What does it mean to say that a cryptonetwork is decentralized? A quick Twitter search suggests that it has to do with how many computers are running in it, the number of miners keeping it secure, their respective hashing power, or the degree to which such things are all geographically strewn about. It’s true that these things have often been correlated with decentralization, but they are a far cry from its definition.
Some have gone as far as to argue that the term has outlived its own usefulness; that it has become fuzzy and come to mean one too many things to mean anything at all. Admittedly, that does often feel true. But I argue that the term is still useful and can be saved. It helps us talk and reason about one of the most important aspects of networks: their power structure. So rather than ditching the word, we should focus it and reclaim it.
Last year, Vitalik suggested that “decentralization” can be used to describe three different things about a cryptonetwork: (1) its physical architecture, (2) its logical data model, and (3) its power structure. I argue that only the third of these uses for the word is useful. It is this use that most people have in mind when casually expressing concern that one network or another is “too centralized”. And, it gets to the heart of what cryptonetworks are all about––they matter because they unlock new mechanisms for decentralizing power.
But what about Vitalik’s two other uses? It’s important to talk about those things too. However, better terms in Computer Science already exist to talk about distributed architectures and logical data models. Terms like redundant and fault tolerant better describe what Vitalik called “architectural decentralization”, and adjectives like unifiedare better suited than “centralized” to talk about the data model of blockchains. So let’s narrow and sharpen the way we think of decentralization; it is all about power and control.
In that spirit, I propose the following definition:
The degree to which power and control over a network is distributed across a large, representative base of independent human actors.
The words large, independent, and representative are key. The set of actors with power over the network must be large because otherwise the network is vulnerable to centralizing forces like collusion and bribery. A large number of actors alone is of no good, however, if their actions happen to already be under the influence of a single company or person. Those actors must also be independent from one another. But even independence is not enough. A network is still centralized if the interests of those with power (e.g. miners) are misaligned with the interests of those without (e.g. users). In order for a network to be fully decentralized, the interests of those who control it must be representative of everyone in the network as a whole.
This definition is, of course, idealistic. There is no such thing as perfect decentralization, only hopeful aspirations for it. Like all good things in life and engineering, it does not come for free. There are unavoidable tradeoffs and widely different approaches to navigating them. All of this makes the concept of decentralization complex. Not to mention that, because hierarchies of power are often concealed, it is also hard to quantify (though good strides to that end have been made). It is still, however, useful as a concept.
Let’s make this usefulness more concrete by breaking down the kinds of things it can refer to. There are four different kinds of power that people can wield over a network. Each one of them is a force that can lead it to become centralized. Let’s call them The Four Horsemen of Centralization (:
The gatekeepers of a network are the sentries who control access. They are those with the power to decide who gets to participate. The word “participate” here is deliberately left abstract. It could refer to a user’s ability to use the network, to contribute to it and get paid (e.g. as a miner or validator), or to any other kind of interaction with the network.
Ideally, a fully decentralized cryptonetwork would have no gatekeepers; or, in other words, it should be permissionless and censorship resistant. Eliminating gatekeepers completely, however, is unrealistic. Let’s look at some examples.
On one end of the spectrum, a fully centralized network like say, Facebook, clearly has a single and all-powerful gatekeeper––Facebook. It has absolute control over who gets to participate. This is the obvious base case. The presence of gatekeepers in cryptonetworks is often more subtle.
Certain networks, for example, might appear to be decentralized, but actually rely on a kind of master list (known as a whitelist) that dictates who gets to participate. Being or not being on the list might completely determine whether you get to be user or a miner on the network. If the whitelist is under the control of single group or individual, then the network has a centralized gatekeeper.
Moving along the spectrum: Cryptonetworks based on Proof-of-Work (PoW) like Bitcoin and Ethereum are designed to have no gatekeepers. In theory, anyone, anywhere should be able to participate without anyone’s permission. If, however, the hashing power in the network becomes concentrated in hands of only a handful of miners, then that group can begin to exclude transactions from certain users or even entire blocks from other miners. Similarly, the hardware manufacturers who control the production of the chips required to meaningfully participate as a miner can likewise become gatekeepers.
Proof-of-Stake (PoS) networks have some advantage over PoW because no special hardware is required to participate as a validator, but they are also vulnerable to censorship if too much of the stake in the network is controlled by a handful of validators.
And finally, exchanges and other points of integration that connect cryptonetworks to the rest of the world also wield varying degrees of power as gatekeepers for both Proof-of-Work and Proof-of-Stake networks alike.
As you might imagine, the Enforcers in the network are responsible for enforcing the rules. They are the people in the network who actually run the code that secures it. Their work, in most cases, is driven by economic incentives. Miners in PoW cryptonetworks, for example, receive block rewards and fees in exchange for validating transactions and contributing hashing power to the network. The power that Enforcers wield is the kind power that most people are thinking about when they talk about the centralization of mining, or centralization of consensus.
