Vitalik: Should there be a demand-based recurring fee for ENS domains?

Original Title: 《Should there be demand-based recurring fees on ENS domains?》

Author: Vitalik

Translation: Guo Qianwen, Chain Catcher

ENS domains are currently very cheap. Registering and maintaining a five-letter domain costs only $5 per year. From the perspective of an individual registering a single domain, this price is quite reasonable, but when viewed globally, everything changes. In the early days of ENS, people could register all 8,938 five-letter words from a "spelling dictionary" (which included various uncommon words), prepaying for a hundred years of ownership for just two Lamborghinis. In reality, as many have pointed out: almost all five-letter domains are already taken, with many people holding onto them waiting for high-paying buyers. A casual glance at OpenSea shows that 40% of the domains are for sale or have already been sold.

The question is, is this really the best way to allocate domain names? By selling domains at low prices, the ENS DAO clearly generates less revenue, which limits its ability to enhance the ecosystem. The current situation also undermines fairness: being able to purchase all domains at low prices was a good thing in 2017, acceptable in 2020, but will severely impact the ecosystem's operation by 2050. The actual cost of purchasing a five-letter domain ranges from 0.1 to 500 ETH, and the so-called low registration fees do not save users much money—relying on the secondary market will make domain prices higher compared to a well-designed, embedded protocol mechanism.

Is there a better way to allocate domain ownership? Can it bring more revenue to the ENS DAO, ensure that domains are acquired by those who will maximize their use, while preserving the value of ENS—trustworthy neutrality and long-term ownership guarantees?

Question 1: Is there fundamentally a need to compromise between property rights and fairness?

Suppose there are N "high-value names" (such as five-letter words from a spelling dictionary, or any similar category). Assume that each year users grab k names, of which p names are taken by unusually stubborn individuals who will not give them up (p can be small, just greater than 0). Then, after Nk∗p years, no one will be able to obtain high-value names anymore.

This is a two-line mathematical theorem that doesn't require much explanation. But it reveals a key fact: allocating limited resources without time constraints will harm long-term fairness. This situation also appears in land issues, which is why there have been numerous land reforms throughout history, and why many advanced thinkers advocate for land taxes. The same applies to domain names, although early .com holders introduced a large number of .io, .me, .network, and many other domains, briefly alleviating the issue through this "forced dilution."

ENS has formally committed to not adding new top-level domains to avoid polluting the global namespace and disrupting its eventual integration with mainstream DNS, so this dilution is not under consideration.

Fortunately, ENS not only charges a one-time fee for registering a domain but also charges recurring annual fees for maintaining the domain. Not all decentralized domain systems have the foresight to implement these actions; Unstoppable Domains does not, and even proudly promotes its focus on short-term consumer appeal rather than long-term sustainability ("No renewal fees ever!"). The recurring fees in ENS and traditional DNS are a healthy solution to address this unrestricted, excessive "pay once, own forever" model.

At the very least, recurring fees mean that no one can accidentally lock a domain forever due to forgetfulness or carelessness. But this may not be enough. You could still spend $500 to lock an ENS domain for an entire century, and there are certainly some types of domains in high demand, so this price will not stop there.

Question 2: Speculators do not create an effective market

If we acknowledge that the low fixed-rate first-come-first-served model has issues, we often encounter the rebuttal: yes, many names will be bought by speculators, but speculation naturally occurs and is a good thing. This is a free market mechanism where those who truly want to maximize profits are incentivized to resell domains in such a way that the person who can best utilize the domain will acquire it, and their excess returns are merely compensation for this service.

However, it has been shown through academic research that profit-maximizing auctioneers do not achieve maximum social welfare! Quoting Myerson from 1981:

"The seller announces a reserve price of 50, which means they bear the risk of the item not selling (1/2n), even if some bidders are willing to pay more than t0; but the seller also increases their expected revenue because when the item is sold, they can demand a higher price.

Thus, the optimal auction may not be ex post efficient. To better understand why this occurs, consider the example in the previous paragraph, where n=1. Ex post efficiency requires that the bidder can always obtain the item as long as their value estimate is positive. However, this means that bidders will never recognize a value estimate greater than an infinitesimal value, because any positive bid will win the item… In fact, the seller's best strategy is to refuse to sell the item for less than 50."

Maximizing the seller's revenue almost requires them to accept the possibility that the domain cannot be sold, rendering the domain completely abandoned. An important nuance of this theory is that when there is at least one buyer (or at least one buyer's bid is far greater than others), the "seller's interest maximization" auction is the least efficient for the seller; as more buyers compete with each other, this inefficiency decreases. But for all domains, they are in the first type of situation mentioned above. A domain name is essentially the name of a person, project, or company, with a natural buyer: that person or project. If a speculator buys that name, they must set a high price, thus accepting the possibility that although orders may appear, they will never be completed, and profits cannot be maximized.

