Governed Bridge — NYCN Handoff Map

Status: draft spec, handoff map. Maps NYCN's fake airlock rehearsal outputs onto ICN's governed-bridge conformance contract, so the fake fixtures at ../../tools/bridge-conformance/nycn-intake-handoff-v0/ (§3–§12) and ../../tools/bridge-conformance/nycn-relationship-handoff-v0/ (§14) can be read alongside their NYCN sources. It defines vocabulary/field mappings only — it implements no bridge, adds no route, reads no source, and does not imply any bridge can import real rows today. Both sides are fake. It advances ICN #2377 without closing it. Package-local source nouns appear ONLY in NYCN-source columns/labels; every ICN-side name is generic (ICN core does not know what the package's local roles mean).

1. Purpose

ICN #2375 landed the governed-bridge conformance contract and its validator (tools/validate-governed-bridge-conformance.py), with a first fake fixture review-coverage-v0. NYCN #85 / #87 / #88 / #89 landed fake airlock fixtures and a local validator on the producing side. Neither side referenced the other. This document is the explicit bridge between the two fake vocabularies, and it backs the fixture nycn-intake-handoff-v0, which expresses NYCN's fake intake rehearsal in the ICN contract and passes the ICN validator unchanged.

The NYCN source is docs/bridge-rehearsals/fake-intake-import-airlock/ (source-records, dry-run-plan, review-decision) in the NYCN repo. Every value on both sides is invented.

2. Non-claims

• Fake data only; nothing was imported, handed off, synced, or written to a node.
• No runtime bridge, no connector, no live sync, no real Drive / Sheets / SimpleTix rows, no private operational data.
• No production, pilot-readiness, or live-federation claim; no claim that NYCN operations are ICN-native today.
• No payment / settlement / wallet / token / cryptocurrency framing. External references are observed, never processed.
• Action cards are derived read views, never write targets.
• ArtifactReceipt (verified-transfer proof) never satisfies ArtifactRegistrationReceipt.
• A receipt records an institutional fact and grants zero authority.

3. Record and field identity

4. Privacy-class handling

ICN's validator treats privacy_class as an opaque string, so NYCN's richer taxonomy passes through verbatim — public, participant_visible, care_sensitive, organizer_only, follow_up_only, external_reference, discard. ICN does not force NYCN to collapse these into a shorter set; the class travels with the field into the custody map. This is the anti-capture property: the contract fixes the safety invariants (custody kinds, receipt obligations, coverage, privacy scan) but leaves the institution's data model to the institution.

5. Custody-target translation

NYCN's free-text future_icn_target strings map onto ICN's closed 7-kind custody_target.kind enum:

The intake flow naturally exercises five custody kinds (artifact_registry, scoped_vault, governed_object, external_reference, discard). It does not naturally exercise policy_gate or policy_block; see §7. No fields were synthesized to reach full taxonomy coverage — review-coverage-v0 already proves that.

6. Gate-field handling

NYCN's follow_up_consent is a gate: consent present authorizes a follow-up; consent absent forbids it. NYCN also treats any field name containing permission as gate-like (e.g. the sponsor flow's public_logo_permission, out of scope here). In this intake fixture the gate is expressed on the follow-up's own field_path: consent present → approve the underlying governed_object; consent absent → an automatic block (see §9).

7. Custody kinds not exercised

policy_gate and policy_block are ICN custody kinds where a field is routed to a gate/block target rather than a write target (e.g. an unapproved publication permission that is blocked from publication, as in review-coverage-v0's sponsor public_logo_permission). The intake flow has no such published-artifact gate; its one gate (follow-up consent) blocks the creation of a follow-up governed_object rather than the publication of an artifact, so it is modeled as a steward block verb on the governed object, not as a policy_block custody kind. A sponsor publication-permission gate — a natural policy_block — is a candidate follow-up slice (§13).

8. Reviewer authority gap

ICN requires a verifiable reviewer_authority_ref (a DID/signature/authority basis); a display role alone is explicitly insufficient. NYCN's fake review decision carries only a role-only reviewer_role (summit_data_steward) and has no verifiable-authority field. The fixture supplies a fake reviewer_authority_ref (authref_nycn_fake_did_sig_001) with the role as a display label. NYCN adding a verifiable authority reference to its airlock review-decision shape is a candidate follow-up (§13).

9. Decision decomposition

ICN requires exactly one atomic decision per (source_record_ref, field_path), using one of the closed verbs (approve, reject, hold, block, discard, request_reobservation, request_reclassification, request_member_consent_review). NYCN's review decisions are compound and record-level (e.g. approve-with-split, approve-public-only-hold-rest) with field-level lists (approved_targets / rejected_fields / held_for_review / blocked_fields / gate_fields). The handoff decomposes each compound record-level NYCN decision into N atomic ICN per-field decisions. For example, a single NYCN "approve-with-split, discard the free text, block the no-consent follow-up" record becomes distinct approve, discard, and block rows keyed by (record, field).

An automatic block (no-consent follow-up) is marked automatic: true and carries no BridgeReviewDecisionReceipt — an automatic policy block is not a human review.

