# PC Form & PCN Workflow Audit, Data Integrity Review, and Unified Flow Plan

**Date:** 22 August 2026  
**Scope:** Codebase at `d:\Projects\performance_management_system` (Laravel 10 / PHP 8.2, Blade UI, MySQL/`pc` connection)  
**Constraint:** No production code was changed. This document is analysis only.

Evidence labels used throughout:

- **Confirmed** — verified in schema, models, services, routes, or tests
- **Strongly inferred** — consistent across multiple call sites but not documented as policy
- **Uncertain** — insufficient evidence in the repository

---

## 1. Executive Summary

The application does **not** have a single PC Form / PCN lifecycle. It has **parallel products** that share naming (`state`, `status`, `is_revised`, `is_modified`, draft tables, meetings) but are **sibling entities under a scheme (`projects.project_id`)**, not parent/child.

Critical facts (**Confirmed**):

1. **There are no incrementing `revision_number` or `modification_number` columns.** Revision and modification are **boolean flags** plus **new database rows** (`parent_id` chain). Display “revision position” is computed in listing code, not stored.
2. **Drafts are implemented at least three ways** for PC Forms (in-place `pc_forms.state = draft`, standalone `pc_forms_drafts` with `pc_form_id` null, and linked `pc_forms_drafts` rows that **accumulate** on save).
3. **Two approval machines coexist:** actor-based **forward/decline** (`workflow_stage`, `current_actor_id`, `pc_form_forwards` / `pcn_forwards`) for PC-I/PC-II and PCN, versus **legacy counters** (`check_count` / `approve_count`) still used for later PC kinds and leftover rows.
4. **PCN is not generated from a PC Form.** Meetings attach **either** a PC Form **or** a PCN via `mws_meeting_subject_links`.
5. **Import is PCFMS-only** (`pc:migrate-pcfms` → `PcfmsPcMigrationService`). It writes directly to `pc_forms`, maps source IDs in `pcfms_import_maps`, and **does not** use `PcFormService::create()`. Imported `parent_id` is often set to **the row’s own id**. PCN import was **not found**.

The recommended target is: **one scheme-scoped document lifecycle, one draft store, one versioning model (immutable submitted rows + flags only as derived labels), one workflow engine, import that exits into that model.**

---

## 2. Scope and Methodology

Inspected:

- Repository layout: Laravel `app/`, Blade `resources/views/`, `routes/web.php`, `database/migrations/`, `tests/`
- Models: `App\Models\PC\PcForm`, `PcFormDraft`, `ProjectConceptNote`, `ProjectConceptNoteDraft`, `PcfmsImportMap`, `MwsMeetingSubjectLink`
- Services: `PcFormService` (~14k lines), `ProjectConceptNoteService`, `PcfmsPcMigrationService`, workflow/stage/recipient/approval-count services
- Controllers: `PcFormsController`, `ProjectConceptNoteController`
- Seeders: `PcFormControlAccessSeeder`, `ProjectConceptNoteControlAccessSeeder`
- Tests under `tests/Feature/PcForm/`, `tests/Unit/*PcForm*`, `tests/Unit/*Pcn*`
- Global searches for revision/modification/draft/import/meeting/permission identifiers

Frontend is **server-rendered Blade + jQuery/Select2**, not a separate SPA. There is no dedicated React/Vue PC Form app.

---

## 3. Repository Findings

| Layer | Finding | Confidence |
| ----- | ------- | ---------- |
| Backend | Laravel 10, `PcFormsController` + `PcFormService` | Confirmed |
| Frontend | `resources/views/pc-forms/`, `resources/views/project-concept-notes/` | Confirmed |
| Database | `pc_forms` + many section tables; mirrored `pc_forms_drafts` + `pc_form_draft_*`; `project_concept_notes` + section tables (split later) | Confirmed |
| PC kinds | `pc_i`, `pc_ii`, `pc_iii_a`, `pc_iii_b`, `pc_iv`, `pc_v` on `pc_forms.pc_form_kind` | Confirmed |
| PCN | Independent module; same scheme; similar flags/workflow | Confirmed |
| Meetings | `mws_meetings` + `mws_meeting_subject_links` (`pc_form_id` XOR `project_concept_note_id`) | Confirmed |
| Import | Artisan `pc:migrate-pcfms`; map table `pcfms_import_maps` | Confirmed |
| Permissions | `control_access` + `can('pc-forms', action)` / `can('project-concept-notes', action)` | Confirmed |

`PcFormService.php` is the gravitational center of PC Form behavior. PCN is smaller and cleaner but **copied the same flag semantics**.

---

## 4. Current State — Core Concepts

### 4.1 What each word actually means

| Concept | What it is in this codebase | Separate table? | Status? | Number? | Confidence |
| ------- | --------------------------- | --------------- | ------- | ------- | ---------- |
| **PC Form** | One row in `pc_forms` plus section child tables | Yes (`pc_forms`) | `state` + `status` (active/inactive) | UID `pc_uid` | Confirmed |
| **PCN** | One row in `project_concept_notes` plus section tables | Yes | Same pattern | `pcn_uid` | Confirmed |
| **Draft (PC new)** | `pc_forms` row with `state = draft` | Same table | Yes | No | Confirmed (`2026_08_20_214202_add_draft_to_pc_forms_state_enum.php`, `PcFormService::create($isDraft=true)`) |
| **Draft (standalone)** | `pc_forms_drafts` with `pc_form_id` null | Yes | `state=draft` | No | Confirmed |
| **Draft (linked)** | `pc_forms_drafts` with `pc_form_id` set; **new row on every “save as draft”** when in-place save is not used | Yes | Yes | No | Confirmed (comment in `PcFormService::update`) |
| **PCN draft** | `project_concept_note_drafts` standalone or linked | Yes | Yes | No | Confirmed |
| **Revision** | New `pc_forms` / `project_concept_notes` row with `parent_id` = root id, often `is_revised=true` | New row | Plus workspace flags | **No stored number** | Confirmed |
| **Modification** | Boolean `is_modified` (and PCN `modified_pcn_source_id` workspace) | Flag / extra row | Not a status | **No stored number** | Confirmed |
| **Version** | PCFMS `VersionNo` used only to **sort import groups**; not persisted on `pc_forms` | No | No | Import-only | Confirmed |
| **Submission** | `status=1`, `submitted_at` set, workflow started | Same row or new row | `state` / `workflow_stage` | No | Confirmed |
| **Approval / rejection** | Reviews (`pc_form_reviews` / `pcn_reviews`) **and/or** forward actions; `state` synced from stage | Review tables | Yes | Counts on legacy path | Confirmed |
| **Meeting** | MWS meeting + subject link | Yes | Meeting lifecycle separate | N/A | Confirmed |
| **Imported form** | Normal `pc_forms` row + `pcfms_import_maps` | Map table | Mapped from PCFMS status | Source VersionNo discarded | Confirmed |
| **System-generated** | User-created via wizard; also PC-II–V dedicated create routes | Same | Same | Same | Confirmed |
| **Historical record** | Prior chain rows with `status=0`; PCFMS files via `PcfmsPcformsHistoricalFilesService` | Same / files | Inactive | Computed position | Confirmed |
| **Snapshot** | Not a first-class snapshot table; copies of section rows | Copied children | N/A | N/A | Confirmed |
| **Audit** | `LoggingService` / activity events; reviews; forwards | Activity + review + forward | N/A | N/A | Confirmed |

### 4.2 Current vs original vs “current version”

- **Original / root:** `parent_id` **null** in live create path (`PcForm::isRoot()`). **Import often sets `parent_id = id`.** (**Confirmed** — `PcfmsPcMigrationService::importPc1Groups` / `importPc2FromPc1`)
- **Current submitted:** `status = 1` (tinyint/boolean “active”). Older chain members set `status = 0` on revision spawn / import of non-latest.
- **Current workspace:** unpublished shells: `post_approval_revision_workspace=1` and `status=0` (PC); PCN also `modified_pcn_source_id` + `status=0`.
- **Multiple rows can represent one logical scheme document.** Listing code specially includes workspaces and filters “current” (**Confirmed** — `PcFormService` listing helpers).

