P10A-1B — Điều 35 Segmentation Candidate v2 (Read-Only Patch)

Date: 2026-04-29 Captured (UTC): 2026-04-29T04:02:21Z (source snapshot reused from P10A-1) Predecessor: knowledge/dev/laws/dieu38-trien-khai/reports/p10a-d35-segmentation-candidate-2026-04-29.md Mode: ZERO MUTATION. No DB queries needed (reused P10A-1 schema/vocab snapshot). Source body re-fetched read-only via GET /api/documents/.../full=true to compute per-segment hashes.

---

0. PASS Outcome

#	Criterion	Result
1	Root body ≤ 200 chars (no full-doc payload)	✅ 78 bytes (12 words)
2	dieu35.preamble.parent_key = dieu35.root	✅
3	§4 SCHEMA split into ≥2 children, each < technical_spec soft_limit (1000 words)	✅ 7 children, max 389 words
4	Section_type diversity ≥ 4	✅ 12 distinct types
5	description present on every unit	✅
6	word_count + length_flag_expected per unit	✅
7	Report uploaded at v2 path	✅ (this file)

ALL 7 PASS.

---

1. Source Re-snapshot (unchanged from P10A-1)

Field	Value
Path	knowledge/dev/laws/dieu35-dot-governance-law.md
Revision	13
SHA256	4353ec6d453411a7c8e207658bbc4457d00f99747cba90551c8a4926894d2e5c
Bytes (UTF-8)	39,938

---

2. Distinct section_types used (12)

appendix, article, changelog, checklist, definition, governance_process, heading, instruction_block, paragraph, principle, process, technical_spec. All ∈ tac_section_type_vocab (active).

---

3. Unit Inventory (30 units)

sort	unit_key	parent_key	section_type	wc	flag	title
-1	dieu35.root	—	article	12	normal	ĐIỀU 35: LUẬT QUẢN TRỊ DOT — v5.2 FINAL (synthetic root)
0	dieu35.preamble	dieu35.root	heading	215	soft_limit	Preamble (title block + meta)
1	dieu35.s1	dieu35.root	principle	212	normal	§1. MỤC TIÊU
2	dieu35.s2	dieu35.root	paragraph	295	normal	§2. PHẠM VI
3	dieu35.s3	dieu35.root	definition	197	normal	§3. DOT 2 CẤP
4	dieu35.s4	dieu35.root	technical_spec	5	normal	§4. SCHEMA (parent)
5	dieu35.s4.1	dieu35.s4	technical_spec	223	normal	4.1 dot_tools — 11 fields
6	dieu35.s4.sub2	dieu35.s4	technical_spec	310	normal	★ v5.1 KHOẢN 1 — UNIQUE PARTIAL + NOT NULL + REFERENCE TABLES
7	dieu35.s4.sub3	dieu35.s4	technical_spec	389	normal	★ v5.1 KHOẢN 1B — ĐỊNH NGHĨA 2 HÀM PHỤ TRỢ
8	dieu35.s4.sub4	dieu35.s4	technical_spec	277	normal	★ 4.1.1 DOT Description Contract
9	dieu35.s4.2	dieu35.s4	technical_spec	40	normal	4.2 dot_domains
10	dieu35.s4.3	dieu35.s4	technical_spec	34	normal	4.3 dot_coverage_required
11	dieu35.s4.4	dieu35.s4	technical_spec	147	normal	4.4 dot_config
12	dieu35.s5	dieu35.root	process	462	normal	§5. QUY TRÌNH TẠO DOT MỚI — 8 BƯỚC
13	dieu35.s6	dieu35.root	governance_process	5	normal	§6. VÒNG ĐỜI DOT (parent)
14	dieu35.s6.1	dieu35.s6	paragraph	19	normal	6.1 Trạng thái
15	dieu35.s6.2	dieu35.s6	governance_process	176	normal	6.2 Flow Sửa Bug DOT (fix_repair_dot)
16	dieu35.s6.3	dieu35.s6	governance_process	96	normal	6.3 Điều kiện từng bước
17	dieu35.s6.4	dieu35.s6	governance_process	85	normal	6.4 Verify 3 tầng
18	dieu35.s6.5	dieu35.s6	governance_process	99	normal	6.5 ADMIN Fallback
19	dieu35.s6.6	dieu35.s6	paragraph	36	normal	6.6 Paired test bắt buộc (NT12)
20	dieu35.s6.7	dieu35.s6	paragraph	21	normal	6.7 Retrofit
21	dieu35.s7	dieu35.root	process	35	normal	§7. ĐO LƯỜNG
22	dieu35.s8	dieu35.root	governance_process	650	soft_limit	§8. DOT TỰ QUẢN TRỊ — 4 CẶP
23	dieu35.s9	dieu35.root	instruction_block	463	normal	§9. BOOTSTRAP — 4 BLOCKS
24	dieu35.s10	dieu35.root	checklist	142	normal	§10. SUCCESS METRICS
25	dieu35.s11	dieu35.root	governance_process	343	normal	§11. RETROFIT CLAUSE
26	dieu35.s12	dieu35.root	paragraph	30	normal	§12. (Đã bỏ) tombstone
27	dieu35.appendix_a	dieu35.root	appendix	209	normal	PHỤ LỤC A — 24 Domain Seed
28	dieu35.changelog	dieu35.root	changelog	316	normal	CHANGELOG
29	dieu35.post_merge_todo	dieu35.root	checklist	132	normal	GHI CHÚ BAN HÀNH (post-merge TODO)

Total: 30 units.

