diff --git a/src/lib/Heading.svelte b/src/lib/Heading.svelte new file mode 100644 index 0000000..ed70e2c --- /dev/null +++ b/src/lib/Heading.svelte @@ -0,0 +1,15 @@ + + + + {text} + + diff --git a/src/routes/writing/+page@.svelte b/src/routes/writing/+page@.svelte index 535e8c3..80da393 100644 --- a/src/routes/writing/+page@.svelte +++ b/src/routes/writing/+page@.svelte @@ -1,5 +1,10 @@ + + +

+ Special thanks to Soumya Basu, + Chad Fowler, + Benjamin Funk, + Patrick O'Grady and + Myles O'Neil for feedback and review. +

+ +

+ 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": +

+ + + +

+ 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:

+ +
    +
  1. + A: A permissionless system with four validators where no single party can influence the system +
    2. +
  2. + B: A permissionless system with a million validators, but where one party can influence which + apps are allowed to operate on the system +
    3. +
+ +

+ 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: +

+ +
    +
  1. + Inclusion: The time from when a transaction is created to when it is guaranteed that it will + eventually be executed +
    2. +
  2. Execution: The time from when a transaction is created to when it is actually executed
    3. +
+ +

+ Price discovery is defined by Hasbrouck (1995) as + the process by which new information is incorporated into an asset's price. Since the updated price is only known after + a transaction is executed, the efficiency of the price discovery process is tied to execution latency, not + inclusion latency. In essence, execution latency is equivalent to the latency of a price discovery engine. +

+ +

+ 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:

+ +
+ + Anatoly describing Multiple Concurrent Leaders + +
+ +

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 Byzantine faults. This collaborative ordering process comes at the cost of higher execution latency, making permissionless + blockchains suboptimal for tasks like efficient price discovery which benefit from low-latency execution. +

+ +

+ 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 $210B+ market with an accelerating arms race amongst institutional players to issue them. Stripe acquired Bridge for $1.1B to advance stablecoin adoption. Prominent firms like Robinhood, Kraken, and Anchorage Digital launched the Global Dollar Network. Even Donald Trump's crypto project plans to launch its own stablecoin. All of these are centralized stablecoins. +

+ +

+ 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 here. +

+ +

+ 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 in The 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. +

+ +
+ +
+ 1 I am actively working on this at Nuveaux, join us! +
+ diff --git a/src/routes/writing/2025/01/06/decentralization-is-not-the-goal/+page.ts b/src/routes/writing/2025/01/06/decentralization-is-not-the-goal/+page.ts new file mode 100644 index 0000000..e73d208 --- /dev/null +++ b/src/routes/writing/2025/01/06/decentralization-is-not-the-goal/+page.ts @@ -0,0 +1,27 @@ +import type { MetaTagsProps } from 'svelte-meta-tags'; + +export const load = () => { + const title = 'Decentralization is not the goal - Dhruba Basu'; + const description = 'What does "decentralization" mean and is it always a good thing?'; + const publishedTime = new Date(Date.UTC(2025, 0, 6, 16)); + + const pageMetaTags: MetaTagsProps = { + title, + description, + openGraph: { + type: 'article', + title, + description, + article: { + publishedTime: publishedTime.toISOString(), + modifiedTime: publishedTime.toISOString(), + authors: ['https://x.com/dhrubabasu_'], + }, + }, + }; + + return { + date: publishedTime, + pageMetaTags: Object.freeze(pageMetaTags), + }; +}; diff --git a/src/routes/writing/2025/01/06/decentralization-is-not-the-goal/multiple-concurrent-leaders.jpg b/src/routes/writing/2025/01/06/decentralization-is-not-the-goal/multiple-concurrent-leaders.jpg new file mode 100644 index 0000000..c6040bb Binary files /dev/null and b/src/routes/writing/2025/01/06/decentralization-is-not-the-goal/multiple-concurrent-leaders.jpg differ