---

## 5. Current PC Form Workflow

### 5.1 Entry points (**Confirmed**)

| Action | HTTP | Controller | Service |
| ------ | ---- | ---------- | ------- |
| Index / list | `GET /pc-forms` | `PcFormsController@index` | `PcFormService` listing + standalone drafts |
| Create PC-I | `GET/POST /pc-forms` | `create` / `store` | `PcFormService::create` |
| Create PC-II–V | `GET pc-forms/create-pc-ii` … `create-pc-v` | dedicated create | same `create` with kind |
| Edit / update / submit | `GET/PUT /pc-forms/{id}` | `edit` / `update` | `PcFormService::update` |
| Autosave draft | `POST pc-forms/draft-autosave` (+ `/{id}`) | `autosaveDraft*` | service draft persist |
| Delete form | `DELETE /pc-forms/{id}` | `destroy` | `PcFormService::delete` |
| Delete standalone draft | `DELETE pc-forms/drafts/{draft}` | `destroyDraft` | `deleteStandaloneDraft` |
| Start revised PC-I | `GET pc-forms/begin-approved-revision` | `beginApprovedRevision` | `spawnPostApprovalRevisionWorkspace` |
| Revision gate | `GET pc-forms/revision-availability` | `revisionAvailability` | service |
| Forward / decline | `POST pc-forms/{id}/forward\|decline` | | `PcFormWorkflowService` |
| Approve / reject (legacy + meeting) | `POST .../approve\|reject` | | `PcFormService::approve` / reject |
| Chairman advise/defer | `POST .../chairman-*` | | `storeChairman*` |
| Submit modified to CDS | `POST .../submit-modified-to-cds` | `submitModifiedToCds` | service |
| Import | Artisan `pc:migrate-pcfms` | `MigratePcfmsPcCommand` | `PcfmsPcMigrationService::migrate` |

### 5.2 Create / draft / submit (happy path)

```
Wizard POST (isDraft)
  → PcFormService::create(true)
  → INSERT pc_forms state=draft, persist sections on MAIN tables via persistDraftPcFormWorkspace

Wizard POST (submit)
  → PcFormService::create(false)
  → INSERT pc_forms, section tables, submitted_at, workflow defaults
  → completeForwardWorkflowAfterSubmit() if PC-I/II
```

**Confirmed:** first-time draft now lives on **`pc_forms`**, not `pc_forms_drafts`.

### 5.3 Saving / editing drafts

| Situation | Mechanism | Confidence |
| --------- | --------- | ---------- |
| `state=draft` or revision workspace or rejected/modified editable | **In-place** update of `pc_forms` + section tables (`pcFormAllowsInPlaceDraftWorkspaceSave`) | Confirmed |
| Existing submitted form, “save as draft”, not in-place eligible | **Always INSERT new** `pc_forms_drafts` row (explicit comment) | Confirmed |
| Never submitted, listing “standalone” | `pc_forms_drafts.pc_form_id` null; `createMainFormFromStandaloneDraft` | Confirmed |
| Autosave | Same persist path as draft workspace | Confirmed |

Saving a draft **does not increment a revision counter** (none exists). It **does** write activity logs on in-place save.

### 5.4 Submit / resubmit

`PcFormService::update(..., $isDraft=false)`:

1. If `shouldCreateRevisionOnSubmit()` → **new row** via `createPcFormRevisionFromSubmit`, source typically inactivated.
2. Else if latest linked draft `state=draft` → `copyDraftToMain`, draft → `requested`.
3. Else patch the same `pc_forms` row from the request.

`shouldCreateRevisionOnSubmit` (**Confirmed**):

- Post-approval revision workspace → always new row
- Else only if `state=rejected` **and** latest rejection in cycle is **approver-type**, **and** not same-row defer (checker/department/PnD/chairman defer)

### 5.5 Approve / reject / forward

**Forward path (PC-I / PC-II when `usesForwardWorkflow()`):** actor-based. Stages in `PcFormWorkflowStageResolver`: `ad_review`, `ad_approve`, `pnd_development_section`, `pnd_internal_circulation`, `ddwp_department_handoff`, `meeting_linked`, `approved`, `declined`. `syncLegacyState()` writes old `state` enum.

`usesForwardWorkflow()` (**Confirmed**, `PcForm.php`): PC-I/II only if `status` truthy and `state` not draft/rejected/approved. Inactive workspaces therefore **do not** use forward until submitted.

**Counter path:** `PcFormApprovalCountService` increments `check_count` / `approve_count` and advances `state` (`requested` → `checking` → `department_approval_pending` → `under approval` → `approved`). Still used for PC-III+ unless `uses_forward_workflow` column is set.

Reviews live in `pc_form_reviews` with expanding action enums (defer, chairman, AA, concurrence, etc.).

### 5.6 Meeting association

- Link table: `mws_meeting_subject_links.pc_form_id`
- `PcForm::hasLinkedMeeting()`
- Stage can become `meeting_linked`
- Defer/hold flags: `defer_repeat_*` / `defer_block_repeat_*` on `pc_forms` change **whether a new working paper** may be generated (`allowsNewWorkingPaperGeneration`) — **meeting type is a workflow discriminator** (**Confirmed**)

### 5.7 Workflow map vs requested list

| # | Flow | Status |
| - | ---- | ------ |
| 1 | New PC Form | Implemented — `create` |
| 2–4 | Draft create/save/edit | Implemented — **multiple mechanisms** |
| 5 | Submit | Implemented — `create`/`update` |
| 6 | Revise submitted | Implemented — new row / workspace |
| 7 | Modify | Implemented as **flag + same-row or new-row resubmit**, not a numbered series |
| 8 | Approve/reject | Implemented — dual machines |
| 9–11 | PCN create/modify/revise | Implemented in PCN module (see §6) |
| 12–13 | Meeting create/associate / type behavior | Implemented in MWS + flags |
| 14 | Generate from another source | **PC-II–V from scheme/PC-I context**; not “generate PCN from PC Form” |
| 15–20 | Import lifecycle | **Ingest only** via PCFMS; after that they are `pc_forms` with map rows. No separate imported-edit API |
| 21 | System-generated | Same create path |
| 22 | Clone/copy/duplicate | **Not a user “duplicate” action.** Internal `copyMainFormToRevision` / `copyDraftToMain` / rejected clone |
| 23 | Reopen submitted | Via reject/defer edit, revision workspace, or modified flags |
| 24 | Cancelled/withdrawn | **Not implemented / not found** as first-class states |
| 25 | Archived | Working papers mention “not archived”; PC Form **archive state not found** |
| 26 | Historical | Inactive chain rows + PCFMS file browser |
| 27 | Other | AI analyze; cover/AA/advice letters; post-forum; actor reconcile commands; workflow count sync commands; **EAED mobile API** for PC-I location drafts (`/api/v1/eaed`, `EaedDraftService`) |

---

## 6. Current PCN Workflow

### 6.1 Entry points (**Confirmed**, `routes/web.php`)

Resource `project-concept-notes` plus: approve, reject, forward, decline, timeline, workflow-actions, revision-availability, **begin-approved-revision**, **begin-modified-revision**, preview, scheme-eligibility.

