+ Special thanks to
+ The opinions expressed in this post are my own and do not necessarily reflect the views of the reviewers. +
+ ++ Speak to any crypto person and you'll almost certainly hear "decentralization" touted as a major benefit. But what + does it mean and is it always a good thing? +
+ ++ Decentralization is rarely defined in an objective way. Even Merriam-Webster only offers a vague definition +
+ +: the dispersion or distribution of functions and powers
+ ++ By dealing in the subjective, we're cursed to engage on beliefs as opposed to data. Look no further than + the decade-long debate amongst crypto enthusiasts and academics about whether certain networks are "decentralized + enough": +
+ +-
+
-
+ 2014: Bitcoin is centralized because GHash.io's mining pool controlled
>51% of Bitcoin's hashpower +
+ -
+ 2017: Balaji offers the
minimum Nakamoto coefficient as a first step to quantify decentralization +
+ -
+ 2018: Cornell researchers use
network connectivity to quantify decentralization +
+ -
+ 2022: Solana is centralized because you need
beefy nodes to run a validator +
+ -
+ 2023: Ethereum is centralized because Lido's liquid staking approached
33% of the total staked ETH +
+ -
+ 2023: L2s are centralized because they use
centralized sequencers +
+ - + [Insert any year]: [Insert your favorite network] is centralized because most validators are on + AWS + +
+ These arguments often get lost in the technical details of the network, missing the point of decentralization. The + true measure of decentralization lies in the power that centralized intermediaries have in a system. +
+ +Consider two systems:
+ +-
+
- + A: A permissionless system with four validators where no single party can influence the system + +
- + B: A permissionless system with a million validators, but where one party can influence which + apps are allowed to operate on the system + +
+ How should we evaluate which system is better? Systems, like the ones above, can be centralized in some aspects and + decentralized in others. We can understand the differences by examining different dimensions, such + as: +
+ +-
+
- User: Who can use the system? +
- Operator: Who can enforce the rules of the system? +
- Power: Who can determine the rules of the system? +
Applying these dimensions to the two systems, we see they rank differently:
+ +-
+
- + User: A = B. Both A and B are permissionless, meaning anyone can use them. + +
- Operator: A < B. A has fewer validators/enforcers than B. +
- + Power: A > B. A has no central authority, while B is controlled by a single party. + +
+ Despite having fewer validators than B, A beats B in the power dimension because no single party can control + application operation. Increasing the number of A's validators at the cost of centralizing power in a single party + would be a bad application of decentralization since it undermines A's power distribution. +
+ ++ Centralizing certain functions, like operator control, may prevent any party from influencing the system, creating a + better environment for users. While it may seem counterintuitive, in some cases, we should tolerate + centralization when it leads to a better distribution of power. +
+ ++ There's a trend amongst crypto enthusiasts to bring everything on-chain. While this is an admirable ideal, it + clashes with the practical drawbacks of decentralized systems that preclude this on-chain utopia. This can be seen + through two growing narratives: Nasdaq and Stablecoins. +
+ ++ Solana co-founder Anatoly Yakovenko has a clear vision for the network: decentralized Nasdaq, a global, + permissionless price discovery engine capable of competing with centralized exchanges. +
+ ++ To analyze the viability of this vision, it's important to first understand price discovery in relation to + transaction latency. Generally, there are two types of transaction latency: +
+ +-
+
- + Inclusion: The time from when a transaction is created to when it is guaranteed that it will + eventually be executed + +
- Execution: The time from when a transaction is created to when it is actually executed +
+ Price discovery is defined by
+ Reviewing the market microstructure literature tells us that a price discovery engine with lower latency is superior + to one with higher latency: +
+ +-
+
-
+
Hasbrouck and Saar (2013) found that "increased + low-latency activity improves traditional market quality measures—decreasing spreads, increasing displayed depth in + the limit order book, and lowering short-term volatility" +
+ -
+
Jones (2013) ;Brogaard, Hendershott, and Riordan (2013) ;Carrion (2013) all found that HFT market + making enhances market quality by reducing spreads and enhancing informational efficiency +
+
With that in mind, let's take a look at how Solana aims to be competitive:
+ +Multiple Concurrent Leaders/Proposers (often stylized as MCP) is the core idea.
