributed ledgers think about decentralisation.

Decentralised network topologies

A network is made of nodes, and edges, or interconnections between the members of the network. In a distributed network every node has roughly the same number of edges, and there are more than one routes in which nodes can connect with each other. This means that the topology of the network does not contain nodes in central or privileged positions, or if there are hierarchies built into the network, each node belongs to more than one hierarchy. This gives distributed networks a special property: the failure of a few nodes (even if they are chosen on purpose) still leaves the network connected, allowing all nodes to communicate with each other (albeit over a possibly much longer path than in the original network). In mathematics (and computer science) Graph Theory is devoted to the study of networks and their properties (Bondy and Murty, 2008).

Figure 1: Various network topologies (Baran, 1964).

Though often used as synonyms, decentralised and distributed networks are not the same. Decentralised networks are built from a hierarchy of nodes, and nodes at the bottom of the hierarchy have only a single connection to the network. Failure of a few nodes in a decentralised network still leaves several connected components of nodes that will be able to communicate with each other (but not with nodes in a different component).

The degree of decentralisation and distributedness varies from network by network. In general networks that are more distributed are more resilient to the failure of individual nodes or loss of connection between them. This resilience applies to both concrete and virtual networks i.e physical network infrastructures (such as the routers, cables, backbones, WIFI hotspots of the internet), and virtual networks running on the physical layer, such as blockchain networks, or file sharing networks.

Initially designed to be a cold-war resilient distributed network, the internet is in fact a decentralised network. Consequently, there are multiple stakeholders, and multiple physical as well as virtual bottlenecks where the network is controllable, or vulnerable to surveillance, and failure (Forte et al., 2016; Kaiser, 2019; Kastelein, 2016; Snowden, 2019). Likewise, while the TCP/IP protocol envisaged a network in which each node (user, machine) could be both an information sender and receiver, in practice, highly centralised virtual networks emerged in knowledge production, communication, or commerce. The recent wave of re-decentralisation (Redecentralize.org, 2020) tries to address the centralisation of the virtual layers - often assuming this will lead to decentralisation in other dimensions including power and political control (Buterin, 2017).

Advantages and disadvantages of decentralisation

Different network topologies come with particular advantages and disadvantages, that vary with the degree of centralisation, and the ways networks become more or less distributed over time. Distributed networks are more resilient to failure but incur a cost to maintain coordination. Centralised networks are much easier to maintain, but the central node can be a performance bottleneck and a single point of failure.

In the following table we summarised the main costs and benefits associated with distributed and centralised networks.

In distributed networks, each node has a wide range of responsibilities and associated costs. A distributed network is only operational if there is a coordination mechanism between the nodes.

In the absence of robust solutions to the problems of coordination and fault tolerance, Lamport et al (2019) have noted, a distributed system is only a network “in which the failure of a computer you didn't even know existed can render your own computer unusable''.

Coordination problems must address, for example, how nodes reach each other (as in the internet routing system); how to deal with competition and race conditions (when multiple nodes want to use the same limited resource, such as a network printer); or how the system’s operational and development processes are governed (Katzenbach and Ulbricht, 2019). While many of the more common technical coordination issues in distributed networks have been addressed and have robust solutions, governance, which might be equally distributed, remains experimental (Arruñada and Garicano, 2018; Atzori, 2017; De Filippi and Loveluck, 2016a). Most of the distributed applications and services have bare-bones, generic governance frameworks. Governance, however, entails more than, for example, an infrastructure of secure voting. Effective participation in the governance mechanisms of a distributed social, political, economic system also requires substantive investment from the individual in terms of knowledge, time, attention, engagement.

The problem of fault-tolerance has to do with failures and attacks[1], and ensures that the overall network remains functional and continues to work to achieve its overarching goal while some of its components fail. Attacks that are particular to distributed and decentralised systems include DDoS (Distributed Denial of Service)[2] and Sybil attacks[3]. Distributed architectures are designed to be tolerant of the failure of a relatively high number (typically 30-50%) of all nodes in a network[4]. But Troncoso et al. (2017) also showed that decentralisation, done naively, may multiply the ‘attack vectors’, and security risks, not least the breach of privacy. Distributed architectures might also be worse in terms of availability and information integrity, as the failure of nodes may have a fundamental impact on these properties.