### 6.2 Create / draft / submit

`ProjectConceptNoteService::create`:

- Draft → **standalone** `project_concept_note_drafts` (`project_concept_note_id` null)
- Submit → `project_concept_notes` + section persist + `finalizeSubmitAndForward`

Linked draft: `createLinkedDraftFromMain` — **new draft row each time** (same accumulation pattern).

**Difference vs PC Form:** first PCN draft is **not** an in-place `project_concept_notes` row (PC Form first draft **is** in-place on `pc_forms`). **Confirmed.**

### 6.3 Revision vs modification

| Action | Method | Persistence |
| ------ | ------ | ----------- |
| Revised PCN (after approved) | `beginApprovedRevision` | New row, `status=0`, `post_approval_revision_workspace=1`, `is_revised=1`, copy sections |
| Modified PCN | `beginModifiedRevision` → `beginModifiedPcnRevision` | New row, `status=0`, `modified_pcn_source_id`, `is_modified=1` |
| Submit workspace | `shouldCreateRevisionOnSubmit` includes modified workspace | New **active** row |

Eligibility: `revisionAvailability()` — no in-progress PCN + approved exists; modified baseline separate.  
**Modified PCN** additionally requires an **active PCN plus a completed linked meeting** (`pcnQualifiesForModifiedRevision` / `pcnHasCompletedMeeting`) — **Confirmed**.

### 6.4 Workflow stages

`PcnWorkflowStageResolver`: similar AD/PnD/meeting path; **terminal `cleared_cdwp`** instead of generic approved-only. Deprecated stages (`working_paper`, `pre_pdwp`, …) remain as constants (**Confirmed**).

### 6.5 PCN import

**Not implemented / not found** in inspected code (no PCN analog of `PcfmsPcMigrationService`).

---

## 7. Draft Lifecycle (deep)

### Current State

**PC Form**

- Authoritative **wizard drafts:** `pc_forms` (`state=draft`).
- **Legacy / linked / standalone:** `pc_forms_drafts` + full parallel section schema (migration `2026_02_10_174449`).
- Drafts are **mutable**.
- **Multiple linked drafts per form are allowed and intended** (“do not update existing draft”).
- Submit copies **latest** `state=draft` via `latest('id')` — older linked drafts become stale leftovers.
- Standalone delete: `destroyDraft` / `deleteStandaloneDraft`. Linked drafts: cascade if `pc_form_id` FK `nullOnDelete` when parent deleted.
- Abandoned drafts: **no expiry job found**.
- Imported drafts: import writes **main** `pc_forms`, not draft tables (**Confirmed**).
- **EAED (mobile):** `routes/api.php` `/api/v1/eaed/drafts*` → `EaedDraftService` creates/updates **standalone** `pc_forms_drafts` (location/photos). Fourth write path into the draft table (**Confirmed**). Web promote remains `createMainFormFromStandaloneDraft`.

**PCN**

- Parallel draft tables; first save is standalone draft.
- Modified/revision workspaces are **main table shells**, not draft table.

### Problems

- Two physical draft models on PC Forms; listing merges them (`is_standalone_draft` stub).
- Latest-id submit can **ignore** a newer in-place main-row edit if a stale linked draft exists (**Strongly inferred** from `latest('id')` precedence in `update`).
- No optimistic lock on drafts.
- Parallel section tables double schema cost and drift (PC-III/IV/V columns added on both sides over many migrations).

### Is the draft model logically sound?

**No.** It is historically layered: shadow tables → in-place workspace → leftover “always insert new draft”. Sound rule would be: **one mutable working copy per (user, logical document) until submit.**

---

## 8. Revision Lifecycle (deep)

### What causes a revision? (**Confirmed**)

| Trigger | PC Form | PCN |
| ------- | ------- | --- |
| First submit on scheme that already had approved PC-I | `is_revised` on **new** create (`shouldMarkNewPcAsRevisedSchemeFirstSubmit`) | N/A (blocked if in progress) |
| Post-approval “Revised” | Workspace row then submit clones active row, `is_revised` | Same |
| Approver-stage reject (not defer) | New row; `is_revised` unless defer-repeat flags force “modified” semantics | Same pattern |
| Import later VersionNo / later PC1 in group | `is_revised=true`, `parent_id=root` | Not imported |
| Display | `revision_chain_position` computed by `id` order | No equivalent found |

**Does not increment on:** draft save, meeting change, metadata-only edit, import VersionNo persistence (VersionNo discarded).

### Storage

- `parent_id` → root id (live) or **self** (import)
- `is_revised` boolean
- `post_approval_revision_workspace`, `revision_source_approved_id` (PC Form)
- Full **row copy** of section tables (`copyMainFormToRevision`)
- Source row `status=0` when superseded (**import** and revision submit)

**Scope:** per scheme + kind (PC-I chain). Not global. Not a numeric sequence stored.

### Problems

- `is_revised` is overloaded (post-approval Revised PC-I **and** some resubmit clones **and** import).
- Computed position can disagree with business “Revision 2” if ids are not chronological or import `parent_id=self` pollutes grouping (`attachRevisionChainPositions` treats `parent_id ?? id` as root).
- No unique constraint on `(parent_id, …)` “one active revision”.

---

## 9. Modification Lifecycle (deep)

### What qualifies? (**Confirmed**)

- **Not** a field-level change detector. No “which columns count” engine.
- `is_modified` set on:
  - Successful resubmit after certain rejections (`resolveIsModifiedOnSuccessfulResubmit`) — elevated, **never cleared**
  - New revision row from approver reject (`is_modified = !fromPostApprovalWorkspace`)
  - PCN modified workspace from day one
- **Same-row** resubmit after checker/department/PnD/chairman **defer** — routing uses `shouldSameRowResubmitAfterDefer`; `is_modified` may stay false for checker/dept defer
- Meeting hold flags **do not** set `is_modified` but **change WP generation** when `is_modified` is already true

### Numbering

**There is no modification counter.** Duplication/skip/reset of numbers **cannot occur as numbers**; inconsistency is **boolean vs row history** (modified without a new row, or new row without clear user-facing “Mod N”).

Draft save does **not** set `is_modified`. Import does not set a modification series (only `is_revised` for later PC1s).

