The Death of ‘Decentralization’
Note: An abridged version of this post was originally syndicated on Token Daily
“The gatekeeper sees that the man is already dying and, in order to reach his diminishing sense of hearing, he shouts at him, “Here no one else can gain entry, since this entrance was assigned only to you. I’m going now to close it.”
~ Franz Kafka, The Gatekeeper.
Decentralization is a myth
Now that I have your attention, I’m asking you to try to limit the number of times you use “decentralization” in conversation, marketing materials, and presentations. The reason for this is the fact that “decentralization” has been reduced to a marketing buzzword, and is losing meaning through its misuse. Your project or idea may have some qualities that decentralization facilitates (censorship resistance, immutability, accessibility, etc) — but it is not “decentralized.” This post has been inspired by a few conversations and two articles that have recently surfaced.
The first article is Nic Carter’s Blockchain is a Semantic Wasteland in which he asks individuals to be more honest in their definitions of “blockchain” to prevent the promotion of marketing buzz over actual development and breakthroughs. Although blockchain has “become a Schelling point,” it would be wise to eventually abandon it or as I like to think, tone it down in order to get back to reality and understand what really makes a blockchain special. For the remainder of this article, I will be policing my words in order to use better phrasing for a bunch of directed blocks.
The second, Qiao Wang’s post entitled Cryptonetworks and the Theory of the Firm, brings back Ronald Coase’s work by explaining how cryptonetworks exist to complement both firms and markets. In this case, search and information costs are subsidized by incentivized recruitment, bargaining costs are subsidized by the transparency and forkability of the code, and enforcement costs are subsidized by decentralized consensus and programmability.
Most of us will eventually capitulate (or already have) to the notion of using a firm. For some it’s an exchange, for others, it’s a third-party wallet. These firms provide a layer of convenience in exchange for a slight bit of trust and centralization. This idea isn’t new and was even mentioned by Hal Finney in the form of “Bitcoin backed banks,” where they would issue their own cash currency which could be redeemed for bitcoin. The banks would then settle net transfers using bitcoin transactions. However, the importance of all this is the option at a base level: the individual’s ability to choose their degree of centralization based on their preferences.
Decentralization isn’t a “bad” word — it’s just a marketing phrase. The ultimate goal of most of these networks is to reach some degree of decentralization, but to call something “decentralized” is a bit misleading. Being centralized isn’t a bad thing either, especially if the application is capturing plenty of legitimate value. Think of all the ICOs that have solicited millions of investor dollars with the promise of something they’ll never achieve in an absolute context. There are plenty of characteristics that it can bank on, but instead, it decides to use the phrase in an effort to win over retail investors.
As I’ll cover later, tokens are on the lower end of the totem pole of “decentralization,” and hopefully by the end of this, you’ll think twice about using the word as well. Let’s break down “decentralization” and see what this spectrum is really about.
The Pillars of Decentralization
To break decentralization down, I’m using a few base characteristics: censorship resistance, immutability, verifiability, accessibility, and social consensus. When individuals use the word “decentralization,” they’re typically referring to these characteristics in particular, as they’re consequences of decentralization. However, it’s important to measure each consequence individually, as “decentralization” is often used to broad-brush and generalize in favor of elucidation. In the case of broad-brushing, these consequences are lost and buried under hype.
Censorship Resistance and Immutability
The first pillar of decentralization is censorship resistance. This is measured by the likelihood of an adversary or nation state to censor transactions on a ̶b̶l̶o̶c̶k̶c̶h̶a̶i̶n̶ spockchain. This deals primarily with the ‘when,’ ‘who,’ and ‘why’ aspects of a transaction, as those variables make up the typical reasoning for an attempted blocking of a transaction. In third-party systems, transactions can be limited to certain time windows, may not be able to be sent to particular parties (if there are sanctions or limits on that party — a great example of this is using Venmo to send money to Cuba from the United States), and any undisclosed reasoning in which a nation-state would want to control ‘malicious’ activity.