In distributed networks, individual nodes must also take care of their own security, and availability. Distributed networks also have issues with efficiency, such as the transaction throughput of blockchain systems, or the bandwidth and latency in the TOR routing network.

In return, when done right, distributed networks offer higher resilience. There is also a lower risk of any central actors taking control, or exercising unilateral power over the network. For this reason, decentralised network topologies are also used to achieve privacy, censorship resistance, availability, and information integrity information security properties (Hoepman, 2014; Troncoso et al., 2017b).

In centralised systems coordination is taken care of by central actors who can specialise, and this leads to efficiency gains. There are costs to this, however, including making the network more vulnerable to the failure, or the abusive behaviour of that central node. Since network transactions run through a specific server, this grants those who control that server significant powers to observe, manipulate or cut off traffic (Troncoso et al., 2017), as well as to control, censor, tax, limit or boost particular social interactions, economic transactions, information exchange among network participants, and unilaterally set the conditions of interactions within the network.

To illustrate this cost-benefit calculus consider the privacy protecting TOR network. TOR is able to give reasonable levels of privacy at the cost of using a distributed network to route messages with lower speeds, and larger latency. These costs are seemingly too large for everyday users who are willing to settle for lower levels of privacy. On the other hand, for political dissidents who fear government retribution, journalists, whose integrity depends on their ability to protect their sources, and other groups for whom strong privacy is essential, the cost-benefit analysis justifies the higher costs of using this distributed network.

Both the costs and the benefits of using distributed network topologies are dynamic in nature, and are heavily dependent on factors both internal and external to the network. For example, the unresolved problem of distributed governance often creates a certain structurelessness in the social, political dimensions of distributed networks. As Freeman (1972) or De Filippi and Loveluck (2016) pointed out, seemingly unstructured social networks risk informal centralisation of their governance. In fact, blockchain networks have highly centralised forms of governance. (Azouvi, Maller, and Meiklejohn, 2018; De Filippi, 2019; De Filippi and Loveluck, 2016b; Musiani, Mallard, and Méadel, 2017; Reijers, O’Brolcháin, and Haynes, 2016). Blockchain networks may also suffer from centralisation in other dimensions of power. For instance, the proof-of-work (PoW) protocol randomly assigns a miner node to validate the latest batch of transactions for a relatively large reward to minimize the risk of a malicious miner hijacking the transaction ledger. The corresponding low chance of being rewarded for being honest forced miners to aggregate into a handful of coordinated mining pools, which control the vast share of this critical resource in an otherwise physically, geographically distributed network. The alternative approach, proof-of-stake (PoS) requires that those who wish to validate transactions stake their decisions with hard (crypto)cash: the larger the stake, the larger the validating power. PoS may remove mining pools, but creates another form of centralised power, namely that of capital. On the other hand, the increasing legal pressure on P2P file sharing networks, in particular on central nodes, pushed these projects towards increasingly distributed architectures, such as bittorrent networks, with distributed hash tables (Giblin, 2011).

Distributed systems in practice

While distributedness, as we have noted earlier, have been proposed as a general template for both the physical and the virtual digital networks, truly distributed networks only established themselves in particular niche applications, due to their particular cost-benefit balance.

P2P systems: P2P networks collectively make a resource (computation, storage) available among all nodes in the network. Examples of peer-to-peer computation networks are Seti@Home[5] and Folding@home[6]. Napster, Kazaa, or the bittorrent networks are peer-to-peer storage and file sharing networks, used to distribute copyrighted works under conditions of limited legal access (Johns, 2010; Patry, 2009). The peer-to-peer nature of these networks made it much harder to censor them and to take down material that infringed on copyrights (Buford, Yu, and Lua, 2009).

Distributed ledgers are distributed data structures where a set of bookkeeping nodes (sometimes called miners), interconnected by a peer-to-peer network, collectively maintain a global state without centralised control (Narayanan et al., 2016). Bitcoin (Nakamoto, 2008) was the first distributed ledger, inventing blockchain as the data structure to store transaction histories of digital tokens capable of digitally representing units of value. Ethereum generalised the distributed ledger from recording transactions to instead process code and store the state of the network. Bookkeeping nodes maintain consensus on the list of executed transactions and their effect on the global state, as long as a specified fraction of the bookkeeping nodes is honest and active.