---

## 10. Version / History Storage

| Mechanism | Role | Authoritative? |
| --------- | ---- | -------------- |
| Extra `pc_forms` / PCN rows | Prior submitted editions | Yes for content |
| `pc_form_reviews` / `pcn_reviews` | Decisions | Yes for actions |
| `pc_form_forwards` / `pcn_forwards` | Actor trail | Yes for forward workflow |
| Activity / `LoggingService` | Audit text | Partial |
| `pcfms_import_maps` | Source identity | Yes for import identity |
| Files `pcfms_pcforms` | Legacy binaries | Yes for files, not structured data |
| JSON on forms | Content fields, not version snapshots | Content |

**No event-sourced revision log. No JSON snapshot of full form per version.**

---

## 11. Meeting Workflow

Meetings are a **separate product** (Meeting Workspace). For PC/PCN they:

- Attach subjects (`MwsMeetingSubjectLink`)
- Gate MoM / PC decisions (see workspace MoM rules; PC approve/defer can require MoM signature completeness — **Confirmed** in project rules and meeting tests)
- Drive `workflow_stage = meeting_linked`
- Store **per-meeting-type** repeat/block flags on the **form/PCN row**

**Meeting type as hidden workflow discriminator — Confirmed:** `allowsNewWorkingPaperGeneration` switches on `pre-pdwp` / `pdwp` / `ddwp` columns. Forum decision tests (`PdwpForumDecisionFlowTest`) encode type-specific PC actions.

**Recommendation:** keep meeting type as a **permission + policy dimension for MWS**, but **stop storing meeting-policy flags as ad-hoc booleans on the document** if a unified model is adopted (move to meeting/hold records).

---

## 12. Permission Model

### Current State (**Confirmed**)

- Stored: `control_access` (module/form/route/action), role grants, field rows `control_access_fields` for **“PC 1”** only.
- Checked: `can($routeSlug, $action)` (`app/helpers.php`).
- PC Form kinds **share** `pc-forms` actions. **Not** split by `pc_form_kind`.
- PCN: `project-concept-notes` list/view/add/edit + workflow-oriented actions; seeder **removes** a long list of legacy PCN and leaked PC-I actions (`ProjectConceptNoteControlAccessSeeder`).
- Meeting types: **separate** MWS control-access (MoM generate, etc.), not the same catalog as PC Form.
- Forward actor: **not** only `can_approve` — `current_actor_id` + recipient resolver (hardcoded role/stage rules in `PcFormWorkflowRecipientResolver` / `PcnWorkflowRecipientResolver`).
- Department scope: `PcFormApprovalCountService` overlap helpers reused by PCN.
- Frontend: Blade `can()` hides buttons; backend also checks — **not frontend-only**, but UI and service rules can diverge (actor vs `can_check`).

### Should permissions split by PC Form type and meeting type?

**Yes, but only two extra dimensions, not a matrix of every verb × every kind × every forum.**

Recommended smallest model:

```
Role
  → grants
      → document_class: pcn | pc_i | pc_ii | pc_iii | pc_iv | pc_v
      → actions: list, view, create, edit_own_draft, submit, delete_draft,
                 revise_approved, modify_resubmit, check, department_approve, pnd_act, approve
  → meeting_class: pre_pdwp | pdwp | ddwp | general
      → actions already owned by MWS (create meeting, WP, MoM, circulate)
```

**Do not** add separate permissions for draft-save vs autosave, or import (import is ops/CLI).

**Why split kinds:** PC-III–V are singleton-per-scheme and skip PC-I sections (`isPerSchemeSingletonKind`, `isShortPcWizardKind`). One “add PC 1” grant already unlocks all create-* routes if the controller only checks `can('pc-forms','add')` (**Confirmed** on `beginApprovedRevision`).

**Why not split every meeting action onto the PC Form grant:** meetings already have MWS permissions; duplicating them on PC Form created the PCN legacy action sprawl the seeder is deleting.

**Drawback of over-granularity:** this app already has field-level PC-I view/edit (~40 fields × 2). Combining that with kind × meeting × action will make roles unmaintainable.

---

## 13. Imported PC Form Workflow

### Pipeline today (**Confirmed**)

```
pcfms.sql
  → PcfmsSqlTableReader
  → PcfmsSchemeResolver (SC_UID / ADP → projects.project_id)
  → group PC1 by scheme, sort VersionNo then PC1ID
  → upsertPcFormFromPc1 / upsertPcForm
  → pcfms_import_maps (source_table, source_id, pc_form_kind) UNIQUE
  → section imports (authorities, objectives, …)
  → PC2/PC3/PC4/PC5 sibling forms, often state=approved
```

Re-import: mapped rows **updated**; workflow counts **preserved** (`PRESERVED_WORKFLOW_COLUMNS`, command comment).

`mapWorkflowState`: PCFMS `PC1StatusID` 4/5/7 → `rejected`; `1` → `requested`; **default `approved`**.

### Imported → normal flow issues

| Issue | Root cause | Consequence | Fix | Data cleanup |
| ----- | ---------- | ----------- | --- | ------------ |
| `parent_id = self` | Import sets root to own id | `isRoot()` false; chain math treats self as parent | Normalize `parent_id` null for roots | UPDATE where parent_id=id |
| Inactive historical rows `status=0` | Latest-only active | Correct-ish, but mixed with live revision rules | Keep; document | None |
| Default state `approved` | Loose status map | Forms skip forward workflow (`usesForwardWorkflow` false when approved) | Map unknown to `requested` or explicit imported_approved | Review unmapped status IDs |
| No `workflow_stage` / actor | Bypass `PcFormService` | Cannot forward until reconciled | `ReconcilePcFormForwardWorkflowCommand` / actor commands | Run reconcile |
| VersionNo dropped | Not mapped to a column | Cannot show PCFMS version | Optional `source_version` on map table | Backfill from SQL if kept |
| PC-III/IV/V forced approved + counts=1 | Import payload | Appear finished; may not match reality | Import as historical approved **explicitly** flagged | `origin=pcfms` |
| No PCN import | None written | PCN always “system” | Out of scope or separate job | N/A |
| Historical files | Separate storage service | OK if paths exist | Keep | N/A |
| Edit/submit after import | Same controllers | **Works if state/status allow**; approved locked (`isPostAaAuthLocked` / approved) | Treat imported approved as terminal unless user starts Revised | None |
| Duplicate PCN | N/A | Import does not create PCN | — | — |
| ID collision | New autoincrement ids; source in map | Safe | Keep map unique | — |

**End state goal (imported becomes normal):** **Safe for content**, if roots are normalized and imported approved rows are explicitly **historical-complete** (same as a live approved PC-I). **Not safe** to drop `pcfms_import_maps` (re-import idempotency).