Length flag summary

• normal: 28 units
• soft_limit: 2 units
  • dieu35.preamble — 215 words (heading.soft=100). Acceptable: preamble is metadata-heavy; could re-type to paragraph if needed (paragraph.soft=300, would flip to normal).
  • dieu35.s8 — 650 words (governance_process.soft=500). Within hard_limit (1500). Recommend P10A-2 split into 4 cặp DOT children if drift continues, or accept soft_limit_exception.
• hard_limit: 0

---

4. Patch-by-patch summary (responses to GPT review)

#	Patch	Implementation
1	Root body minimal	dieu35.root.body = "Điều 35 — Luật Quản trị DOT v5.2 FINAL (synthetic structural root)." (78 bytes, 12 words). body_sha256 = 5f32d336…3982bd.
2	Preamble child of root	parent_key = "dieu35.root", section_type = "heading".
3	§4 split	1 parent (dieu35.s4, body=placeholder) + 7 children (s4.1, s4.sub2, s4.sub3, s4.sub4, s4.2, s4.3, s4.4) all technical_spec. Word counts: 5 / 223 / 310 / 389 / 277 / 40 / 34 / 147 — all < soft_limit 1000.
4	§6 split	Detected 7 ### 6.x subheadings → split. Parent dieu35.s6 (placeholder body) + 6.1..6.7 children. 6.1/6.6/6.7 demoted to paragraph due to short bodies; 6.2/6.3/6.4/6.5 kept governance_process.
5	section_type diversity	12 distinct types vs 6 in v1. Mapping table: §1=principle, §2=paragraph, §3=definition, §4=technical_spec, §5=process, §6=governance_process, §7=process, §8=governance_process, §9=instruction_block, §10=checklist, §11=governance_process, §12=paragraph. Deviations from GPT guidance: none kept; all suggestions adopted. (§7 kept process per dispatch note "quá ngắn để đổi").
6	description + word_count + length_flag_expected	Added to every unit. description derived from first_para heuristic (≤80 words), capped & cleaned of leading list/table/code-fence noise. length_flag_expected computed via vocab soft_limit_words / hard_limit_words per section_type.

Caveats on auto-generated description

• For sections whose body opens with a table or SQL code fence (e.g. §3, §4.x, §5, §9), the heuristic falls through to the first non-decorative line, which sometimes lands inside an SQL block (e.g. dieu35.s3.description starts with CREATE OR REPLACE FUNCTION fn_dot_enforce_paired()…). Functionally valid as a "first prose-or-fenced summary" placeholder, but P10A-2 should regenerate description per Đ43 description-governance template (e.g. [MỤC ĐÍCH]: …) before INSERT.
• dieu35.s3.section_type = definition has description_required=true, body_required=true — both satisfied (description non-empty, body 1215 bytes). ✅
• dieu35.preamble.section_type = heading has description_required=false, body_required=false — descriptions provided regardless (defensive). ✅

---

5. Vocab-validation table (section_type vs body_required / description_required)

section_type	body_req	desc_req	soft / hard	units using	all bodies present	all descriptions present
article	t	t	500 / 1500	1	✅	✅
heading	f	f	100 / 300	1	✅	✅ (defensive)
principle	t	t	300 / 800	1	✅	✅
paragraph	t	t	300 / 1000	5	✅	✅
definition	t	t	200 / 500	1	✅	✅
technical_spec	t	t	1000 / 3000	8	✅ (parent placeholder is short prose, not empty)	✅
process	t	t	500 / 1500	2	✅	✅
governance_process	t	t	500 / 1500	6	✅ (s6 parent is placeholder)	✅
instruction_block	t	t	1000 / 3000	1	✅	✅
checklist	t	t	300 / 1000	2	✅	✅
appendix	t	t	2000 / 5000	1	✅	✅
changelog	t	t	500 / 2000	1	✅	✅

No body_required / description_required violations.

---

6. Open Items for P10A-2 (carry-forward, unchanged scope)

• D1 root: kept as article (synthetic), 12-word body. Acceptable per article.body_required=true; body_sha256 hashed over the literal placeholder string. P10A-2 may keep or drop; if dropped, all §-units must be re-parented to publication-level.
• D2 §4 children: split adopted; some carry ★ markers (KHOẢN 1, 1B, 4.1.1) — keys dieu35.s4.sub2/sub3/sub4 chosen because the source uses ### ★ v5.1 KHOẢN ... headings without numeric prefixes. P10A-2 may rename to canonical dieu35.s4.kh1 / s4.kh1b / s4.1.1.
• D3 (heading-only units): dieu35.preamble is the only heading-typed unit. Body 1397 bytes carries the title-meta block. Accept.
• D4 description backfill: see §4 caveat — auto-gen heuristic is good enough for candidate but must be regenerated per Đ43 template at P10A-2 time.
• D5 §12 tombstone: now typed paragraph per GPT guidance (was process).
• §8 soft_limit: P10A-2 should split into 4 children (one per cặp DOT) or apply length_exception_reason.

---

7. Đ41 / VPS Code Hygiene

• Zero repo files modified. No git activity required.
• Temp artifacts on VPS (kept for audit; SHA256 recorded):
  • /tmp/seg2.py — segmenter v2 (sha256 90ebb12d76902c63579c799f1232244f4a1db109a30364dea7275927983b8d29)
  • /tmp/d35-segments-v2.json — full candidate JSON 26,111 bytes (sha256 fb32786c33f8fc459bbaddd6c539c7def4e21331ed0ef44993068eab5c1f7ee6)
  • /tmp/units-v2.json — units array, 24,679 bytes
• Local mirror: /tmp/seg2.py (source).

---

8. STOP

P10A-1B closes here. All 7 PASS criteria met. Hand off to GPT for v2 review and P10A-2 dispatch.

---

P10A-1B report v2 | S187 | 2026-04-29 | Read-only patch | 7/7 PASS