Secure multiparty computation allows several participants to collectively compute a common output, which is based on each of their private inputs. Instead of sending the private inputs to one central coordinator (that would therefore learn the values of all private inputs), the algorithm to compute the value is distributed and the computation is done on the devices of the participants themselves, thus ensuring that their inputs remain private (Cramer, Damgard, and Nielsen, 2015; Yao, 1982).

Decentralisation as a social template

Distributed networks have brought experimentation with new coordination mechanisms, new ways to manage risks, and failures, lowering transaction costs and removing central powerful positions in technical terms. Proponents of disintermediation hope that these same logics provide new tools for horizontal social coordination, and the removal of political, economic, or social intermediary institutions, previously fulfilling those tasks.

The centralisation/decentralisation dichotomy is often framed in terms of power asymmetries, where distributed architectures are proposed as an alternative to authoritarian, coercive forms of political power. This dichotomy rests on a number of assumptions about power, and often does not fully account for the ways that, in practice, decentralisation in one dimension might produce or be enabled by centralisation in another. In terms of economics, distributed digital networks often align with the concept of perfectly competitive markets, designed to prevent the emergence of entities in a monopoly position, whether information, resource, or other monopoly (Brekke, 2020). Yet in practice, markets tend to rely heavily on a regulatory body to ensure fair competition. Distributed ledger technologies (DLT) have also offered a possible technical solution to the loss of trust in institutional actors (Bodó, 2020), by setting up networks with little reliance on trusted third parties, and minimising the need to have trust in interpersonal relations (Werbach, 2018). Yet in practice, DLT brings along new kinds of intermediaries, from interface designers and wallet developers, to exchanges, miners, full nodes and core developers, therefore requiring new forms of accountability methods.

The recent popularity of distributed technical networks raised important questions about the preferred modes of social, political, or economic organisation. Digital innovation changes the costs and benefits of coordination and collaboration (Benkler, 2006). This highlights questions about the roles that intermediaries play in those relations (Sen and King, 2003). For example, cryptocurrency technology may have successfully demonstrated that there is no need for a centralised intermediary to keep accounts, or even run an asset exchange. However, that is not the only function of banks and exchanges. Trust generation, due diligence, risk assessment, conflict resolution, rules provision, accountability, insurance, protection, stability, continuity, and education are arguably also core functions of the banking system, offered in conjunction with the bookkeeping function. A second set of questions address the various layers which constitute a complex techno-social system, and the fact that a distributed topology at one layer, may not produce, require, or allow a distributed form of organisation at the other. In fact, often highly centralised governance is a precondition of a distributed system to function, as is currently the case in blockchain based systems. Another example would be the role of governments to ensure fair and open competition on various markets, such as anti-trust regulation, or in politics.

Conclusion

Decentralised and distributed modes of organisation are well defined in computer science discourses and denote a particular network topology. Even there, they can be understood either as an engineering principle, a design aim, or an aspirational claim. In the decentralisation discourse these three dimensions are often conflated without merit. A decentralised network design might not produce decentralising effects and might not either necessarily be decentralised in its actual deployment.

When the technical decentralisation discourse starts to include social, political, or economic dimensions, the risk of confusion may be even larger, and the potential harms of mistaking a distributed system for something it is not, even more dangerous. Individual autonomy, the reduction of power asymmetries, the elimination of market monopolies, direct involvement in decision making, solidarity among members of voluntary associations are eternal human ambitions. It is unclear whether such aims can now suddenly be achieved by particular engineering solutions. An uncritical view on decentralisation as an omnipotent organisational template may crowd out alternative approaches to creating resilient, trustworthy, equitable, fault resistant technical, social, political or economic modes of organisation.

References

Arruñada, Benito, and Luis Garicano. 2018. “Blockchain : The Birth of Decentralized Governance.”

