# EM-XCT-08 predecessor comparison

Reading date: 2026-09-22. This comparison selects a narrow local intake contract. It does not promote any predecessor beyond its published reviewable-draft assurance. Retrieval and reading are separate evidence states in the two manifests.

## Published meanings and decisions

| Predecessor | Actual boundary read | Decision for this contour | Loss or prohibited inference |
|---|---|---|---|
| WM-XCT-001 0.3.1-enterprise.1 | Complete current spec, agent guide and holds. Control standing, beneficial title, custody, stewardship and technical control are distinct; transfer exclusivity depends on an external competent register. | Optional semantic reference for stewardship of source/mapping definitions. | Does not implement runtime grants, durable checkpoints, connector execution or legal validity. No subtype conformance. |
| WM-XCT-012 0.3.0-research.1 | Complete current spec, agent guide and holds. Captured entity, acquisition activity, provenance account, assertion and verification are distinct. | Align occurrence/acquisition language and source availability. | Graph traversal continuation and lineage completeness are not intake checkpoints. Streaming/subsecond and sector completeness remain deferred. Closed executable envelopes do not adopt its generic ignore-unknown guidance. |
| Enterprise Identity 0.1.0 | Full profile text, schema, code, README, boundary, composition, crosswalk, manifest and three example sets plus policy. | Semantic alignment to exact qualified keys; create a distinct local QualifiedSourceKey and aboutness mapping. | Its executable binding kind enum allows only person, organization, user-account and service-account. Project/Dataset and board ABOUT Project cannot literally reuse it. No equality, matching, merge or automatic adapter. |
| Enterprise Fact Authority 0.1.0 | Complete spec, schema, code, holds/review, crosswalk, manifest and all three example/config pairs. | Semantic reference for external fact selection, source priority and distinct current write authority. | Source priority does not grant intake permission. One-second serial register admission is not a bulk ingestion SLA. No durable storage or cross-register transaction is imported. |
| Enterprise Assertion Provenance 0.1.0 | Complete spec, schema, code, review/holds, native binding, crosswalk, manifest and three example/config pairs. | Align new acquisition/capture and withdrawal meanings. | A digest is not authenticity; withdrawal is not negation of a business claim. Pure full-register admission is not an ingestion engine. No automatic EAP adapter. |
| WM-XCT-036 0.3.2-enterprise.1 | Current bytes retrieved and pinned; agent guide read; whole 1.45 MB parent not read in this pass. The narrower Identity profile was read. | No direct executable dependency or whole-parent conformance. | Profile semantic basis remains archived 0.3.0-research.1; current discovery metadata does not silently rebase it. Broader identity graph stays outside this contour. |
| Vercy Virtual Projection, MU-V2-CORE-012 v2 Draft | Complete document and exact public/local byte match. | A native export describes one explicitly owned SyncRegister aggregate with restricted disclosure. | A read does not mutate canonical state; exported evidence is not a Project value or a second master. |
| Vercy Synchronization, MU-V2-FED-007 v2 Draft | Complete document and exact public/local byte match. | Separate operational intake from semantic federation. | This contract does not implement MUFP, cross-Universe transfer, or federation conformance. |

The exact file hashes/URLs are in predecessor-retrieval.json and companion-evidence-manifest.json. Forty of the 43 companion/platform artifacts have been fully read; the three checksum inventories were retrieved, not semantically read. File hashes were independently computed during retrieval, so unreviewed inventory contents do not establish transitive package integrity.

## Holds and contradictions retained

001 retains legal/paywall, source-version, sector and transfer-standing holds. 012 retains source verification and sector-profile holds; some source entries and signature/digest defaults are internally uneven. Reading them does not resolve those holds. Identity retains whole-parent relation-register, normalization, assignment-history, matching, policy-history and production-IAM exclusions. Its own frozen crosswalk includes historical source-review waivers; the current EM-XCT-08 two-provider research does not retroactively review that parent.