In the ideal case, the power that Enforcers can wield over the network is strictly limited to enforcing the rules of the protocol. If Enforcement power is decentralized (as per our definition above), then independent Enforcers will keep one another in check. But, of course, if too much of this kind of power becomes concentrated in the hands of a small number of people (i.e. centralized), then Enforcers can begin to bend the rules or, in extreme cases, break them completely (as with double spend attacks).
Gatekeepers and Enforcers are often one and the same, but there are cases in which they aren’t. Cryptonetworks that are gated by whitelists, for example, might have a single Gatekeeper (she who controls the whitelist), but many Enforcers (those who are included in the whitelist).
The Architects of a network are those who have the power to change its rules. They are the ones who have influence over the network’s governance. Their presence and integrity is crucial for the network to evolve.
Architects come in many different shapes and forms. Some of them wield power directly by determining what changes make it into the protocol’s codebase(s) or by voting with their tokens whenever governance is done on-chain. Others wield power indirectly through their influence in the network’s community.
As with the other archetypes, it’s important that Architects keep one another in check and for them be representative of everyone else in the network. In many ways, this is the most important kind of power, for it can easily be transformed into the other kinds.
The profiteers in a network are, quite simply, those who profit from its rules. Maybe the rules state that they are entitled to some percentage of the network’s original supply of tokens or, in some cases, to a fraction of the tokens minted thereafter for some time (known as a “founder’s reward”).
This kind of power often manifests itself indirectly through wealth; it affords its holders influence over the three other archetypes. Profiteers also wield some degree of power over the markets in which the network’s crypto-assets are traded, for their activity often moves markets. And, in networks where token holders have the ability to vote on questions of governance, Profiteers might also easily become Architects.
It is undoubtedly true that powerful entities within networks invariably wield some combination of all four kinds of power. It is also true that absolute power of any one kind, can likely be transmuted into power of the three other kinds. All kinds of power must therefore be decentralized in order for the network as a whole to be decentralized in turn.
In the end, the question that matters when thinking about decentralization is the following:
Who exactly do you have to trust to believe that your interactions with a network will be fair?
The vision that underlies the crypto movement suggests that the answer to this question should be “nobody”; that is, that the network should function in a way that is completely “trustless” and self-policing. But of course, in practice, there is no such thing as perfect decentralization. There is always someone, somewhere who wields a little bit too much power and has to be trusted, at least to a degree.
One thing is clear: decentralization is a high-dimensional concept. It is important for us to figure out ways to clearly talk about the different kinds of power that people can wield over a network. It will clarify our trust assumptions and will, thereby, help us build better cryptonetworks.
Ali Yahya is a General Partner at Andreessen Horowitz where he invests in crypto. Before joining Andreessen Horowitz, he was a software engineer and researcher at Google X, where he worked on the Everyday Robot project, which aims to build a robot that is smart and affordable enough to be viable as a consumer product for the home. He also worked at Google Brain, Google’s Artificial Intelligence lab, where he was a core developer on TensorFlow, Google’s primary library for machine learning.
The views expressed here are those of the individual AH Capital Management, L.L.C. (“a16z”) personnel quoted and are not the views of a16z or its affiliates. Certain information contained in here has been obtained from third-party sources, including from portfolio companies of funds managed by a16z. While taken from sources believed to be reliable, a16z has not independently verified such information and makes no representations about the current or enduring accuracy of the information or its appropriateness for a given situation. In addition, this content may include third-party advertisements; a16z has not reviewed such advertisements and does not endorse any advertising content contained therein.
This content is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. You should consult your own advisers as to those matters. References to any securities or digital assets are for illustrative purposes only, and do not constitute an investment recommendation or offer to provide investment advisory services. Furthermore, this content is not directed at nor intended for use by any investors or prospective investors, and may not under any circumstances be relied upon when making a decision to invest in any fund managed by a16z. (An offering to invest in an a16z fund will be made only by the private placement memorandum, subscription agreement, and other relevant documentation of any such fund and should be read in their entirety.) Any investments or portfolio companies mentioned, referred to, or described are not representative of all investments in vehicles managed by a16z, and there can be no assurance that the investments will be profitable or that other investments made in the future will have similar characteristics or results. A list of investments made by funds managed by Andreessen Horowitz (excluding investments for which the issuer has not provided permission for a16z to disclose publicly as well as unannounced investments in publicly traded digital assets) is available at https://a16z.com/investments/.
Charts and graphs provided within are for informational purposes solely and should not be relied upon when making any investment decision. Past performance is not indicative of future results. The content speaks only as of the date indicated. Any projections, estimates, forecasts, targets, prospects, and/or opinions expressed in these materials are subject to change without notice and may differ or be contrary to opinions expressed by others. Please see https://a16z.com/disclosures for additional important information.