---

## 14. PC Form → PCN Relationship

**Confirmed:**

- **Not** parent/child. No `pc_forms.pcn_id`. PCN has `project_id`, not `pc_form_id` on the note (meeting links have both columns).
- One scheme can have **many** PC Form rows (kinds + revision chain) and **many** PCN rows (chain + workspaces).
- One PCN does **not** reference a PC Form revision.
- Changing a PC Form after PCN exists **does not** update the PCN.
- PCN is a **live row** (plus copied sections on revise), not a snapshot of a PC Form.
- Imported PC Form **can** later have a user-created PCN on the same scheme (no code block found that forbids it).
- Meeting change does not rewrite PCN↔PC Form FKs (there are none).
- **Meeting decisions share one pair of routes** for both document types: `POST meeting-workspace/meetings/{id}/linked-pc-form/approve` and `.../defer` (`MeetingWorkspaceMeetingController` → `MwsPcFormMeetingDecisionService` / `ProjectConceptNoteService::approveFromMeetingWorkspace`). MoM signature gate applies per subject (`MwsMomSchemeSubject`: exclusive `pc_form_id` **or** `project_concept_note_id`). **Confirmed.**

**Referential risks:** two in-flight documents on one scheme; meeting link hydrates missing PCN by **latest non-terminal PCN for project_id** (`hydrateMissingProjectConceptNotes`) — can attach the **wrong** PCN (**Confirmed**).

---

## 15. Functionality-to-Workflow Matrix

| Functionality | Current flow | Tables | Draft | Revision | Modification | PCN | Meeting | Import | Problems |
| ------------- | ------------ | ------ | ----- | -------- | ------------ | --- | ------- | ------ | -------- |
| Create PC-I | `PcFormService::create` | `pc_forms`+sections | In-place or none | Flag if scheme already approved | No | Independent | Later | No | Dual draft leftover |
| Create PC-II–V | Same + kind routes | Same | Same | Singleton kind rules | No | Independent | Later | PCFMS siblings | Shared `add` permission |
| Edit draft | In-place or new draft row | Main or drafts | Multiple | No | No | Parallel | — | — | Stale draft overwrite |
| Submit | `update`/`create` + forward complete | Main | Convert latest linked | Maybe new row | Maybe flag | `finalizeSubmitAndForward` | — | — | Dual workflow |
| Approve/reject | Forward **or** counters + reviews | reviews, forwards | — | — | Defer vs reject | Same split | Forum actions | Bypass | Two state machines |
| Revised | Workspace + submit clone | New `pc_forms` | In-place on shell | `is_revised` | Cleared on shell | Same | WP rules | Import `is_revised` | Flag overload |
| Modified | Flag + same/new row | Same | In-place if rejected | Sometimes new row | Boolean | `modified_pcn_source_id` | Hold flags | No | No number |
| Import | CLI upsert | `pc_forms`, map | Unused | Chain by VersionNo | No | None | None | Authoritative path | Divergent parent_id/state |
| Generate WP | Working paper services | `working_papers` | — | New row allows WP | Hold if modified | PCN WP type | Type-specific | — | Hidden discriminator |
| Clone | Internal copy only | Sections | — | Copy | — | Copy sections | — | — | No product clone |
| Historical files | File service | disk | — | — | — | — | — | PCFMS | Detached from rows |
| Reopen | Reject/workspace | Same | Yes | Yes | Yes | Yes | Defer flags | Edit if not locked | Easy to stuck |
| Cancel/archive | Not found | — | — | — | — | — | — | — | Soft-delete only via `delete` |

---

## 16. Duplicate / Competing Flows

| Duplicate | Where | Why it exists | Who uses it | Keep? |
| --------- | ----- | ------------- | ----------- | ----- |
| In-place draft vs `pc_forms_drafts` | `PcFormService` | Schema first, UX later | Listing + update | **Merge to in-place + optional standalone only for never-submitted** |
| Linked “always new draft” vs in-place | `update($isDraft)` | Comment-era behavior | Older edit path | **Deprecate insert-always** |
| Forward vs counter workflow | I/II vs III+ and legacy | Migration mid-flight | `usesForwardWorkflow`, reconcile commands | **One engine**; counters derived |
| `state` enum vs `workflow_stage` | Model + resolver sync | Compatibility | All listings | **Stage authoritative**; state generated |
| PC `is_revised` vs PCN same flags | Copied design | Feature parity | Both UIs | Keep concept, **one shared policy class** |
| Approve POST vs forward POST | Controller | Old checkers + new actors | Both live | **Forward is canonical** for I/II/PCN; keep approve for meeting/forum |
| PCFMS upsert vs `PcFormService::create` | Command vs HTTP | Bulk legacy | Ops | Keep ingest; **normalize then stop special-casing** |
| Meeting hydrate PCN by project | `MwsMeetingSubjectLink` | Missing FK history | MWS | **Require explicit PCN id** |

---

## 17. State Machine

### PC Form `state` values (**Confirmed**)

`draft`, `requested`, `checking`, `department_approval_pending`, `under approval`, `approved`, `rejected`  
(`draft` added later; original enum lacked it.)

### Parallel `workflow_stage` (PC-I/II)

`ad_draft` (const) / `ad_review` → `ad_approve` → `pnd_development_section` → `pnd_internal_circulation` → (`ddwp_department_handoff`) → `meeting_linked` → `approved` | `declined`

`PcFormWorkflowStageResolver::syncLegacyState` maps these onto `state`.

### Who moves

- Submitter: draft → requested / first actor
- Current actor: forward/decline
- Meeting/forum: approve, defer, chairman actions (reviews)
- Import: jump to approved/rejected/requested **without transitions**

### Illegal / messy transitions

- Import → `approved` without forwards (**Confirmed**)
- `status=0` workspace with `state=requested` looks “in flight” to naive queries
- Rejected + `is_modified` + same-row resubmit resets counts but keeps flags
- Frontend may show Draft for workspaces while `state` is `requested` (`PcFormService` listing comments ~1580)

### Canonical target machine (recommended)

```
draft → submitted → in_review → (returned_to_creator) → submitted
                         → meeting
                         → approved
                         → declined
approved → (optional) revision_draft → submitted … (new edition row)
```

Store **one** `lifecycle_status`. Derive listing labels in one resolver.

---

## 18. Database / Table Audit

| Table | Purpose | Authoritative? | Used by | Duplicate? | Problems | Recommendation |
| ----- | ------- | -------------: | ------- | ---------- | -------- | -------------- |
| `pc_forms` | Document header + many kind columns | Yes | HTTP, import, MWS | Kind-specific JSON/text on same row | God table | Keep; stop adding policy flags |
| `pc_form_*` sections | Body | Yes | Service persist | Mirrored in drafts | Schema clone | Keep; generate persist |
| `pc_forms_drafts` + `pc_form_draft_*` | Shadow body | Partial | Listing, old save | **Full duplicate of sections** | Drift | **Deprecate** except standalone pre-create |
| `pc_form_reviews` | Decisions | Yes | Approve/forum | Overlaps forwards | Action enum growth | Keep |
| `pc_form_forwards` | Actor trail | Yes | Forward WF | — | — | Keep |
| `pc_form_department_approvals` | Dept sign-off | Yes | Counter/dept | — | — | Keep until engine merge |
| `project_concept_notes` | PCN header | Yes | PCN HTTP | Was wide; split | Flags copy PC | Keep |
| `pcn_*` section tables | Body | Yes | Persistence | Draft twins | — | Keep |
| `project_concept_note_drafts` + draft sections | Drafts | Partial | Create/update | Duplicate | Same as PC | Same deprecate path |
| `pcn_reviews` / `pcn_forwards` | WF | Yes | PCN | — | — | Keep |
| `pcn_department_approvals` | Created in `2026_07_27_120000` | **Uncertain / unused** | No Eloquent model found | Likely leftover of PC Form dept table | Dead table risk | **Confirm row counts; drop or wire** |
| `pcfms_import_maps` | Source map | Yes | Re-import | — | — | **Keep forever** |
| `mws_meeting_subject_links` | Subject | Yes | MWS | Hydration fallback | Nullable FKs | Require one subject FK |
| `working_papers` | WP | Yes | MWS/PC | `pc_form_id` / `project_concept_note_id` | — | Keep |
| `control_access*` | Perms | Yes | `can()` | Field-level explosion | PC 1 only fields | Keep actions; trim fields |
| Activity/logs | Audit | Partial | LoggingService | Not full snapshot | — | Keep |

