Troubles with the ENS and decentralization

I recently found out about the Ethereum Naming System, or the ENS. In order to understand exactly what the ENS is, we have to understand the problem it's trying to solve.

The way the internet works is fundamentally very centralized. It has to be. When you type in natechoe.dev into your web browser, your web browser uses the Domain Name System (DNS) to convert natechoe.dev into an IP address, and asks that IP address for the natechoe.dev webpage. The DNS has to be internally consistent so the financial advisor Nate Choe can't take the same domain as the programmer Nate Choe. The organization that manages the DNS and makes sure that everything is internally consistent is the Internet Corporation for Assigned Names and Numbers (ICANN). This gives ICANN the power to effectively regulate the entire internet. They don't, they're a politically neutral not-for-profit corporation, but this danger is why the ENS was created.

The ENS works through an Ethereum smart contract to allow human readable domains to get converted into an Ethereum address. You just make a request to the smart contract to create a new domain, and now it's yours. You make another request to the smart contract to set a resolver for that domain, then you make a resolver smart contract that converts domains to Ethereum addresses, which are used instead of IP addresses.

Ignoring all of the environmental, economic, and legal concerns with blockchains, there are still a few computational problems with this system. Firstly, blockchains are incredibly unwieldly. The ENS doesn't actually have a method to do a resolve. You can transfer a domain, you can set various factors related to a domain, but you can't actually do a resolve. Instead, you just look at all the times that a variable was set and figure it out yourself. This makes sense as repeatedly executing a smart contract is incredibly wasteful and expensive, but it also means that you have to go through a decent chunk of the Ethereum blockchain for every single ENS resolve. This means that while a DNS resolve takes just a few dozen milliseconds, an ENS resolve takes several minutes and is very expensive. With something as ubiquitous as domain names, any solution trying to replace the DNS better be at least as fast.

There is one solution to the fact that the ENS is incredibly slow, you could cache every single ENS lookup. This would be an incredibly time-consuming and resource intensive process, so it would probably be outsourced to a few companies, and we've just reinvented the DNS. The fact is that blockchains are just too unwieldy to avoid centralization. This is the reason why NFTs can effectively disappear if an NFT marketplace decides that it breaks their terms of conditions.

The second problem with the ENS is that it's still not free from the ICANN. How do you actually receive the Ethereum blockchain to do an ENS resolve? The math behind blockchains allow you to verify that a blockchain is correct if you have it, but how do you actually receive it? You could receive several hard drives through the postal service, but in all likelihood you're just going to use the internet. Even if you didn't do a single DNS resolve and used raw IP addresses to get the blockchain, a procedure which is incredibly difficult without any centralization, you'd still have to use an IP address, and IP addresses are allocated by the Internet Assigned Numbers Authority (IANA), which is controlled by the ICANN. In order to get rid of ICANN, you'd have to restructure the fundamental foundations of the internet. Every single Windows 98 machine still running a hospital for some reason, every single web server written from scratch, every single website that's been using the same tech stack for 30 years, they'd all have to be updated to use some new hypothetical cryptographically secure protocol which would almost definitely be more computationally expensive by several orders of magnitude.

The fact is that there's an inverse relationship between centralization and cost, and for things as important and ubiquitous as the internet, not only is it worth it to do some centralization, it's inevitable. Once the numbers are allocated, the process of actually connecting to a server is theoretically decentralized such that I could get thousands of miles of cable and connect my computer to thousands of other people's computers, but because that's so incredibly impractical for the average user everybody just centralizes that job into internet service providers. Theoretically, every tech company could build their own servers and host them on site, but that's so difficult to get right and so expensive to get wrong that everybody just centralizes their web hosting to Amazon, Google, or Microsoft. Theoretically, anybody could create their own Distributed Denial of Service (DDoS) protection, but this is such an expensive task that everybody just centralizes their DDoS protection to Cloudflare. The advantages of economies of scale make it cheaper to centralize than to decentralize, and that's not necessarily a bad thing.

In conclusion, the costs of decentralization almost always outweigh its benefits simply due to the removal of economies of scale, and any attempt to decentralize will be halted by its extreme costs.