# Placement Decision Matrix

Version: `0.1.0`

Status: `HISTORICAL EXPERIMENT MATRIX / STACK-B RECOMMENDATION WITHDRAWN`

Maintainer correction: this matrix compares placement for the bounded
nine-carrier projection only. It does not define Stack B's function and cannot
decide placement for the stack as a whole.

| Criterion | Transversal APPLICATIO capability | Existing OPERA alone | Minimal asymmetric stack candidate | Reject capability | Defer insufficient source |
|---|---|---|---|---|---|
| authority fit | strong: capability and provider resolution are explicit | weak: OPERA must not self-admit or own sources | possible, but creates unnecessary family identity | safe but discards demonstrated projection | safe if exact projection cannot be demonstrated |
| lifecycle | thin candidate then separate human creation/registration | would blur admission, execution, and mediation | larger lifecycle than evidence requires | none | research only |
| input admission | composed AREPO provider remains explicit | missing if OPERA is treated alone | can be local, but duplicates existing admission | n/a | n/a |
| execution ownership | exactly one effective OPERA can be selected | strong only for execution substep | possible | none | none |
| output mediation | composed SATOR provider remains explicit | missing if OPERA is treated alone | possible, but duplicates existing mediation | none | none |
| reuse | highest across domains and targets | risks root/global overreach | low until a genuine new class function exists | none | none |
| duplication | low | medium | high | none | none |
| source fit | nine carrier records supported; sentence template declared as projection | predicate application supported, whole closure not supported | no evidence for stack identity | carrier projection evidence would be ignored | justified only if traceable complete projection fails |
| five-owner clarity | explicit asymmetric resolution | incomplete | explicit but artificial locality | n/a | n/a |
| compiler interface | external typed IR consumed | could consume, but lacks orchestration | likely reimplementation pressure | unused | deferred |

## Former experiment recommendation

The experiment recommended `EXISTING_SURFACE`, specifically a thin APPLICATIO-local capability which
resolves existing five-role providers and selects OPERA T plus
`INTERPRETATIO_CONTEXTUALIS` for the execution substep.

“Existing surface” does not mean “already registered or executable.” Current
law requires a separate human decision to create a domain APPLICATIO. The
former recommendation therefore ended at a candidate contract and proof probe.
It was withdrawn as a Stack B disposition on `2026-08-23` pending functional
rewrite of the M2M.

## Rejected alternatives

- `EXISTING_OPERA_ALONE` collapses admission and mediation into execution.
- `MINIMAL_STACK_CANDIDATE` gives alphabetic symmetry architectural authority
  and duplicates existing providers.
- `REJECT_STACK_B` is correct only for the numbered-stack hypothesis, not for
  the traceable projection capability.
- `DEFER_INSUFFICIENT_SOURCE` is not required for the controlled-record
  projection, but remains correct for any claim that the historical source
  mandates nine natural-language sentences.