research / distributed systems & protocol theory · v1.0.0
Network-Native Computation
When does moving state resolution into naming, content addressing, and Merkle witness protocols beat shipping raw state — and where is the exact information-theoretic floor?
Question
Protocol substrates (DNS aliases, Git packfiles, content-addressed DAGs, IPFS, AO logs) already perform structural computation during resolution. How do we measure that work honestly without hiding cost in pre-indexing or uncounted round trips?
Theory
Introduces a closed four-term accounting law across setup, transport, local compute, and verification, alongside Theorem 1 & Lemma 1 (Bounded-Probe Structural Ambiguity), proving sharp query lower bounds when structural invariants are not committed in hashes.
Empirical suite
Eight reproducible experiments (Exp01–Exp08) spanning DNS DNAME redirection, content-addressed Merkle deduplication, stock Git v2.55.0 tree-same pruning, IPFS Bitswap/DHT resolution, and AO checkpoint witness replay.
Negative result retained
Exp08 (prospective latency-aware routing under noisy probes) fails to beat static baseline selection and is preserved verbatim in both the manuscript and benchmark runner.