Background - The RIBLT checksum, its linearity over F2 (yangl1996/riblt#3), why AdHash does not repair it, and keyed and group-based alternatives.
Problem statement - The adversary, the beacon-salted setting and its reference work factor, per-namespace curves against precomputation, curve selection and certificates, design criteria and scope.
Workload - The multiset digest, the operations a RIBLT performs on it, and their counts in repeated reconciliation.
Candidate constructions - The fields, curve models, accumulators, encodings and hash maps compared, and the models excluded.
Security considerations - Nominal rho costs per family, extension-field attacks, the limits of the security evidence, and the known weaknesses.
Measurement methodology - What the benchmarks measure, the decision table and dominance rule, how to run checks, benchmarks and reports, and how results are published.
Field and curve code generation - Static instruction counts of field and point operations across 24 compiler targets and CPU models, from the asm-report run at revision 2d6c7e3c.
Summary across runs - Measurements, batching, build comparisons and selection cost.
Annotated literature - For each key of references.bib, what the work contributes to the evaluation of ECMH-based RIBLT checksums over per-namespace curves; in the book, headed by the formatted entry.
Repository organization - Where the field arithmetic, curves, group capabilities, benchmarks, reports, published results and reference programs live.