Censorship resistance also spills into other pillars such as accessibility and verifiability, because you need an exact history in order to verify which party has which assets to begin with, and you need to be able to send transactions without a gatekeeper dictating the types of transactions you can send. Another example of censorship resistance in action was Wikileak’s acceptance of Bitcoin in 2011. Due to the U.S. government blocking Visa, Mastercard, PayPal and other payments processors from dealing with Wikileaks, they still managed to accept donations in the form of Bitcoin.
Associated with censorship resistance is the cost of attacking the actual network. A network may have a degree of censorship resistance, but it is definitely not censorship proof. A network that takes a higher amount of cost and resources to attack has a greater degree of censorship resistance. Again, we’re trying not to deal in absolutes here. Coupled with censorship resistance is immutability — if the ̶b̶l̶o̶c̶k̶c̶h̶a̶i̶n̶ TurokChain was modified either through an adversarial attack or social engineering, then it loses a few points on that scale as well. Just like censorship resistance not being measured in absolutes, nothing in this world is immutable — however, we can assess the immutability of a chain based on how many actors or how much capital it would take to modify transactions.
Verifiability
For verifiability, all blinders must be removed. Is everything open source and auditable — are all the cards on the table for everyone?
If the software is completely open source, congratulations — the code is now subject to scrutiny from you and the greater community running it as well, and everyone is watching their own back in the same system. This leaves you as a runner of open source software with a lower risk because while everyone is looking out for themselves, everyone is technically looking out for everyone else. If someone notices an issue with the code and fixes it, everyone can benefit from that level of self-preservation. Every line is auditable, and you are responsible for the code that you are running (as you have full access to peruse through it).
Part of the verifiability of a ̶b̶l̶o̶c̶k̶c̶h̶a̶i̶n̶ timechain also includes the ability to audit the entire history of the ledger. Can all transactions and the entire ledger history be verified? A great an often ignored example of not being able to audit an entire history comes from Ripple. Ripple, or “XRP” as some have decided to call it recently, has 32,570 blocks missing from it with no chance at recovery. Full nodes can’t acquire the data and nobody has a chance at seeing the entire ledger’s exact history. The reason for the issue was simple: a bug in the Ripple server caused the loss of all of the headers in that period. This loss accounted for an entire week’s worth of history, and the earliest ledger available for XRP is #32570.
Accessibility
The first part of accessibility with ̶b̶l̶o̶c̶k̶c̶h̶a̶i̶n̶s glockchains deals with gatekeepers and the ability to transact on the network. Can anyone transact on the network or are there parties that can possibly prevent this. What does it take to transact from the highest degree of solitude with the elimination of all possible third parties, and are the actors processing those transactions potentially subject to nation-state influence? The goal of b̶l̶o̶c̶k̶c̶h̶a̶i̶n̶s timecubes is to make the action of censoring transactions as costly and as economically irrational as possible. Are there rules to transacting, or are all transactions created equal. In the case of EOS, some transactions are more equal than others.
I’m saying this in reference to the EOS constitution — a document by which the participants on the network must abide by, or be censored by the block producers. Articles included in the constitution include rules on how users can transact on the network, and those pertaining to what they can and cannot do with the network. Article VI in particular, outlines the penalties for a breach in the constitution, including fines, a loss of one’s account, or “other restitution.” A social enforcement process by “leaders” in a ̶b̶l̶o̶c̶k̶c̶h̶a̶i̶n̶ guac-chain sets you back quite a few points on the decentralization scale.
Oh — you also thought I was done with Ripple? Guess again.
The second part of accessibility comes from a user’s ability to run and synchronize a full node. The heavier the weight of the node or greater amount of time it takes to synchronize, the lesser the accessibility of the chain. For example, to sync Ripple’s incomplete history took someone six months and there were only two public endpoints with the entire history available. Accessibility of the system is a core component to decentralization because it doesn’t require a reliance on a trusted party to provide a history of the ledger or validation of your own transactions. Bitcoin on the other hand only requires under 200 gigabytes of storage for the ledger’s entire history, and for me personally, took under a day to synchronize.