Atzori, Marcella. 2017. “BLOCKCHAIN TECHNOLOGY AND DECENTRALIZED GOVERNANCE: IS THE STATE STILL NECESSARY?” Journal of Governance and Regulation 6(1):1–37. doi: 10.22495/jgr_v6_i1_p5.

Azouvi, Sarah, Mary Maller, and Sarah Meiklejohn. 2018. “Egalitarian Society or Benevolent Dictatorship: The State of Cryptocurrency Governance.” The Fifth Workshop on Bitcoin and Blockchain Research.

Baran, Paul. 1964. “On Distributed Communications Networks.” IEEE Transactions on Communications Systems 12(1):1–9.

Benkler, Yochai. 2006. The Wealth of Networks : How Social Production Transforms Markets and Freedom. New Haven: Yale University Press.

Bodó, Balázs. 2020. “Mediated Trust: A Theoretical Framework to Address the Trustworthiness of Technological Trust Mediators.” New Media & Society 146144482093992. doi: 10.1177/1461444820939922.

Bondy, J. A., and U. S. R. Murty. 2008. Graph Theory. New York: Springer.

Brekke, Jaya Klara. 2020. “Hacker-Engineers and Their Economies: The Political Economy of Decentralised Networks and ‘Cryptoeconomics.’” New Political Economy. doi: 10.1080/13563467.2020.1806223.

Buford, John F., Hong Heather Yu, and Eng Keong Lua. 2009. P2P Networking and Applications. Amsterdam ; Boston: Elsevier/Morgan Kaufmann.

Buterin, Vitalik. 2017. “The Meaning of Decentralisation.” Retrieved February 10, 2019 (https://medium.com/@VitalikButerin/the-meaning-of-decentralization-a0c92b76a274).

Cramer, Ronald, Ivan Bjerre Damgard, and Jesper Buus Nielsen. 2015. Secure Multiparty Computation and Secret Sharing. Cambridge: Cambridge University Press.

De Filippi, Primavera. 2019. “Blockchain Technology and Decentralized Governance: The Pitfalls of a Trustless Dream.” SSRN Electronic Journal. doi: 10.2139/ssrn.3524352.

De Filippi, Primavera, and Benjamin Loveluck. 2016a. “The Invisible Politics of Bitcoin: Governance Crisis of a Decentralised Infrastructure.” Internet Policy Review 5(3). doi: 10.14763/2016.3.427.

De Filippi, Primavera, and Benjamin Loveluck. 2016b. “The Invisible Politics of Bitcoin: Governance Crisis of a Decentralised Infrastructure.” Internet Policy Review 5(3). doi: 10.14763/2016.3.427.

Fischer, Michael J., Nancy A. Lynch, and Michael S. Paterson. 1985. “Impossibility of Distributed Consensus with One Faulty Process.” Journal of the ACM 32(2):374–82. doi: 10.1145/3149.214121.

Forte, Andrea, Vanesa Larco, Amy Bruckman, Andrea Forte, Vanessa Larco, and Amy Bruckman. 2016. “Decentralization in Wikipedia Governance.” 1222(March). doi: 10.2753/MIS0742-1222260103.

Freeman, Jo. 1972. “The Tyranny of Structurelessness.” Berkeley Journal of Sociology 151–64.

Giblin, Rebecca. 2011. Code Wars: 10 Years of P2P Software Litigation. Cheltenham, UK ; Northampton, MA: Edward Elgar Publishing.

Hoepman, Jaap-Henk. 2014. “Privacy Design Strategies BT  - ICT Systems Security and Privacy Protection.” Pp. 446–59 in, edited by N. Cuppens-Boulahia, F. Cuppens, S. Jajodia, A. Abou El Kalam, and T. Sans. Berlin, Heidelberg: Springer Berlin Heidelberg.

Johns, Adrian. 2010. Piracy: The Intellectual Property Wars from Gutenberg to Gates. University Of Chicago Press.

Kaiser, Brittany. 2019. Targeted. New York: HarperCollins.

Kastelein, Richard. 2016. “World Wide Web Creator Tim Berners-Lee Wants to Decentralise the Internet with P2P and Blockchain Technologies.” BlockchainNews.

