Macro-4 Delete-Fast Master Requirements — R2-B2 (2026-06-19)

Date: 2026-06-19 · Workstream: R2-B2-MACRO-4-STAGING-WORKBENCH-IO-CONTRACT-TD-ENTRY-GATE-2026-06-19 (Deliverable 37 of 90) · Editorial revision: rev1 Class: delete-fast master requirements · READ-ONLY · NON-ENACTING · NON-AUTHORIZING · NO write performed.

Metadata convention. Editorial revision (rev1) only. Storage revision/content_length authoritative at read time.

---

0. Status and non-authorization

STATUS: PASS — engineering / design-only. The master delete-fast requirements every draft surface must satisfy. Engineering PASS ≠ authority PASS. Default: HOLD.

1. Purpose

Answer macro question 4 — what delete-fast guarantee is required? — as a single master spec.

2. Sources / evidence read

Workbench delete-fast requirements (10); draft-object delete contract (17); bad-input/delete-fast plan §8; Macro-3 delete-fast risk map. Main process, no reader-agents.

3. Accepted baseline (carried)

Every draft surface is disposable as one unit; born/checked/replaced/rolled-back separately; rollback is a per-block unit, not a global script.

4. Evidence / analysis — master requirements (DF)

#	Requirement
DF-1	Single unit of deletion (Deliverable 38)
DF-2	No production dependency (the unit holds no production rows)
DF-3	Before/after proof (Deliverable 39)
DF-4	Delete evidence + rollback evidence (Deliverable 39)
DF-5	No leftover references (no dangling pointer/edge/row)
DF-6	No canonical address, no certified row, no KG edge created by the draft
DF-7	Failure modes enumerated and fail-closed (Deliverable 40)

5. Contract / requirement / matrix / result

These requirements bind any future staging build; satisfying them is a precondition for the staging-build gate (Deliverable 57). The mechanism is FUTURE_TECHNICAL_DESIGN_REQUIRED; no SQL/DROP written here. Nothing is deleted or built in this macro.

6. Owner-gated future work

Building/executing delete-fast is Owner-gated; forbidden now.

7. What remains unresolved

HOLD-2 OPEN (no atomic birth-certify-promote txn); runtime delete-fast unproven.

8. Ready for GPT/Codex review

Yes — Codex should confirm the requirements are total, isolated, and proof-bearing.