---

## 19. Column Audit (important headers)

### `pc_forms` — keep

`id`, `pc_uid`, `project_id`, `pc_form_kind`, content columns, `created_by`/`updated_by`, `submitted_at`, timestamps.

### Dual-purpose / problem columns

| Column | Issue | Depends on | Replace with | Migration |
| ------ | ----- | ---------- | ------------ | --------- |
| `status` (0/1) | Means **active edition**, not workflow | Listing, import, `usesForwardWorkflow` | `is_current` or `superseded_at` | Backfill from status |
| `state` | Legacy + synced from stage | Everywhere | Generated from `workflow_stage` | Keep column temporarily |
| `workflow_stage` | Real machine for I/II | Forward | Canonical | Backfill III+ |
| `parent_id` | Null vs self | Root/chain | Null = root; CHECK | Fix import |
| `is_revised` | Overloaded boolean | UI, WP | Derived: `edition_kind=revision` | Optional new enum |
| `is_modified` | Sticky boolean | WP hold, UI | `edition_kind=modification` or review reason | Don’t clear history |
| `post_approval_revision_workspace` | Workspace vs edition | Listing, submit | `row_role=workspace` | |
| `revision_source_approved_id` | Good FK | Spawn | Keep | Index already |
| `check_count` / `approve_count` / `required_*` | Second machine | Legacy + import | Derived from reviews | Freeze writes |
| `defer_repeat_*` / `defer_block_*` | Meeting policy on document | WP generate | Meeting/hold table | Move later |
| `uses_forward_workflow` | Kind opt-in | III+ | Remove when all forward | |
| `workpaper` longText | Old notes | Some UI | WP table | Deprecate |

**Not present:** `revision_number`, `modification_number`, `version`, `import_id` on `pc_forms` (import id is map table).

### `project_concept_notes`

Same flag set plus `modified_pcn_source_id`, `workflow_stage`, `current_allowed_actions`, `cleared_for_cdwp_at`. Same recommendations.

---

## 20. Data Integrity Risks

### Referential

- `parent_id` FK exists (`nullOnDelete`) — **Confirmed**
- Import self-parent **violates the live root convention** (FK still valid)
- Draft `pc_form_id` nullOnDelete — orphans become standalone
- Meeting links can lack both FKs and hydrate by project — **wrong subject**
- Section tables generally `cascadeOnDelete` — good

### Version

- No unique “one current per (project, kind)” except business rules in PHP (`isPerSchemeSingletonKind`, eligibility)
- Concurrent `begin-approved-revision` can create two workspaces (delete-then-insert in transaction per user, **not** global unique)
- `revision_chain_position` not stored → listing-only

### Modification

- Flag without history row
- History rows without incrementing mod id (by design)

### State

- Dual `state` + `stage`
- Import approved stuck outside forward
- `status=0` + `state=requested` workspace

### Concurrency (**Confirmed / Strongly inferred**)

- Create/update use `DB::transaction` — good
- **No** `updated_at` version token / lockForUpdate on submit
- Two users: last write wins; two submits can both pass eligibility if no unique constraint
- Import in one transaction per migrate — good
- Linked draft `latest('id')` race: two drafts, submit picks max id

---

## 21. Concurrency / Transaction Recommendations

Atomic (already partly transactional — **extend locks**):

| Operation | Need |
| --------- | ---- |
| Submit / create revision | Transaction + `lockForUpdate` on current row + unique `(project_id, kind, is_current)` |
| Spawn workspace | Unique partial index: one open workspace per `(project_id, kind, created_by)` or per scheme |
| Create PCN | Same uniqueness for in-flight PCN per scheme |
| Import upsert | Keep unique `(source_table, source_id, kind)` — **already exists** |
| Increment counters | Eliminate or serialize via review insert only |
| Meeting link | Unique subject per meeting |

Prefer **optimistic** `lock_version` on `pc_forms` / PCN for editor conflict. Use **pessimistic** only on submit/revision spawn.

---

## 22. Coding Issues (severity)

| Sev | Issue | Evidence |
| --- | ----- | -------- |
| Critical | Dual workflow + dual draft can lose or mis-apply submit payload (`latest` draft vs in-place) | `PcFormService::update` |
| Critical | Import `parent_id=self` + default `approved` | `PcfmsPcMigrationService` |
| High | `PcFormService` god-class (~14k lines); business rules in controller + service | File size, `PcFormsController` |
| High | Meeting subject hydration guesses PCN | `MwsMeetingSubjectLink::hydrateMissingProjectConceptNotes` |
| High | No optimistic concurrency | Absent |
| High | `FOREIGN_KEY_CHECKS=0` on draft management-structure insert | `PcFormService` ~8124 |
| Medium | Magic strings for stages/states | Resolvers exist but many raw comparisons |
| Medium | Hardcoded meeting type columns | `allowsNewWorkingPaperGeneration` |
| Medium | Permissions: all kinds under `pc-forms` `add` | Controller |
| Medium | N+1 risk on listing with many relations | `show()` eager list is huge |
| Medium | Silent catch around revision workspace logging | `spawnPostApprovalRevisionWorkspace` |
| Low | Deprecated PCN stage constants still present | `PcnWorkflowStageResolver` |
| Low | `workpaper` column vs WorkingPaper model | Dual |

---

## 23. Flow / Business Issues (not mere bugs)

1. **Revision and modification are flags + row copies, but the product language implies numbered editions.**
2. **Imported forms are the same table but not the same lifecycle** until reconciled.
3. **PCN and PC Form look like a pipeline; they are peers.** Users can run both on one scheme without a defined handoff.
4. **Meeting type changes document behavior** (WP, defer hold) without being a formal workflow type.
5. **Draft and submitted rules differ by kind and by which draft store is used.**
6. **PC-I/II vs PC-III–V** are different products sharing one controller/service.

---

## 24. Frontend vs Backend

- Blade `can()` hides create/revise; backend repeats `can()` and service exceptions.
- Revision/modified eligibility is **AJAX** (`revision-availability`) then redirect — backend is source of truth (**good**).
- Listing **relabels** workspace as Draft while DB `state` is `requested` — UI/backend status mismatch (**Confirmed** comments in `PcFormService`).
- Autosave vs full submit: backend section-gated persist (`draftWorkspaceShouldSave`) — rules live in PHP; JS must send the same section keys or data is skipped (**Strongly inferred**).
- Business rules in JS (PCN modified resume in `crud-scripts.blade.php`) **mirror** service messages — keep backend authoritative.