Social Consensus
The last part of decentralization comes from social consensus — are all participants actively aware of their choices and can meaningfully contribute individually, or does one individual or foundation have a disproportionate amount of social control. To address this, we must first ask the following questions about the network under scrutiny:
Does your network have an operating company?
Does your network have a foundation?
Does your network have a Justin Sun?
Then you might have a bit fewer decentralization-points than the b̶l̶o̶c̶k̶c̶h̶a̶i̶n̶ LinkedList-Premium next to you.
Humans naturally form hierarchies, as we come from a tribal upbringing. There can be those that have respect in the space, but might not be formally designated “leaders.” For example, I can respect Nick Szabo’s contributions to the cypherpunk canon, but he’s not the CEO of Bitcoin. He didn’t even make it to the google search results page. Typically token projects are open about their teams and operations, but can something that has scheduled maintenance truly be decentralized?
As an aside, those that resort to forms of promoting their own ‘legitimacy’ through meritocratic means not directly relevant to the space are dangerous. Remember those that contribute, not those that put on a show.
“Non-Custodial” Isn’t as Sexy as “DEX” — A Case Study
Nothing illustrates this problem as clearly as “decentralized exchanges.” The phrase “decentralized exchange” has caught on quite well with many people that now refer to them simply as a “DEX” or “DEXes.” However, the typical “decentralized” aspect of their offering has to do with a non-custodial element, whereas typical exchanges hold custody of a user’s assets. This is great, and a step up, but doesn’t justify the fact that these exchanges are not censorship resistant. For example, Paradex is selective about which jurisdictions they operate in due to their fear of regulatory scrutiny.
I was actually pleased to find that Paradex and Radar Relay don’t actually include the word “decentralized” in much of their front-facing materials, with the exception being Paradex’s Wrapping and Unwrapping Ether help article. However, upon looking at two other popular “decentralized” exchanges, DDEX and IDEX, this was not the case. Both exchanges go into maintenance cycles, have centralized listing policies, and can freeze the trading of particular assets at will (as do most of the rest of them). EtherDelta billed itself as a “decentralized exchange” but was more accurate in that it was just an auditable contract that a user could interface with at-will. The order books for these exchanges are on centralized servers because running them on-chain would be absurd.
IDEX, in particular, has the authority on submitting signed trades to the network, which lets them control the transaction ordering. They also take 0.2% taker fees and 0.1% taker fees which goes to their team. There are minimum trade orders for both sides of the trade and has a minimum withdrawal amount for all tokens. DDEX, in particular, has a 0.05% trading fee, and a selective lower trading fee for particular token holders. They have rebate programs for market makers, barriers to entry for listing on the exchange, a front-facing order book that they still broadcast to users, and like IDEX, a team behind the relayer. They are acting as a firm, providing users an easy way to use the 0x protocol to exchange assets, but is still a centralized middle-man. Their only claim-to-fame is the non-custodial element, but that doesn’t sell — “decentralized” does.
Don’t even get me started on “decentralized securities exchanges” that also have a non-custodial element as their main offering.
While We’re at it, Stop Using “Trustless” Too
Just like extending decentralization as a broad spectrum rather than a fixed status, it’s time to start treating “trustless” the same way, and convert it to a system of thinking about minimizing or reducing trust. The concept of something being “trustless” does not exist, just as you’re slowly trusting in my arguments. Just as absolute decentralization will never be achieved, a complete removal of having to trust any party won’t be possible as well. Let’s use Bitcoin here as the strongest example of minimized trust.