10. Receipt translation and the action-card conflict

NYCN's fake planning materials name a follow-up as an action-card creation and use a receipt named ActionCardCreationReceipt. ICN forbids that name: it is a deprecated placeholder (see governed-bridge-receipts.md), and the conformance validator rejects it as a literal anywhere inside a fixture directory. The doctrinal reason is ADR-0027: an action card is a derived read view of an underlying governed object; it is never a write target, so there is no card-write receipt. The handoff routes a consented follow-up to an underlying governed_object with FollowUpObjectCreationReceipt; a card derives from that object downstream. This deprecated name is named here (outside the fixture directory) only to document the conflict; it never appears in the fixture.

Otherwise the receipt vocabularies align: BridgeDryRunReceipt, BridgeReviewDecisionReceipt, BridgeImportReceipt, VaultObjectWriteReceipt, ArtifactRegistrationReceipt, FollowUpObjectCreationReceipt, ExternalReferenceObservationReceipt, DiscardDecisionReceipt, and a consent policy-block receipt are shared by both sides.

11. External references are observe-only

NYCN's registration_reference (an external ticketing/registration id) maps to ICN's external_reference custody kind with ExternalReferenceObservationReceipt. ICN records that an external system holds authority over a fact — it does not import the external document, does not process any settlement, and does not claim to be the source of truth. The binding marks external_reference_policy.observe_only: true.

12. Plan-hash gap

ICN binds the review to the dry-run via reviewed_plan_hash == plan_hash (a hash-bound review). NYCN's airlock has no plan-hash concept. The fixture supplies a fake planhash_nycn_fake_0001 on both sides. NYCN adding a plan-hash / reviewed-plan-hash binding to its airlock shape is a candidate follow-up (§13).

13. Candidate follow-ups (not opened by this document)

• A relationship-flow handoff fixture (exercises policy_block publication gates) — landed as nycn-relationship-handoff-v0 (§14).
• A receipt-naming decision for non-follow-up governed objects — resolved: the generic GovernedObjectCreationReceipt landed in the receipts spec and validator family floor; the binding pins the per-field receipt.
• NYCN-side plan-hash / reviewed-plan-hash support in the airlock review shape.
• NYCN-side verifiable reviewer authority reference (beyond the role-only field).
• A stronger NYCN fake-fixture privacy scan (real-name and external-id shape detection, which the NYCN validator does not yet perform).
• Record-state-dependent custody — partially addressed (v0): a field's custody entry may now carry an optional condition predicate declaring the source-state precondition under which its route is eligible (§15). Per-state multi-kind routing — the same field path routing to a different custody kind by record state — remains a future change (one field path still has one custody kind).

These are recorded as direction only; this document opens no issues.

14. Relationship / recognition / obligation handoff (nycn-relationship-handoff-v0)

Maps NYCN's fake sponsor-pipeline airlock rehearsal (its package-local name; docs/bridge-rehearsals/fake-sponsor-pipeline-airlock/ in the NYCN repo) onto generic ICN custody. The word "sponsor" is NYCN source vocabulary and appears only in the source column below — the ICN fixture, field paths, custody kinds, and receipts are entirely generic.

Source authorities translate as: sponsor-provided → counterparty_provided, organizer-derived → institution_derived, external-accounting-system → external_accounting_system. Privacy classes translate as: sponsor_restricted → relationship_restricted, organizer_only → internal_only; the rest pass through. All are opaque strings to ICN core.

15. Record-state-dependent custody (v0: eligibility conditions)

§14's declined-prospect row exposed a real v0 limit: the custody meaning of a source field can depend on the source record's state. A declined record is an internal closure record, not a commitment; a public-recognition field must stay unproposed unless the record's state permits publication. The original fixture handled this by omission — the declined record simply does not propose public_listing_name — which is safe but invisible to the conformance validator, and the tempting-but-forbidden shortcut is to invent a renamed field to bypass source state.

v0 mechanism — an optional condition predicate on a field_custody_map entry. A field's custody entry may declare the source-state precondition under which its route is eligible:

relationship.public_listing_name:
  custody_target: { kind: artifact_registry, namespace: public-recognition }
  required_receipts: [BridgeImportReceipt, ArtifactRegistrationReceipt]
  condition:
    source_field_path: relationship.record_state   # opaque source field
    op: equals                                      # closed ICN operator set
    value: active                                   # opaque literal

• Generic and opaque. source_field_path and value are opaque strings/scalars ICN core never interprets — exactly like field_path, privacy_class, and object_class. Only op is ICN vocabulary, drawn from a closed set: equals, not_equals, in, not_in, present, absent. The predicate's keys are strict (source_field_path / op / value only) so a package noun (e.g. a sponsor_status key) cannot ride in — the Meaning Firewall holds.
• Explicit, not silent. The condition records why a field is unproposed for a given record (its state does not satisfy the predicate) rather than leaving it to omission. It is producer-declared intent the validator can shape-check.
• v0 scope — shape only. The validator validates the predicate's shape; it does not evaluate it or route custody by state. A conditioned field must still be exercised by at least one dry-run action where its condition holds (the existing binding-coverage rule is unchanged), and a field path still has exactly one custody kind.
• Deferred (future change). Per-state multi-kind routing — the same field routing to a different custody kind under different record states — would require the binding to hold per-condition custody variants and the dry-run kind-match check to become condition-aware. That is a larger change, recorded here as direction only; this v0 slice does not open it.

Non-claims. Fake-fixture / conformance design only: no runtime bridge, no connector, no live source scan, no real data, no package-specific ICN custody kind or receipt class, and no claim that NYCN operations are ICN-native today.