# Changelog

## 0.6.0 - 2026-09-22

- Add a streaming Common Crawl domain Web Graph adapter for harmonic-centrality,
  PageRank, and member-host evidence while preserving exact provider fields and
  source-line coordinates.
- Authenticate and validate the complete rank stream but retain only registrable
  domains already asserted by imported public identity sources, avoiding a
  multi-gigabyte runtime catalog whose unrelated rows cannot resolve routes.
- Bind every compact graph projection to the SHA-256 of its sorted candidate
  domains and fail closed when identity evidence, the PSL, the bound scope, or a
  selected graph row is absent.
- Teach scheduled updates to import identity sources before automatically binding,
  downloading, and importing Web Graph evidence with `{candidate_domains}`.
- Strictly allowlist official HTTPS domain-rank objects and retain Common Crawl
  Terms-of-Use attribution without treating authority as ownership, safety,
  reviewer approval, or query popularity.
- Update Rustls to 0.23.45, remediating RUSTSEC-2026-0285 in the dataset
  acquisition path.

## 0.5.0 - 2026-09-13

- Add a streaming ROR 2.1 ZIP adapter with exact schema checks, declared domains,
  website links, names, aliases, external identifiers, status, type, country and
  administrative metadata while preserving ROR identities independently.
- Advance source manifests to v3 with source-bound content/provider checksums,
  explicit direct and upstream lineage, authenticated per-record limits and exact
  provenance for every imported or derived fact.
- Stream multiline Wikidata JSON arrays and typed JSON deltas, including auditable
  tombstones when official-website statements disappear.
- Add compact runtime generations backed by content-addressed cold audit bundles,
  complete verification and export, signed no-deletion retention checkpoints and
  exact active-source binding in generation receipts.
- Add lineage-aware evaluation reports and a reproducible, resource-bounded
  benchmark harness with exact input, output, process and hardware evidence.
- Harden ZIP expansion limits, audit reads, foreign-key restoration, disk-full
  import recovery and output-exhausted benchmark process cleanup.

## 0.4.0 - 2026-09-13

- Add typed full/partition/delta source coverage, active-record masking, a
  receipt-bound coverage digest, active export and complete audit export.
- Separate name, website-edge and entity-equivalence identities so aliases cannot
  inherit routes and unrelated metadata does not invalidate stable edges.
- Replace latest-review admission with authenticated, policy-bound reviewer votes,
  independent identity/group/physical-key quorum, trusted writer acceptance time,
  bounded expiry, risk-class threshold overrides and sticky explicit revocation
  supersession.
- Bind strict reviewer trust roots and review policy into releases and enforce
  publisher identity and physical-key separation.
- Store immutable subject-bound observation batches and add exact/reverse lookup,
  bounded DNS-pinned candidate observation, cache-only replay and structured HTTP,
  redirect, canonical, hreflang, JSON-LD, sitemap, country, DNS, TLS and failure
  evidence.
- Add deterministic evidence bundles, risk-ordered review queues, drift classes,
  revocation-candidate export, schema migrations 4 and 5, 0.3 compatibility
  guidance, architecture decisions and strict end-to-end security fixtures.
- Add publisher-signed cumulative emergency revocation feeds with seven-day
  freshness, rollback-safe continuity, block-only application to older compatible
  registries, and reinstatement validation against an exact full generation.
- Add safe scheduled typed-snapshot supersession and a serialized, bandwidth-capped
  observer service; reject compressed bodies, redirect loops and duplicate update
  coordinates before acquisition.

## 0.3.0 - 2026-09-13

- Snapshot authenticated SQLite bytes into a private file before queries and sign
  sealed receipt bytes, removing mutable-path time-of-check/time-of-use windows.
- Require consumer-supplied reviewer trust roots during signed verification and
  activation; support retired reviewer keys with OpenSSH validity epochs.
- Hash the exact source descriptor stream consumed by parsers and add configurable
  expanded-byte, record, database-growth, query-output and diff-output limits.
- Bind identity decisions to entity metadata, redact descriptions from typed diffs,
  include diff attribution and keep CrUX job IDs stable across output-only limits.
- Migrate writer stores to schema/rules version 3 while allowing activation over
  pinned version 1 and version 2 generations for revocation checks.

## 0.2.0 - 2026-09-13

- Open generations from an authenticated immutable SQLite file descriptor and
  reject sidecars, symlinks and unexpected generation files.
- Authenticate exact reviewer decisions with SSH signatures, retain their proofs
  and re-verify every decision before release signing.
- Explain resolver abstentions with typed statuses and decision counts.
- Add exact entity, URL/domain, source-separated popularity, redacted category and
  registry-statistics audit views.
- Stream typed diffs across source selections, projections, decisions and explicit
  equivalences.
- Add bounded native evaluation with multilingual, regional and deceptive-query
  regressions plus latency evidence.
- Normalize rights-bearing future crawler observations without admitting them as
  identity or ownership evidence.

## 0.1.0 - 2026-09-12

- Initial independent extraction with five rights-reviewed source adapters,
  deterministic normalization, immutable generations and conservative regional
  resolution.