ACDOCA Fold Plan — the SAP headline¶

The marquee claim, stated once. The Fold Engine folds SAP S/4HANA's Universal Journal (ACDOCA) into the same canonical model it folds iDempiere and Odoo into — to the cent, in a browser, with no HANA. This page is the doctrine and the plan for that headline: defined ahead of the folding engineers, so the work behind it follows a fixed target, not a moving one.

Companions: Cross-ERP Rosetta Stone (the term matrix) · ERP Rosetta Stone (dev idiom) · HolyGrail (the migration solvent, the SAP boundary) · Coverage Matrix. Execution card: prompts/SAP_FOLD_POC.md (the headless witness POC) + scripts/sap_adapter.js (the blind SD→FI hypothesis already mapped to the six verbs).

1. Why ACDOCA is the headline (and not a stretch)¶

Every other migration target makes you argue that one universal journal is the right shape. SAP already agreed. In the 2015 S/4HANA redesign SAP collapsed FI (BSEG), CO, Material Ledger and Asset Accounting into one line-item table — ACDOCA, the Universal Journal — and made document lineage an explicit graph, VBFA (document flow). That is exactly the engine's own shape:

SAP S/4 converged on	The Fold Engine has always been
ACDOCA — one append-style universal line-item journal	`journal` / `fact_acct` — one books table every doc posts into
VBFA — document flow as an explicit derivation graph	`kernel_ops` chain — the signed op-log, lineage by construction
Universal posting (one journal, many sources)	the closed verb set — `POST` folds every doc class into one journal

So the test is sharp, not lucky: a system whose own engineers concluded "the future is one journal of effects + an explicit flow graph" should fold cleanly under a model built on that premise. The headline writes itself — "we fold the Universal Journal SAP bet the company on" — and it is falsifiable: if even SAP's standard flow needs a verb outside the six, newVerbs will say so, loudly.

2. The fold target — ACDOCA → the canonical model¶

The standard Sell-side chain and its posting, mapped onto the canonical column (see the Cross-ERP Rosetta for the full term grid):

VBAK/VBAP  →  C_Order        (sales order)        ┐
LIKP/LIPS  →  M_InOut        (delivery)           │  the document chain — CREATE_DOCUMENT / CREATE_LINE / SET_STATUS
VBRK/VBRP  →  C_Invoice      (billing)            ┘
        ↓ posting
ACDOCA     →  fact_acct / journal                 the books — POST folds DR/CR, ΣDR==ΣCR
SKA1/SKB1  →  C_ElementValue                      the G/L account each line hits
VBFA       →  the kernel_ops chain                lineage / "what produced what" — the derivation spine (§0.13)

The six verbs carry it (the standing closed set, newVerbs=[] the acceptance bar): CREATE_DOCUMENT · CREATE_LINE · SET_STATUS · MATCH (GR/IR → M_MatchInv) · ALLOCATE (clearing) · POST (→ ACDOCA-shaped journal). sap_adapter.js already encodes this hypothesis blind; this plan is the path to proving it against real SAP output.

3. The SOP, applied to SAP¶

The universal protocol (unchanged — SAP is a harder instance, not a new method):

Re-key KNA1/LFA1→C_BPartner, MARA→M_Product, SKA1→C_ElementValue, the SD docs→C_Order/M_InOut/C_Invoice.
Project SAP status management (BSVZ) onto the DocStatus / C_DocType FSM frame (DR→…→CO).
Classify each rule by op-log effect — DERIVE / VALIDATE / ACT. Standard customising is declarative → folds; ABAP user-exits/BAdIs are imperative → effect-captured or plugin.
Absorb the declarative config + the document effects via the six verbs; route Z-code (per-engagement custom ABAP) to plugins — never auto-import.
Prove by diffing the folded journal to real ACDOCA to the cent (maxDiff=0c), with §FALSIFIER and newVerbs=[] as the two canaries.

4. The honest boundary (read twice — the asymptote)¶

SAP is the asymptote of the thesis, and the difficulty is not that the model is alien (§1 shows it fits). The real costs, named so the headline stays honest:

The oracle is costly. Odoo gave us source + a live oracle (odoodemo); SAP gives us output only — no readable ABAP, and a real posted oracle (IDES / a productive S/4) is licensed and hard to obtain. The adapter is therefore built blind from exported table rows. (Today: the /DMO/ Flight reference model is the toe-in — a master + lifecycle stub with no FI; see build/erp/14-sap-chain.json. It proves the DDIC world re-keys into AD; it does not prove a posting cycle.)
Z-code is out of scope by construction. Customer ABAP (user-exits, enhancements) is per-engagement and imperative → plugin lane, not fold. We fold the standard SD→FI flow.
Breadth is not parity. We aim for one document cycle to the cent, not "we did SAP." The IMG config sprawl is folded only where the standard flow touches it.

A leg that needs one of these is marked ⛔ with the missing fact — never green-washed (the lane discipline).

5. The plan — R → E → V, one leg at a time¶

Phase	Goal	Acceptance (the witness)	Status
S-0 · Oracle	Obtain real posted SAP output: `ACDOCA` + the SD chain (`VBAK…VBRP`) + `VBFA`, exported rows (IDES, a productive sandbox, or a vendor extract).	`§SAP-ORACLE present` (today: `§SAP-ORACLE unavailable`, honestly gated). The gating leg — resolve before any fold claim.	⛔ data-blocked
S-1 · Masters	Re-key `KNA1/LFA1 · MARA · SKA1` into `C_BPartner/M_Product/C_ElementValue`; counts == source.	shard counts == SAP `SE16` counts; §FALSIFIER (short shard → gap detected).	🟡 stub (`/DMO/` masters only)
S-2 · One O2C cycle	Fold `VBAK→LIKP→VBRK` and post to an ACDOCA-shaped journal; diff to real `ACDOCA`.	`§ACDOCA-FOLD maxDiff=0c` vs exported ACDOCA for one billing doc; `newVerbs=[]`; `VBFA` reproduced as the `kernel_ops` chain. This is the headline witness.	⛔ pending S-0
S-3 · Clearing / reverse	GR/IR (`MATCH`) + payment clearing (`ALLOCATE`) + reversal annihilate-to-zero.	`maxDiff=0c` vs ACDOCA clearing lines; reversal nets every account to 0.	⛔ pending S-0
S-4 · Scorecard row	SAP added to the critic scorecard on the honest axes — surpass / match / trail, each citing its witness.	one row, no hype.	—

The one thing that unblocks everything is S-0 — a real ACDOCA export. Until then the adapter stays a blind, falsifiable hypothesis (sap_adapter.js), the /DMO/ stub holds the master-mapping demonstration, and this plan states the headline as a target, not a claim. The moment S-0 lands, S-2 is the marquee: the Universal Journal, folded to the cent, serverless.

6. What a folding engineer does next¶

Read prompts/SAP_FOLD_POC.md (the witness POC) + scripts/sap_adapter.js (the verb hypothesis) + scripts/poc_sap_fold.js (the gated runner).
Pursue S-0: any path to a real exported ACDOCA + SD chain + VBFA (the artifact, not the system).
Fold S-2 through the existing six verbs; if a seventh is needed, stop and log it — that is the discovery, not a failure.
Diff to ACDOCA to the cent; add the scorecard row; cross-link back here.

The doctrine is fixed; the folding follows. ACDOCA is not a reach — it is SAP meeting us where the engine already stood.