**Move to domain:** eligibility, revision spawn, submit row-split, `is_modified` elevation — already mostly in services; **stop adding new rules only in Blade.**

---

## 25. Recommended Unified Architecture

### Target State (not current)

```
Scheme (project)
  └── Document (type: pcn | pc_i | …)
        ├── editions[]          immutable after submit
        ├── working_copy        0..1 mutable (draft / returned / workspace)
        ├── workflow            one engine (actor + stage)
        ├── reviews / forwards  append-only
        └── origin              user | pcfms (+ map)
Meetings reference edition_id (or working_copy only if explicitly allowed)
```

Principles:

- One lifecycle per document type
- One draft mechanism (working_copy)
- One edition mechanism (new row on publish of a new edition)
- `is_revised` / `is_modified` become **edition_reason**, not independent counters
- Import **normalizes then exits**
- Meeting type is MWS policy, not a second PC state machine

---

## 26. Recommended Revision / Modification Model

### What is a revision?

**A new submitted edition of an already approved (or import-historical approved) document, started as an explicit “Revised” action.**  
First-ever submit is edition 1, not a revision.

### What is a modification?

**A change to an in-flight or deferred document that remains the same edition** (same public identity / same `pc_uid` row), typically after return/defer.  
If policy requires a **new legal instrument** after approver reject, that is a **new edition** with reason `resubmit_after_approver_reject` — do not call it both revised and modified.

### Deterministic increments (target)

There is **no business need for two integer counters** given current code. Recommend:

- **`edition_no`** integer, per `(project_id, document_type)`, assigned **only when a working_copy is published** (`submitted_at` set, `is_current=1`).
- **`edition_reason`:** `initial` | `revised_after_approval` | `resubmit_after_return` | `imported`.
- Draft save → **no** edition_no change.
- Meeting / attachment / metadata on a draft → **no** edition_no change.
- Import historical rows → assign edition_no from PCFMS `VersionNo` **or** 1..n by sort order, then freeze.

**Same action ⇒ same edition behavior for imported and native** after ingest.

Drop sticky `is_modified` for **new** writes; keep column as generated: `edition_reason = resubmit_after_return`.

---

## 27. Recommended Import Architecture

```
External SQL
  → staging (optional tables or temp)
  → validate scheme map
  → normalize (parent_id null, edition_no, origin=pcfms, lifecycle_status)
  → write canonical pc_forms + sections
  → write pcfms_import_maps
  → stop
Normal HTTP workflow only
```

- Staging: drop after success; **map table is permanent**
- Source IDs: **only** in `pcfms_import_maps`
- Duplicates: existing unique key
- Re-import: update content, **do not** reset `edition_no` or destroy reviews
- History: inactive editions `is_current=0`
- Errors: fail the scheme group, not half a chain (today: one big transaction — **keep**)
- Idempotency: source unique key (**already correct**)

---

## 28. Recommended Permission Architecture

See §12. Final:

- **Keep** route-level `can('pc-forms'|'project-concept-notes', action)`
- **Add** document_class on create/revise (or separate routes already exist — grant per route)
- **Keep** MWS meeting-type permissions where they are
- **Do not** duplicate MWS verbs onto PCN/PC Form
- **Shrink** field-level grants to section-level if roles cannot be administered
- Actor assignment remains **data** (`current_actor_id`), not a permission

---

## 29. Keep / Merge / Deprecate / Drop

| Component | Keep | Merge | Deprecate | Drop | Reason |
| --------- | ---: | ----: | --------: | ---: | ------ |
| `pc_forms` + sections | ✓ | | | | Authoritative |
| `pc_forms_drafts` linked path | | | ✓ | later | In-place working_copy |
| Standalone drafts | ✓ short term | into working_copy without parent edition | | | UX for unsaved new |
| `is_revised` / `is_modified` | read | into `edition_reason` | writes | | Flags overloaded |
| `state` | read | from stage | | | Dual machine |
| `workflow_stage` + forwards | ✓ | extend to III+ | | | Canonical WF |
| Counter columns | read | | ✓ writes | later | Legacy |
| `PcFormService` | | split by use-case | | | Maintainability |
| `PcfmsPcMigrationService` | ✓ ingest | output via normalizer | special states | | Needed |
| `pcfms_import_maps` | ✓ | | | | Idempotency |
| PCN services | ✓ | shared edition policy | | | Parallel OK |
| `hydrateMissingProjectConceptNotes` persist | | | ✓ | | Wrong-link risk |
| Approve POST (I/II) | meeting/forum | | actor approve | | Forward canonical |
| Clone product feature | | | | N/A | Not found |
| Event sourcing | | | | ✓ don’t add | Overkill |

---

## 30. Target Data Model (logical)

```
DocumentEdition          ← authoritative submitted (or imported historical)
  project_id
  document_type          pcn | pc_i | …
  edition_no
  edition_reason
  is_current
  lifecycle_status
  origin                 user | pcfms
  parent_edition_id      previous edition (linear), not self
  created_by, submitted_at

DocumentWorkingCopy      ← 0..1 mutable (can be same table with row_role=working)
  edition_id nullable    (null = first create)
  same sections

WorkflowAssignment
  current_actor_id, stage, allowed_actions

Review / Forward         ← append-only

ImportMap                ← source_table + source_id + document_type → edition_id

MeetingSubjectLink       ← meeting_id + edition_id (type implied)
```

- **Current:** `is_current=1` edition, or working_copy if open
- **History:** other editions
- **Immutable:** submitted editions’ section rows (new edition = copy-on-write)
- **Mutable:** working_copy only
- PCN and PC Form **share this shape**; they stay separate types, not one table, unless you later unify (not required)

---

## 31. Migration Strategy

### Phase 1 — Audit (measure)

- Count `parent_id = id`, `parent_id is null`, multiple `status=1` per (project, kind)
- Count linked drafts per `pc_form_id`
- Count rows with `workflow_stage` null vs counters-only
- PCFMS status ID histogram vs `mapWorkflowState`
- Open workspaces per scheme

### Phase 2 — Introduce canonical columns

Add nullable `edition_no`, `edition_reason`, `origin`, `row_role`, `lock_version` **without** switching writes.

### Phase 3 — Normalize existing data

- Roots: `parent_id = id` → `null`
- `edition_no` = `revision_chain_position` or import order
- `origin=pcfms` if map row exists
- `row_role=working` if workspace or `state=draft`

### Phase 4 — Imported records

- Tag origin; set `lifecycle_status=approved` only when PCFMS says so
- Do not invent forward trails
- Allow `begin revised` using same spawn as native approved

### Phase 5 — Application

- New submits only create working_copy + publish edition
- Stop inserting linked `pc_forms_drafts` on save
- Forward engine for all kinds that still use HTTP workflow

### Phase 6 — Deprecate

- Stop writing counters except backfill
- Listing ignores draft table except leftovers

### Phase 7 — Remove

- Drop unused draft section tables after zero rows
- Drop `uses_forward_workflow` if universal

**Rollback:** keep old columns; feature-flag publish path; import map unchanged.

---

## 32. Backward Compatibility

Must remain readable:

- All existing `pc_forms` / PCN ids (WP, meetings, letters, MoM)
- `pc_uid` / `pcn_uid`
- Review/forward history
- PCFMS file links
- Inactive chain rows

Do **not** keep insert-always drafts or self-`parent_id` as **behavior** — keep **rows**.

