Design and Implementation of the Zero-Query Name System (ZNS): Evaluating Transport-Independent State Replication Against Distributed DNS Querying

The Domain Name System (DNS) operates on a distributed request-response paradigm formulated under the constraints of 1980s network topologies: resolvers issue queries over the network, cache answers for a bounded Time-To-Live (TTL), and re-query upon expiration. While resilient, this model inherently incurs wide-area latency penalties on cache misses, exposes user navigation patterns to intermediate network observers, and fails during network partitioning even when authoritative state has not changed. This paper presents the Zero-Query Name System (ZNS), an alternative architecture that replaces the query-first loop with transport-independent, authenticated state replication. In ZNS, authoritative publishers commit zone records into a versioned, Merkleized state tree authenticated by Ed25519 signatures and domain-separated BLAKE3 hashes. Resolvers selectively synchronize verified state objects and compact binary deltas over multiplexed QUIC streams, resolving client lookups authoritatively from local memory. To facilitate adoption without operating system or browser modifications, ZNS incorporates an RFC 1035 UDP/TCP compatibility gateway that transparently proxies non-authoritative queries. We evaluate an open-source Rust implementation of ZNS across bare-metal hardware and synthetic workload simulators. In physical microbenchmarks, in-process ZNS lookups achieve a median latency of 320 ns (an 81.5× speedup over warm dnsmasq socket lookups), while the RFC 1035 gateway sustains 54,550 queries/second across 8 worker threads with zero packet drops. Incremental base-relative deltas reduce update transmission volume by 60.9% to 96.5% compared to full zone packages, converging across live QUIC relays in 34.7 ms. We also report on a 1,000,000-domain trace simulation, establishing that while selective replication eliminates cache staleness and WAN round trips for active domains, naive unselective fanout on long-tail namespaces incurs unacceptable bandwidth amplification, validating our tiered subscription design. Project Website: https://zns.is-cod.inResearch Prototype: https://github.com/Araan-Sheikh/zns (Private Prototype)Contact: [email protected]