+ +
+ Anatoly's tweet is partially correct. MCP can greatly improve Solana's inclusion latency, but it
+ cannot lower Solana's execution latency. Broadly speaking, permissionless blockchains
+ emphasize censorship resistance through a decentralized validator set, where a majority of the
+ validators must agree on transaction ordering to protect against
+ As we learned above, a price discovery engine with lower latency is superior to one with higher latency. MCP + therefore falls short of making Solana's execution latency competitive with the execution latency of centralized + exchanges. +
+ ++ Even though Solana cannot compete with centralized exchanges on execution latency, Solana wins in power + distribution. This characteristic is especially valuable in the broader context of open financial systems. It + enables Solana to dominate financial functions like asset issuance, not price discovery. +
+ +
+ Stablecoins are often hailed as the "killer use case" for cryptocurrency. They are now a
+ Assuming decentralized stablecoins like DAI could technically function, would most people actually trust them over + centralized stablecoins? Unlike USDC, where Circle can recover lost or incorrectly transferred funds, decentralized + stablecoin transfers are irreversible, making them far riskier for the average user. Because of this, centralized + stablecoins will likely remain the dominant form of currency movement. People feel safer knowing they have someone + to contact (or even sue) when problems arise. +
+ ++ This reality means that any amount of decentralization in a blockchain is largely meaningless for centralized + stablecoins. Whether USDC is issued on a blockchain with a billion solo stakers or a blockchain with a single + centralized sequencer, it's ultimately secured by Circle. If Circle chooses not to honor your USDC on a given + network, you lose access to those dollars. +
+ ++ To take it further: When Ethereum undergoes a hard fork, the canonical chain is determined only by centralized + entities, not by any solo staker. This is not a hypothetical scenario. After the DAO hack in 2016, the Ethereum community + was split between a new chain reversing the hack and the original chain which did not. +
+ +
+ Ethereum co-founder Vitalik Buterin supported the new chain. Coinbase, Kraken, and Bitfinex all supported the new
+ chain as the canonical chain. The new chain then came to be known as Ethereum (ETH) while the original chain is
+ referred to as Ethereum Classic (ETC). A good summary of the entire history can be found
+ Since centralized entities control liquidity and accessibility for a large number of users, they effectively + determine the practical value of each chain. If Ethereum forked again today, whichever fork Circle honors the USDC + on will be the canonical chain. +
+ ++ While decentralization isn't suitable for every use case, we've collectively done an incredible amount of work to + use it in ways that empower individuals. It's important to recognize the tremendous progress we've collectively made + in reducing reliance on centralized authorities. +
+ ++ With self-custody support, blockchain users have full unilateral control over their assets via a private key they + can generate themselves. This gives blockchain users the exclusive right to claim and manage their + assets without needing permission from anyone. There's still a long way to go before self-custody becomes a + practical alternative to holding assets on centralized exchanges or custodians, especially when it comes to + accessing centralized liquidity1. +
+ ++ With the networks we're building, the barrier for asset issuers to onboard and seamlessly interoperate is + continuously getting lower: +
+ +-
+
- + Solana: A permissionless single chain with high bandwidth and low latency, designed to support + all users with native interoperability + +
- + Avalanche: A permissionless network of networks that allows anyone to create their own network, + with the ability to interoperate with other networks (as long as both networks agree) + +
- Canton: A permissioned network of networks laser-focused on an institutional approach +
+ These different models show that there isn't a single way to decentralize finance. Instead, we can adjust various + dimensions of decentralization to suit the needs of different use cases and users. This flexible, practical approach + is what will drive real distribution of power. +
+ +++ ++ Decentralization measures are like some potent drugs, however: when prescribed for the relevant illness, at the + appropriate moment and in the correct dose, they can have the desired salutary effect; but in the wrong + circumstances, they can harm rather than heal. +
+ + — Rémy Prud'homme inThe Dangers of Decentralization (1995) +
+ As we continuously experiment in the blockchain and crypto space, we must remain focused on the goal of + decentralization: the distribution of power. We don't use smart contracts to build + an app for the sake of using smart contracts; we use them because they enable trustless interactions. In the + same way, decentralization should be applied deliberately when it can contribute to a more open and equitable system. + Without this focus, we risk getting hooked on decentralization itself. +
+ ++ +