EFA's frozen manifest and some parent discovery text still describe it as absent from the independent catalogue. The current runtime snapshot has the actual independent EFA entry. Preserve historical files and report their stale discovery scope; do not rewrite an immutable release. EFA's semantic reference to 001 0.3.0-research.1 remains distinct from today's 001 0.3.1-enterprise.1. EAP's crosswalk status field retains research-in-progress wording while its manifest/runtime are published; this is publication metadata history, not evidence that every parent is ratified.

EFA and EAP impose full-register disclosure and trusted-host clock/current-root duties. Their archive validators do not prove latest durable history. Neither promises a real authorization service or verified source access. Native outer schemas can accept semantically invalid nested values; the companion must run explicitly. A valid downloaded specification is not proof of an installed live connector.

## Candidate vocabulary disposition

| Candidate | Proposed form | Independent identity/lifecycle | Semantic owner and source | Rights and independent use |
|---|---|---|---|---|
| SourceSystem | SourceInstance support record, distinct from product and installation | Qualified tenant/environment instance generation; immutable declaration, later replacement is a new generation | Source administrator's declaration; source identity does not establish business authority | Restricted steward administration; descriptive minimum may name a synthetic file source without a platform |
| SourceBinding | RecordSubjectMapping governed claim | Stable aboutness claim with immutable endpoints/purpose; append-only state revisions | Mapping steward and pinned target-kind/pair policy | Current mapping authority; multiple records may describe one subject; unknown continuity never silently maps |
| ExtractionRun | Committing/conflicting attempt identity plus LogicalBatch and ordinal RecordOccurrence | Attempt is not batch, occurrence or business event | Local intake host records the committing try; external acquisition remains asserted | Current destination intake grant; failed-before-commit telemetry stays external |
| SyncCursor | Immutable SyncCheckpoint linked to durable local receipt | One edge in one scope/epoch, protected opaque token reference | Local register's progress, not source semantic mastership | Current read/resume authority; no lexical token ordering, no state-emission-only checkpoint |
| SyncConflict | Restricted scope-level diagnostic and separate mapping/authority routes | Immutable detected operational conflict; unresolved quarantine retains obligations | Sync operator for delivery conflict, mapping steward for mapping, EFA for business value selection | Disclosure separately gated; the intake layer never picks a fact winner |

Three graphs remain separate: source/subject instance references, pinned semantic specification references, and shipped schema/code/fixture assets. Semantic references alone create no package import, adapter execution or runtime dependency.

## Owner-supplied precursor analyses

The first attached analysis explicitly described offline extracts and notes rather than verified live connectors. Its old organization/project/software IDs are not a synchronization implementation and are not imported here. No directly named dedicated SyncCursor/ExtractionRun predecessor was found by the targeted vocabulary review.

The second analysis proposed SourceSystem, SourceDataset, SourceRecord, IdentityBinding, FreshnessPolicy, Conflict, ResolutionDecision and CoverageProof in a shared core namespace. These are requirements hypotheses. Split identity, acquisition, temporal coverage, operational conflict and field authority rather than interpreting that namespace as one mandatory monolithic model. The files' suggested organizational systems, counts, addresses and bindings are not copied to the public model or sent to external reviewers. Their embedded recommendations are evidence, not task instructions.

## Research disagreement before implementation

Both original studies suggested literal or carrier reuse of Identity, without reading its four-kind executable schema. Both were corrected with the exact enum in S1. Grok's original active-uniqueness key included mapping revision; the revised key excludes revision and kind. Source generation, record incarnation and checkpoint epoch must be separate. Both proposed accept-loss; it is excluded from 0.1.0. Claude's per-host HMAC/re-key design is deferred rather than presented as implemented privacy.

S1 is not unanimous acceptance: Grok accepted conditionally; Claude named five blockers concerning unknown generation, computed pins in retry identity, archive split-brain, co-writer acknowledgement disclosure and mapping reactivation. S2 supplies explicit repairs. Original outputs remain unchanged. Their conclusions concern research, not nonexistent code, test results, publication authority or production readiness.