Design a Blockchain DNS Alternative (Name Service)

Problem statement

Design a blockchain-based alternative to DNS: a decentralized name service that maps human-readable names such as alice.eth or payment.base.eth to wallet addresses per coin type, contract addresses, content hashes, and arbitrary text records. The system must support registration and renewal through a bidding or first-come mechanism, ownership transfer and subdomain assignment, resolution for wallets and dApps at DNS-like read volume, expiry with recovery semantics, and a dispute path for squatting and trademark conflict without a central registrar authority.

The defining property is that the source of truth for ownership is a public ledger whose history can reorganize, whose writes cost gas and settle asynchronously, and whose reads are free but latency-bound by RPC. DNS, by contrast, replicates a zone file through a cached hierarchy with authoritative servers. Therefore this design is not DNS with a different database: it is a cache-coherence problem over a probabilistically final truth, plus a namespace-security problem (homoglyphs, squatting, front-running), plus a governance problem (who may add a TLD, change pricing, or revoke a name).

Why the problem is distinctive

A DNS answer can be cached for an TTL because zone updates are rare and authoritative servers are stable. A blockchain name answer can be invalidated by a transaction in the next block or, worse, resurrected by a reorg that reverses the invalidation. Value-bearing decisions (sending funds to a resolved address) must therefore never trust a stale cache, while cosmetic decisions (displaying a profile name) may. The design separates resolution freshness classes the way the safety plane separates motion from mission progress in robotics: correctness-critical reads go to chain or finalized height; convenience reads go through labeled caches.

Public operating baseline versus design assumptions

Public evidence shows the category is real and large. ENS's public site describes an open naming layer with more than 35 million registered names including subnames, while on-chain .eth registrations are on the order of millions, with Dune-derived reporting of 2.2 million new .eth names in 2022 alone. [[1]] [[5]] Unstoppable Domains reported four million domains sold with one-time purchase pricing and Polygon minting. [[26]] [[24]] The incumbent DNS industry closed Q4 2025 with 386.9 million registrations across all TLDs, which sizes the ceiling a decentralized alternative competes against. [[44]] These are cited public figures, not requirements for our fictional system.

For capacity planning this answer explicitly assumes: 3.2 million active registered names, 40 million issued subnames, 250 million resolution lookups per day across platform endpoints, 8,000 registrations per day average with 60x drop-day spikes, 60 million reverse lookups per day, and a 12x read peak. Unless a number is tied to a citation, it is a stated design assumption, target, or budget.

The four architectural planes

1. Registry plane: on-chain registry mapping namehash nodes to owners, resolvers, and TTL-like metadata; root and TLD controllers; name wrapper semantics.
2. Resolution plane: universal resolver logic, CCIP-Read gateways, wildcard off-chain resolution, multi-tier caches, indexers, and DNS-bridge exports.
3. Registration plane: registrars (auction, Dutch auction, first-come with rent), commit-reveal anti-front-running, renewals, grace periods, price oracles, subname issuance.
4. Governance and trust plane: DAO or multisig policy, normalization standards, dispute and revocation policy, artifact signing, evidence and audit.

A strong interview answer keeps these planes separate: the resolution plane may degrade to stale-labeled answers for cosmetic use, but the registry plane never admits two owners for one node, and the registration plane never lets a mempool observer steal a name.