Frozen measurement artifact for Paper 1 (rev1)

Frozen artifact accompanying the manuscript Why Additive Elliptic-Curve Set
Commitments Do Not Help Ethereum Plasma: A Negative Result from an End-to-End
UTXO Implementation.

This release archives the implementation, the exploit and property tests, the
analysis pipeline, and the complete run-level measurement data behind every
number in the paper. No figure in the paper was transcribed by hand; all tables
and figures regenerate from the archived data.

Measurement datasets

Two frozen code versions were used, and each dataset is reported against its
own commit. They are stored in separate directories and are never pooled.

Experiment	Tag	Commit	Run identifier
E1 commitment cost, E2 per-operation gas, E4 exploit and property tests	paper1-rev1-frozen	949ccf744bb7	20260916-161911-949ccf7
E3 throughput	paper1-rev1-frozen-e3	d4842b9203cb	20260917-091033-d4842b9

The second tag exists because the first throughput harness kept one local node
alive across all runs, so the chain grew monotonically and later cells were
measured against a larger state than earlier ones. The corrected harness starts
a fresh node for every run. No contract and no measured quantity differs between
the two tags.

Contents

• contracts/ — the evaluated L1 and L2 contracts, the seven commitment
  variants, the exploit tests (A1, A2) and the property tests (witness vacuity,
  non-binding), and the parity tests proving that the as-submitted and
  memory-optimized implementations produce identical digests
• data/raw/ — run-level JSONL for every experiment, append-only
• data/receipts/ — 120 Sepolia transaction receipts; every recorded gasUsed
  was verified against its receipt
• analysis/ — aggregation, statistics, table and figure generation
• ANALYSIS_PLAN.md — the statistical plan, frozen before the campaign, with
  Amendments 1–4 recording every deviation and the reason for it
• docs/EXPERIMENT_PRD.md — the experiment specification
• legacy/ — the pre-restructuring implementation, retained for provenance

Reproducing the paper's tables and figures

See the "Regenerating the paper's tables and figures from the frozen data"
section of the README. Regeneration needs no RPC endpoint and no testnet funds;
running make tables twice produces byte-identical output.

Correction notice

The performance figures in earlier versions of this repository and in the
earlier version of the manuscript are withdrawn. They were produced by an
implementation whose block-commitment cost was dominated by avoidable
per-element scalar multiplication and per-operation memory allocation, and by a
throughput metric that counted logical UTXO operations as transactions.
Throughput is now reported as hot-path operations per second, measured on a
single local node, and is explicitly not an end-to-end Plasma throughput figure.