Katzenbach, Christian, and Lena Ulbricht. 2019. “Algorithmic Governance.” Internet Policy Review 8(4).

Lamport, Leslie, Robert Shostak, and Marshall Pease. 2019. “The Byzantine Generals Problem.” Pp. 203–26 in Concurrency: the Works of Leslie Lamport.

Musiani, Francesca, Alexandre Mallard, and Cécile Méadel. 2017. “Governing What Wasn’t Meant To Be Governed: A Controversy-Based Approach to the Study of Bitcoin Governance.” in Bitcoin and Beyond: Cryptocurrencies, Blockchains, and Global Governance, edited by M. Campbell-Verduyn. Routledge.

Nakamoto, Satoshi. 2008. “Bitcoin: A Peer-to-Peer Electronic Cash System.” Www.Bitcoin.Org.

Narayanan, Arvind, Joseph Bonneau, Edward Felten, Andrew Miller, and Steven Goldfeder. 2016. Bitcoin and Cryptocurrency Technologies. Princeton, NJ: Princeton University Press.

Patry, William F. 2009. Moral Panics and the Copyright Wars. New York, NY: Oxford University Press.

Redecentralize.org. 2020. “Redecentralize.Org.” Retrieved November 17, 2020 (https://redecentralize.org/).

Reijers, Wessel, Fiachra O’Brolcháin, and Paul Haynes. 2016. “Governance in Blockchain Technologies & Social Contract Theories.” Ledger 1:134–51. doi: 10.5195/LEDGER.2016.62.

Sen, Ravi, and Ruth C. King. 2003. “Revisit the Debate on Intermediation, Disintermediation and Reintermediation Due to E-Commerce.” Electronic Markets 13(2):153–62. doi: 10.1080/1019678032000067181.

Snowden, Edward. 2019. Permanent Record. Metropolitan Books.

Troncoso, Carmela, Marios Isaakidis, George Danezis, and Harry Halpin. 2017a. “Systematizing Decentralization and Privacy: Lessons from 15 Years of Research and Deployments.” doi: 10.1515/popets-2017-0052.

Troncoso, Carmela, Marios Isaakidis, George Danezis, and Harry Halpin. 2017b. “Systematizing Decentralization and Privacy: Lessons from 15 Years of Research and Deployments.” Proceedings on Privacy Enhancing Technologies (4):307–29. doi: 10.1515/popets-2017-0052.

Werbach, Kevin. 2018. The Blockchain and the New Architecture of Trust. Cambridge, MA: Mit Press.

Yao, Andrew C. 1982. “Protocols for Secure Computations.” Pp. 160–64 in. IEEE.

1. Failure can mean multiple things: the unavailability of a node; the unreliable, unexpected, or unaccounted for behaviour; and any malicious, manipulative or destructive behaviour. Failures can happen for a number of reasons: stochastic processes which may equally affect any node in a network due to their intrinsic properties; failures in some of the underlying layers: energy failures, environmental force majeure; as well as failures due to attacks by malicious actors. ↑

2. A DDoS attack is when the bandwidth of a network is overloaded by flooding it with traffic coming from a distributed set of nodes. ↑

3. A Sybil attack is when some actor/s create/s many nodes such that the network seems distributed, when in actual fact it might be controlled by a single or small set of actors. ↑

4. The so-called Byzantine Agreement protocols allow a system to agree on a common output even if at most one-third of the members are faulty (in the Byzantine sense, meaning that they are malicious) (Lamport, Shostak, and Pease, 2019). But this is only the case under certain conditions. In particular, fully asynchronous systems (where there is no bound on the time it can take for a message to arrive or the time a node may take to complete a step) defy solutions to the Byzantine Agreement problem (Fischer, Lynch, and Paterson, 1985). This highly theoretical line of research re-emerged with the birth of Bitcoin and the subsequent explosion of distributed ledger technologies that exactly needed what Byzantine Agreement offered: reaching agreement on the global order of transactions, when faced with potentially malicious adversaries. ↑

5. Started in 1999, it’s aim is detecting intelligent life outside Earth, see https://setiathome.berkeley.edu ↑

6. Started in 2020, it’s aim is to simulate protein dynamics, see https://foldingathome.org ↑