---

## 33. Testing Strategy

Existing coverage is **workflow/forward/scope** heavy, **not** edition/import integrity (`tests/Unit/PcFormRevisionForwardSubmitTest`, `PcnModifiedRevisionTest`, forward feature tests).

### Unit

- Edition assignment on publish
- Draft save does not change edition_no
- `parent_id` normalization
- `mapWorkflowState` exhaustive PCFMS ids
- Permission: PC-II create without PC-I add (once split)

### Integration

- Create → draft → submit → forward → meeting → approve
- Approved → revised workspace → submit → new edition current
- Defer → same-row resubmit → edition_no unchanged
- Import → map → HTTP edit rules
- PCN modified workspace → publish
- Meeting link cannot hydrate wrong PCN

### Integrity

- Unique current edition
- Unique import source
- Concurrent submit (two processes)

### Critical regressions (never break)

1. PC-I first submit assigns actor
2. Post-AA-auth lock on approved row
3. Revised workspace not locked by parent AA
4. Defer hold WP generation per meeting type
5. PCFMS re-import does not wipe counts (until counters retired, then do not wipe reviews)
6. MoM signature gate for PC decisions (existing product rule)
7. Standalone draft delete does not delete a submitted form
8. Scheme eligibility on **create**, not same-scheme resubmit

---

## 34. Risk Assessment

| Risk | Level | Mitigation |
| ---- | ----- | ---------- |
| Listing/WP/MoM keyed by `pc_forms.id` | High if you renumber | **Never change ids** |
| Silent wrong PCN on meeting | High | Stop hydrate-persist |
| Import default approved | High | Status map + origin |
| God-class rewrite | Medium | Extract, don’t rewrite in place |
| Dual-write period | Medium | Feature flag |
| Over-fine permissions | Medium | Cap dimensions |

---

## 35. Final Recommended Flow (plain language)

1. An officer **creates** a PCN or a PC form (I–V) for a scheme (or scheme-less PCN).
2. They **save a working copy** as often as they want. That does not create a new official edition.
3. They **submit**. That working copy becomes the **current edition** and enters **one** review chain (forward to the next responsible officer).
4. Reviewers **forward, comment, or return**. Return lets the creator edit the **same edition**.
5. If a forum **defers**, the same edition is corrected and sent again. Meeting “hold” is a meeting policy, not a new form type.
6. When the edition is **approved** (or PCN cleared for CDWP), it is locked.
7. A later **Revised** action copies the approved edition into a new working copy. Submit creates **edition N+1** and makes it current. Old editions stay visible and unused for new meetings unless explicitly chosen.
8. **Imported PCFMS** records are loaded as editions with origin “imported”. After load they follow the same rules: approved stays locked; Revised starts edition N+1 the same way.

PCN and PC-I are **two document types on the same scheme**, not “PCN produces PC Form automatically.” A meeting is attached to a **specific edition**, not “the scheme in general.”

```text
Create → Working copy → Submit (edition N)
       → Review / Return (same edition)
       → Meeting (optional)
       → Approved (lock)
       → Revised working copy → Submit (edition N+1)
```

---

## 36. Executive Decision Checklist

### Keep

- `pc_forms` / PCN / section tables as content stores
- Forward trail + reviews
- `pcfms_import_maps` and CLI ingest
- MWS meetings with **explicit** subject FKs
- `revision_source_approved_id` / `modified_pcn_source_id` as lineage FKs

### Merge

- Draft stores → one working copy
- `state` + `workflow_stage` → one lifecycle
- PC vs PCN edition policy → shared rules
- Flag pair revised/modified → `edition_reason`

### Deprecate

- Insert-always `pc_forms_drafts`
- Counter-driven workflow writes
- Meeting-policy booleans on the document
- Persist-on-hydrate meeting PCN

### Drop (later)

- Unused draft section tables
- Self-`parent_id` convention
- Dead PCN stage constants (code only)

### Introduce

- `edition_no` + `is_current` + unique constraint
- `origin`, `row_role`, `lock_version`
- Import normalizer (roots, stages)
- Kind-scoped create permission
- Domain services extracted from `PcFormService`

### Highest-priority bugs

1. Submit may apply **wrong draft** (`latest id` vs in-place)
2. Import **self-parent** + **default approved**
3. Meeting **wrong PCN** hydration
4. Race on two revision workspaces / two current rows

### Highest-priority flow problems

1. Two approval machines
2. Two draft machines
3. Revision/modification are not numbers but product language pretends they are
4. Import forever special unless normalized
5. PCN/PC Form assumed pipeline — they are peers

### Highest-priority data problems

1. Multiple “current” interpretations (`status`, workspace, listing filters)
2. `parent_id` semantics split live vs import
3. Sticky `is_modified` without edition identity
4. No uniqueness for one in-flight document per scheme/kind

### Recommended permission model

Role → document class (PCN, PC-I, PC-II, PC-III, PC-IV, PC-V) → small action set; meeting class stays in MWS. **Do not** split every verb by meeting type on the form itself.

### Recommended revision model

**`edition_no` on publish only**; linear `parent_edition_id`; first submit = 1; Revised after approval = N+1.

### Recommended modification model

**Same edition**, reason `returned`/`deferred`. No second integer. Stop using `is_modified` as a second version axis.

### Recommended import model

Map → normalize → canonical editions → **same HTTP lifecycle**. Keep source keys. Do not keep a second workflow.

### Recommended target architecture

```text
Scheme
  ├─ PCN editions + 1 working copy + 1 workflow
  └─ PC-{kind} editions + 1 working copy + 1 workflow
        └─ Meetings / WP / MoM reference edition_id
Import ──normalize──► same editions
```

---

## 37. Confidence and gaps

| Topic | Confidence | Gap |
| ----- | ---------- | --- |
| No revision/modification integers | Confirmed | — |
| Dual draft + dual workflow | Confirmed | — |
| PCN not child of PC Form | Confirmed | Product intent (should they be sequenced?) **Uncertain** — not in code |
| PCN import | Confirmed absent | Future requirement unknown |
| Exact PCFMS `PC1StatusID` meaning | Strongly inferred from match() | Official PCFMS dictionary not in repo |
| Whether production data already has `parent_id=id` | Strongly inferred (code always writes it on import) | Needs Phase 1 SQL |
| Cancel/withdraw/archive | Confirmed not found as states | Soft delete / `status=0` only |
| API vs web | Confirmed web-centric for full CRUD | Sanctum **EAED** API writes standalone PC-I drafts only (`/api/v1/eaed`). No full PC/PCN REST CRUD. Product FR: `docs/PC-Form-Management-Functional-Requirements-Workflow.md` (may lag live forward-chain / MoM rules) |

---

*End of audit.*