I really appreciated Mark Nottingham’s memo RFC 9518 on Centralization, Decentralization, and Internet Standards. In it he provides a wide-ranging and balanced introduction to the topic in quite a short article. I have been collecting bits of arguments for decentralization (in a centralized location of course 😉) for the past couple of years, with the intention to write something round-up, but have struggled. I am so glad Mark did it, and so much better than I could!

I wholly recommend reading the piece. Below I will expand on it with some more wide-ranging sources. You don’t have to read his article to understand this post, but I hope by the end you will be even more motivated to go check out his.

The RFC 9518 memo covers different types and harms of centralization in an internet context, and indicates what can be done by standards organizations like the Internet Engineering Task Force (IETF). Nottingham notes that standards orgs cannot prevent centralization, but can make sure to consider “centralization risk” when evaluating a standard, and promote those that enable decentralization.

Yet he warns:

while decentralized technical standards may be necessary to avoid centralization of Internet functions, they are not sufficient

This is a problem for a certain sector of blockchain and web3 enthusiasts and why some pragmatic proponents came up with the term “sufficient decentralization.” Nottingham covers blockchain-style consensus in his memo. I also found Alice Yuan Zhang’s challenge of the “myths” of decentralization in web3 is a fun read. A quote from her:

Decentralization as praxis is rooted in direct action, striving to abolish capitalistic economics and supply chains which encode mass oppression into large-scale systems with many actors and minimal accountability.

Nottingham agrees, writing that centralization becomes a problem “when it has no checks, balances, or other mechanisms of accountability”. Architectural decentralization can certainly help, but it isn’t enough as explained in Capture Resistance by Robin Berjon.

Furthermore, we may not want total decentralization. Nottingham gives the example of content moderation as something that benefits from centralization, and makes the point that centralized structures such as “governments, corporations, and nonprofit organizations” can be useful. This level of nuance is what makes Nottingham’s shortish piece extra valuable. Nobody is under the illusion that decentralization is a panacea.

In discussing internet fragmentation, “public interest technologist” Mallory Knodel writes:

Sovereignty needs to be balanced against interdependence. Globalization – and its technologies – arose from the idea that interdependence prevents conflict. Economic interlinkage, or entanglement, disincentivizes attacks because it raises the risk of collateral damage and unintended consequences for the attacker. Perhaps some healthy fragmentation provides greater opportunity for interdependence beyond interconnectivity, and is an example of how to embrace a more complex internet landscape.

“Healthy fragmentation”… centralization and decentralization in balance. How do we combine them appropriately? Mike Masnick used the canonical example of the US interstate system in his essay Decentralization in All the Things:

having centralized infrastructure that is open and on which others can build in a decentralized manner can open up tremendous possibilities.

One main way we can do this with regards to internet protocols is to ensure interoperability and lower switching costs, which is where Nottingham says standards orgs can play a role. In this he is in agreement with Cory Doctorow, whose latest non-fiction book The Internet Con is a podium-pounding preachment on how “Network effects are merely how Big Tech gets big. Switching costs are how Big Tech stays big.” Interop lowers switching costs, and puts agency back in the hands of the people (see the presentation version of the book by watching Cory’s talk at DefCon last year).

Cryptography and Privacy researcher Sarah Jamie Lewis says:

Decentralization is important because building systems that distribute power is important. Building systems that resist abuses of power is important.

She argues that centralization/decentralization should not be conceptualized in terms of simple things like % control over a number of nodes, but “how trust and power are given, distributed and interact.” I think she puts it admirably when she writes, “Decentralized systems are systems that rely on the distribution of power to secure the system.”

That last sentence is certainly one to mull over.

But it is her point about who has access to power that makes decentralization such a wide-ranging topic. Once we start to go beyond the bounds of internet protocols the topic becomes more complex.

Twenty-five years ago the United Nations Development Programme’s 1999 report on decentralization stated that “Decentralization is not an alternative to centralization. Both are needed.” The report notes that subsidiarity is an important principle for:

… increasing the overall quality and effectiveness of the system of governance, while increasing the authority and capacities of sub-national levels. … Decentralization could also be expected to contribute to key elements of good governance, such as increasing people’s opportunities for participation in economic, social and political decisions; assisting in developing people’s capacities; and enhancing government responsiveness, transparency and accountability.

In general we might say decentralization is a substrate for agency.

In her lovely book How Infrastructure Works, Deb Chachra thinks about the future of infrastructure over the next 50 years in the face of climate change. For her, decentralization is a key to resilience, an important ability for the climate fight we are in as we rebuild our massively centralized infrastructure systems so communities can build to their own needs. Diffuse, diverse, and distributed… which also happen to be the founding principles (values?) of the internet itself.

… like forests our infrastructure systems have the potential to be modular, networked, decentralized, responsive, and resilient

Decentralization is a technological, architectural, political, and philosophical topic that I think underlies most if not all of our coordination efforts, whether in the realm of making the net better, or battling climate change. Because it is such a wide-ranging topic, entangled in so many debates, I have struggled to solidify my personal take and capture it in a succinct blog post. The lack of a single canonical answer to “how much decentralization” makes it a slippery topic. It’s a valuable principle to guide our decision-making and I am glad to see it enshrined here in an RFC so that standards organizations like the IETF can take it under consideration as we build the next generation of internet infrastructure. Mark Nottingham has done an excellent job writing such balanced and accessible memo in a handy, shareable document to get people familiarized with the concept and its nuance. That link again: RFC 9518: Centralization, Decentralization, and Internet Standards