Let’s consider the scenario where Bob needs to give Alice a single U.S. dollar. In this first case, Bob will be hand delivering the cash to Alice. We’ve now introduced trust angles between Bob and Alice, and both of them to the U.S. government. On a micro level, Bob trusts Alice to be in the agreed upon location, at the right time, and Alice trusts Bob to do the same. They trust that each counterparty has their intended exchange value on their person. Bob and Alice also both trust that the counterparty won’t rob them upon arrival. On a macro level, they both trust the U.S. government to honor that U.S. dollar’s value and to control it to an extent that won’t harm them — the equivalent of using “thoughts and prayers” in the wake of tragedy.
Now to prevent these issues of trust on a local level and safety, we’ve employed the use of payments processors and banks for Bob to send Alice the dollar. Now Bob can send Alice the dollar from the comfort of his home, and not have to worry about Alice putting a knife in his back. However, we’ve traded safety for a greater amount of trust in this case. We trust the banks to safely store our dollar, the payments processor to handle our transaction, and the government still to honor the value of that dollar. I don’t think Bob and Alice would like to give up the digital element of the transaction, as it creates a convenient and safe method to transact. So how can we now reduce trust in this case?
Let’s start with ending the trustline to the federal government. We can use gold, but that gets tricky and it’s tough to move that digitally for our safety element. In 2008, Satoshi Nakamoto introduced Bitcoin, which gave us value secured through auditable code (feel free to use this for your next white paper introduction). Bob and Alice no longer had to trust the U.S. dollar which had inflation and untraceability, because Bitcoin now has an auditable fixed supply and a ledger to track transactions on, with a complete history of transactions since its genesis. Great — now who takes care of the payments processor and storage part?
Fully minimizing trust on storage and processing payments comes with running your own full node — in this case, software that validates transactions and blocks. Along with this comes the ability to actually run your own full node. There are a few of these, with Bitcoin Core being the most widely used iteration. Trust here is completely minimized because the code is auditable, as it is completely open source. Private key management for the associated wallet is in the user’s hands, which falls in line with “being your own bank.” The user is responsible for the only keys to the safe, and no party needs to be trusted for this process.
As for the payments processing side, this is two-fold between miners who are incentivized to mine blocks onto the longest chain for the sake of a financial incentive, and users to continue to run full nodes to validate transactions. We essentially trust that both parties will continue to do so, and as the network expands, the need for trust in a single entity continues to reduce and minimize itself. In this case, trust in the incentives to continue to be sufficient and trust in the network continuing to expand between hashpower and full nodes. It’s minimized trust, and it’s hell of a lot better than trusting a central party.
Any action a user takes on the storage or validation side by using a firm only increases a dependency on trust. If I choose to use a third-party wallet, I’m not validating my own transactions. If I choose to use a custodian, I’m trusting a third-party to safeguard my funds. If I choose to use an exchange to purchase bitcoin rather than mine my own, I trust that the exchange will execute (regardless of all the trust issues with centralized exchanges). Even using an exchange such as Bisq incurs a degree of trust. Any action you take for the sake of convenience only increases trust, as it’s a continual tradeoff between what is secure, and what is easy. The importance of a system that minimizes trust is giving a user the ability to decide on that scale.
Alphabit Soup
“Decentralization” has been used countless times on websites, whitepapers, marketing materials, and even during presentations as an inaccurate blanket phrase. Decentralization is a moving target — we can take steps toward it, but will never achieve it in an absolute format. Networks can continually increase their censorship resistance through incentivized reinforcement methods, accessibility through open source code and ease of use, verifiability through providing an accurate and complete history with each full node, and limit their social attack vectors through founders stepping away and limiting any control to one entity or organization.
Buzzword bingo is a symptom of an attempt at using one-size-fits-all phrases for very nuanced issues. Let’s try avoiding it for marketing materials and not give the press any more sensationalism to deliver — shall we?
***
For more reads on attempts to concretely measure decentralization, check out the work of Aaron Schneider, and Balaji S. Srinivasan.
Special thanks to Nic Carter, Phil Bonello, Matthew Garza, Tor Bair and the Alpine team for providing feedback on this post.
Follow me on Twitter
Nothing in this article should be taken as legal or investment advice.