Therefore, the statement that "speculators capture a large portion of the domain allocation revenue as their reward for ensuring the market operates effectively" is incorrect. On the contrary, carefully designed mechanisms within the protocol can encourage domains to be sold directly at fair prices, whereas speculators can easily worsen the market.

Why support stricter property rights: stable domain ownership can bring positive externalities

For a long time, it has been recognized that overly strict property rights for non-homogeneous assets can lead to monopoly issues. The Harberger tax was initially designed to address this problem through market means: it requires each asset owner to set a price at which they are willing to sell to others, and pay an annual fee based on that price. For example, an annual fee of 0.5% of the sale price. Thus, holders are incentivized to keep the asset at a reasonable price for purchase, while "lazy" holders who refuse to sell will incur losses each year, and hoarding assets without use is often economically unfeasible.

However, the risk of being forced to sell something at any time brings significant economic and psychological costs. For this reason, advocates of the Harberger tax generally focus on industrial property applications, as market participants are more mature. Where do domain names fit on this spectrum? Let's consider the costs of "relocating" businesses in three different cases: data centers, restaurants, and ENS names.

It turns out that domain names do not maintain well. Domain owners are often not sophisticated, the costs of changing domain names are usually high, and the negative externalities of name changes can be significant. The highest value holder of coinbase.eth may not be Coinbase; it could also be a scammer who manages to seize that domain, immediately creating a fake charity or ICO claiming to be operated by Coinbase and asking people to send money to that address. Therefore, imposing a Harberger tax on domain names is not a good idea.

Alternative Solution 1: Demand-based recurring pricing

Currently, maintaining ownership of ENS domains requires paying recurring fees. For most domains, this is just a low fee of $5 per year. The only exceptions are four-letter domains ($160 per year) and three-letter domains ($640 per year). But what if we don't do this, and instead let the market determine the price based on the actual demand level for the domain?

This is different from the Harberger tax, which does not require a specific price to make the domain immediately sellable. Instead, the pricing process will be driven by the bidders. Anyone can bid on a specific domain, and if they keep their bid open long enough (e.g., 4 weeks), the valuation of the domain will rise to that level. The annual fee for the domain will be proportional to the valuation (e.g., it could be set at 0.5% of the valuation). If no one bids, the fee may decrease at a constant rate.

When bidders submit their bid amounts into a smart contract for bidding, the asset owner has two choices: they can accept the bid or reject it, but rejecting it means they may have to start paying a higher price. If the bidder's bid exceeds the actual value of the domain, the owner can sell it to them, making the bidder pay a hefty price.

This attribute is important because it means that making domain holders "unfortunate" is risky and costly, and may even ultimately benefit the original victims. If you own a domain, and a powerful entity wants to harass or censor you, they can attempt to place a very high bid on that domain, significantly increasing your annual fee. However, if they do this, you can directly sell it to them and collect a large reward.

Compared to the Harberger tax, this provides more stability and is more suitable for newcomers. Domain owners do not need to constantly worry about whether they have set the price too low. Instead, they can relax and just pay the annual fee. If someone bids, they can decide within 4 weeks whether to sell the domain or continue holding it and accept a higher fee. But even so, this may not provide enough stability. To go further, we need a compromise solution.

Alternative Solution 2: Demand-based capped recurring pricing

We can modify the above proposal to provide domain holders with stronger guarantees. Specifically, we can try to offer the following attributes.

Strong time-limited ownership guarantees : For any fixed number of years, there is always a fixed amount that can be calculated, which you can prepay to unconditionally guarantee ownership of the domain for at least that number of years.

In mathematical terms, there must be some function y=f(n), where if you pay y dollars (or ETH), you will receive a hard guarantee that, regardless of what happens, you will hold that domain for at least n years. This may also depend on other factors, such as what has happened to the domain previously, as long as these factors are known at the time of registration or extending the domain transaction. Note that the maximum annual fee after n years will be the derivative f'(n).

The new price after bidding will be capped by the implied maximum annual fee. For example, f(n)=1/2*n^2, so f'(n)=n; if you receive a bid of $5 after 7 years, the annual fee will rise to $5, but if you receive a bid of $10 after 7 years, the annual fee will only rise to $7. If there are no bids to raise the fee to the maximum value within a certain time (e.g., a whole year), n will reset. If someone bids but is rejected, n will reset.

