Research: richer Contract C attribution sidecar RC1

[camerontjs-dot]

Classification

Draft Research Infrastructure / alternative Contract C representation evidence. Do not merge. No released Contract C mutation, canonical sidecar standard, successor-version assignment, Decision Engine production change, Contract E, release, tag, promotion, or operational authorization.

Exact lineage

• protected Apparatus base: c3563cff66d2c85dcbf575c693056e2d8e4563d4
• exact base branch: research/contract-c-attribution-sidecar-rc1-base-20260910
• released Contract C authority: 5fe55f9ed5d0ee9f026ca1b077e9d70ce0487ea1
• released validator blob held fixed: 9c75ccfbf2223578a8d1a7bf0c39673b394fbea4
• released schema blob held fixed: b0369de9b5c156322d6787261bbc7658a3b33781
• CAL PR #100 multiplicity head: aa5f0f1313e65e6d31493095214c76d743ca6d89
• CAL multiplicity run: 34506201889, artifact 10163899726, digest sha256:23ba4cb13870f0b2c5f213d3279faede9d17f19e638dea49444249328a0efb19

Question

Can the old sidecar candidate's sole demonstrated defect be repaired by adding an explicit evidence-level causal-form operator and member identities, while leaving Contract C 1.0 bit-for-bit unchanged?

Preserved apparatus failure

Run 34532296461 at 8d2f68a3fad53dc4ef44b8d61fb16675827508d3 did not reach semantics. The no-producer-import guard matched the independent consumer's docstring phrase saying it does not import the producer implementation. The exact-base/research-surface gate had passed.

The correction changed only that workflow grep to match actual Python import statements. No sidecar, consumer, carrier or evaluator semantics changed. This failed run remains an apparatus-invalid record rather than being discarded.

Decisive execution

Run 34532374495: PASS.

• exact tested head: 132a11587af3cfc7d0e05122390babe630cf26f4
• terminal record head at PR creation: f58f531abf6f2c8ab264db41346592b1498514d0
• disposition: SUPPORTED_RICHER_SIDECAR_TECHNICALLY_SUFFICIENT
• artifact: 10174010034
• artifact digest: sha256:94b4e6f3e5497b3c577c66bf869854de8b30a162633c1540a6c8c006b5cd4567
• released Contract C regression: 17 passed, 8 skipped

Frozen carrier and receipt

Released-1.0 carrier:

• SHA-256 sha256:f17903c8d829b3ca29c59a71f91c49635945c9ad54ebe1a10fdb055b2275de09
• completed / not_checkable / unresolved_categorical_relation
• one opaque causal state: basis member
• zero contribution records naming the unresolved evidence
• Contract C alone cannot reconstruct U1/U2.

Richer sidecar:

• schema cal-producer-attribution-sidecar-rc1-v1
• SHA-256 sha256:3819b662f844b9a0cd5516001101a7bf4d93342d8a58426921600227bafd70b1
• receipt producer-attribution:013530def9a0a5407ac65af312baac8d77d23bf3e9d3a9836803cc848712f994
• exact causal passages U1, U2
• role causal_non_deciding
• evidence-level causal form independent_sufficient_alternatives.

The consumer is standard-library-only with respect to the sidecar and imports no producer-side sidecar implementation or CAL semantic code.

Hostile controls

The cohort rejected stale/wrong Contract C hash, result identity and version; proposition/hash/state substitution; wrong/missing Contract-B evidence; duplicate member/ref; unknown role/form; invalid causal cardinality; stale member/receipt identity; noncanonical member order; and external sidecar whole-object digest mismatch.

Reverse producer input order generated identical canonical sidecar bytes.

Bounded inference

For this exact counterexample, widening the Contract C wire is not technically necessary solely to transport the missing unresolved provenance and multiplicity. A separately immutable companion receipt can carry them while released Contract C 1.0 remains unchanged.

That falsifies the technical-necessity claim for the in-band major successor. It does not establish sidecar architectural preference.

Remaining conflict

Contract C's published purpose says it is the decision-agnostic representation of CAL-attributable epistemic/result state required by legitimate downstream consumers. Under this candidate, Contract C alone remains insufficient and a second artifact is mandatory for deep causal reconstruction.

The next decision must therefore be grounded in the original Contract C authority/EDR record: whether required attribution and multiplicity state had to live in Contract C itself, or whether a separately immutable co-transported companion was permitted.

If Contract C itself was required to preserve those demonstrated invariants, this sidecar is architecturally falsified despite technical success. If a composed handoff package was explicitly allowed, the sidecar remains live and a MAJOR Contract C revision is not yet justified.

[camerontjs-dot]

Boundary disposition after EDR-002 / promotion-authority reconciliation

The technical result in this PR remains valid: the richer sidecar is sufficient to transport the exact U1/U2 unresolved causal provenance and independent_sufficient_alternatives multiplicity while Contract C 1.0 bytes remain unchanged.

It is not selected as the CAL→downstream governing-interface repair under current authority.

Authoritative evidence

EDR-002 / issue #17 defined Contract C itself as:

the immutable, decision-agnostic representation of CAL-attributable epistemic/result state required by legitimate downstream consumers

and made retained contribution / provenance state a required Contract C semantic family, including stable contribution identity, channel/type, exact Contract-B evidence reference, and non-deciding/residual contributions. It separately required the terminal basis / causal-multiplicity distinctions demonstrated by RC2-D.

The Contract C 1.0 promotion PR #18 then compressed the candidate only after mapping each EDR-supported invariant to retained canonical v1 fields. Its frozen conformance manifest maps:

• retained_evidence_and_residual_provenance → propositions[].contributions[];
• terminal_basis_and_causal_multiplicity → propositions[].conclusion.{causal_form,basis_members,residual_contribution_ids,rule_roles,...}.

The repository-wide accepted Contract/Apparatus Separation Invariant additionally requires the boundary model upstream apparatus -> governing contract -> downstream apparatus, states that the contract owns cross-apparatus authority, and forbids producer-private behavior/fields absent from the governing contract from silently becoming downstream authority.

Disposition

RICHER_SIDECAR_TECHNICALLY_SUFFICIENT_BUT_BOUNDARY_FITNESS_FALSIFIED_UNDER_CURRENT_AUTHORITY

The sidecar therefore falsifies the claim that an in-band wire change is technically necessary for information transport, but it does not provide a smaller authorized boundary repair. Making the sidecar mandatory would require a new/superseding governance decision that redefines the CAL→downstream boundary as a composed Contract-C-plus-companion-contract interface. That is a broader authority change than the tested in-band non_deciding repair.

No merge, promotion, release, version assignment, or production change is authorized by this disposition.