dot-iu-cutter v0.2 — BR-6 Closure Options

document_path: knowledge/dev/laws/dieu44-trien-khai/v0.2-planning/dot-iu-cutter-v0.2-br-6-closure-options-2026-05-16.md
revision: r1
date: 2026-05-16
author: Agent (Claude Code CLI, Opus 4.7 1M)
phase: v0.2 — BR-6 absorption (PLANNING/OPTIONS ONLY)
mutation_performed: false
ddl_written: false
manifest_design_started: false
status: PENDING_GPT_REVIEW (decision owner: GPT + User)

Companion to the BR-6 TD analysis and the BR-6→P0-2 impact analysis (2026-05-16). This document presents closure options only. It selects a recommendation but does not close BR-6 — closure is GPT's decision.

---

Option A — Fully design split/merge now, before P0-2

Design the complete SPLIT/MERGE pipeline (edge redistribution, birth/lifecycle propagation, alias emission, hierarchy re-parenting, detection→handling wiring) and only then design P0-2 manifest tables against the finished pipeline.

• Pros: maximal forward-compat; P0-2 designed against a fully known executor; zero manifest rework risk.
• Cons: very large scope; pulls P1 execution work (and P0-6 review_decision, CUT/ VERIFY) into the critical path; violates the scope-backlog §4 step-freeze (split/merge execution is explicitly P1); long lead time before any P0-2 progress.
• Risks: scope creep into unauthorized CUT/VERIFY territory; GOV decisions get bundled into a mega-design that is hard to review; high chance of analysis stalling. Highest schedule + governance risk.

Option B — Add manifest hooks only, defer split/merge execution to P1

P0-2 manifest tables include the structural columns that describe a split/merge (operation_kind, block_role, source_span, soft refs, candidate_edges JSONB) but no execution logic. The split/merge executor is wholly P1.

• Pros: matches scope-backlog phasing (manifest = proposal layer, executor = P1); bounded scope; P0-2 can proceed; future-proofs the schema.
• Cons: the three GOV decisions (address-coining, authority inheritance, manifest↔ alias) are still needed before P0-2 DDL freeze — "hooks only" does not by itself answer them; risk that hooks are under- or over-specified without invariants stated.
• Risks: moderate — if invariants aren't explicitly recorded alongside the hooks, P1 could enact split/merge in a way the hooks didn't anticipate.

Option C — No hooks now, document only

P0-2 manifest tables are designed for first-cut only. BR-6 is documented as a known TD and revisited entirely at P1.

• Pros: smallest P0-2 scope; fastest to a P0-2 design.
• Cons: directly contradicts the closeout GPT review and scope-backlog §5 ("v0.2 design MUST read this TD before authoring P0-2 DDL; it is a prerequisite, not a follow-up"). Guarantees manifest_unit_block lacks span-partition identity / block_role → certain P0-2 re-cut when P1 lands.
• Risks: highest rework risk; re-opens a shipped governance table; violates Nguyên tắc 8 (a cut is not final — the schema must anticipate later topology change). Effectively rejected by the controlling reviews.

Option D — Hybrid: minimum hooks + recorded invariants (RECOMMENDED)

P0-2 carries the minimum viable structural hooks (Option B's columns) plus the BR-6 invariants written as binding constraints carried into P0-2 design, plus the three GOV decisions explicitly escalated for ratification before P0-2 DDL freeze (not before design start). Split/merge execution stays P1.

• Pros: satisfies the closeout review + scope-backlog §5/§6 ("absorb BR-6 before P0-2"); bounded scope (no P1/CUT/VERIFY pulled in); future-proofs the schema and the behavior (invariants bind P1, not just columns); GOV decisions are surfaced, not silently deferred; aligns with the Phase-α discipline (empty tables, soft refs, no CHECK/trigger, DROP-TABLE rollback).
• Cons: requires GPT/Council to rule on GOV-1/2/3 before P0-2 DDL can be frozen (design can still start) — one extra ratification gate.
• Risks: low — the only residual risk is a GOV decision changing a column's semantics, which is contained because design (not DDL) starts first and DDL freeze is explicitly gated on the GOV rulings.

---

Recommendation

recommended_option: D   # hybrid: minimum hooks + recorded invariants + escalated GOV decisions
rationale:
  - only option consistent with closeout GPT review + scope-backlog §5/§6
  - bounded: no P1 / CUT / VERIFY / review_decision pulled into the critical path
  - durable: invariants bind future P1 behavior, not just present columns
  - honors Nguyên tắc 8 (a cut is not final) without over-building
rejected:
  A: scope creep into P1/unauthorized execution; highest schedule+governance risk
  B: acceptable but incomplete — omits explicit invariants and GOV escalation
  C: contradicts controlling reviews; guarantees P0-2 rework; effectively prohibited
gpt_decision_required: true
agent_self_close_BR_6: false

---

Hard Boundaries

no_DDL_written: TRUE
no_manifest_design_started: TRUE
no_mutation: TRUE
no_migration: TRUE
no_production_touch: TRUE
BR_6_self_closed: FALSE
output_form: br_6_closure_options_only

End of BR-6 closure options.