Of course, we have a very subjective standard that f(n) must be "reasonable." We can propose compromise suggestions by trying different function shapes:

Please note that the amounts in the table are merely the theoretically highest amounts required to guarantee holding the domain for that number of years. In practice, almost no bidder for a domain is willing to bid very high prices, so almost all domain holders will ultimately pay fees far below the maximum price.

One of the fascinating aspects of "capped annual fees" is that some versions of it are strictly more favorable to existing domain holders than the current situation. In particular, we can imagine a system where unbid domains do not need to pay any annual fees, while bids can raise the annual fee to a maximum of $5 per year.

Demand from external bidding clearly provides some signals indicating the value of the domain (i.e., how much the owner is willing to spend to exclude others and maintain control over it). Therefore, regardless of what you think about the level of fees required to maintain a domain, you should choose some demand-based fee parameters.

I still advocate for the superlinear function f(n), where the maximum annual fee rises over time, as a good idea. First, paying more for long-term security is a common feature of all economies. Fixed-rate mortgages typically have higher rates than variable-rate mortgages. Offering longer lock-in periods for deposits allows you to earn higher interest; this is the compensation banks pay you for providing long-term guarantees to them. Similarly, long-term government bonds usually have higher yields. Second, annual fees should adjust with any market value of the domain; we just don't want them to fluctuate too quickly.

The value of superlinear f(n) can still reasonably provide strong guarantees for ownership over a considerable time frame: using a linear rate growth formula f(n)=p0n+15/2n^2, you can ensure 25 years of domain ownership for just $6,000 (or $120 per year), and you will almost certainly pay less. For anti-censorship services, the ideal "register and forget" model remains very viable.

Conclusion

Weakening property norms and increasing fees is psychologically unpleasant for many. Even if these fees can bring significant economic benefits, even if you can convert fee revenues into UBI and mathematically show that most people will gain net economic benefits from your proposal, the situation remains the same. Increasing congestion fees in cities is difficult, even when people are very clearly aware that the only two options are to pay congestion fees in dollars or painfully crawl through slow traffic, paying the cost of wasted time and diminished mental health. Despite the fact that land value taxes are among the most effective and least harmful taxes in many ways, they are still difficult to adopt. In my view, Unstoppable Domains' public pride in claiming "never any renewal fees" is very shortsighted, but it clearly has some effect. So why do I think it is possible to increase fees and conditions for domain ownership?

The cryptocurrency space will not solve the deep challenges humanity faces in political psychology—humanity has failed for centuries. But we do not necessarily have to solve the problem; I see two possible answers that may bring hope for success from a realistic perspective:

The first is democratic legitimacy: proposing a genuinely compromise solution that satisfies enough people, possibly even improving the situation for some existing domain holders (not just potential domain holders).

For example, we could implement demand-based annual fees, with a maximum fee of $640 per year for domains up to 8 letters long, and $5 per year for longer domains, with no fees if no one bids. Under such a proposal, many ordinary users would save money.

The second is market legitimacy: legitimacy does not require overturning people's expectations of the existing system; rather, it involves creating a new system (or subsystem).

In traditional DNS, this could be accomplished simply by creating a new top-level domain that is as convenient as existing top-level domains. In ENS, there is a clear idea of only sticking to using .eth to avoid conflicts with the existing domain system. Using existing subdomains is not entirely feasible: foo.bar.eth is much worse than foo.eth. One possible middle ground is for the ENS DAO to completely hand over single-letter domains to projects that run trustworthy neutral markets for their subdomains, as long as they give at least 50% of the revenue to the ENS DAO.

For example, perhaps x.eth could apply one of my proposed pricing schemes to its subdomains, while t.eth could implement some mechanism that gives the ENS DAO the right to forcibly transfer subdomains for anti-fraud and trademark purposes. If it is to serve as some alternative to foo.eth, foo.x.eth is just barely acceptable.

If changing the pricing of ENS domains is not possible, then a market-based approach should be strongly considered, explicitly encouraging markets with different rules in subdomains.

For me, the cryptocurrency space is not just about currency; I acknowledge that my interest in ENS does not revolve around the concept of domain ownership, which is similar to property rights, unconditional and infinitely strict. Instead, my interest in this space is more about trustworthy neutrality and property rights that are strongly protected, especially against political censorship and arbitrary, targeted interventions by powerful entities. Nevertheless, strong ownership guarantees are still very important for the operation of the domain system.

The mixed suggestions proposed above are my attempt to maintain fully trustworthy neutrality, continue to provide high levels of ownership protection, while increasing the cost of domain occupation, bringing more revenue to the ENS DAO, enabling it to commit to important public goods, and giving those who cannot obtain their desired domains a chance to acquire one.

Special thanks to Lars Doucet, Glen Weyl, and Nick Johnson for discussions and feedback on various topics.