PM-DET-0010

For a given model release, an object's State SHOULD be a total function of that object's data and the data of the objects its Restrictions reference, independent of evaluation order, of any prior computed State, and of anything outside the model and the data.

Non-normative: OPEN, not SRC-, because no source states this. It is the property the sources' claims require*, reconstructed so that it can be attacked. It is written SHOULD rather than MUST deliberately: asserting MUST for something we cannot yet define would be exactly the overclaiming this specification exists to avoid.

Three obstacles were recorded here. Two are now closed by a close reading of the patent description, which had not been done when they were written:

1. state() is not yet a function. Restrictions in the sources' own style overlap and no selection rule is given — PM-STR-0030, MID-0001, Q-G-0001. Still open, and now the only one. 2. ~~Filters are not yet propositions.~~ Closed. The quantifier is existential on both positions — PM-FLT-0060, superseding PM-FLT-0040. 3. ~~The Condition evaluation point is undefined.~~ Closed. A Condition is evaluated against the State the Operation would achieve — PM-CND-0050, superseding PM-CND-0040.

The remaining obstacle is also smaller than it was. PM-STR-0040 establishes that exclusive States are expressible — the cardinality 0 form is first-class — so the defect is a missing requirement and a missing tie-break, not a missing capability. PM-STA-0040 establishes that an Operation achieves exactly one State, so the ambiguity is confined to classification and never reaches the transition function. PM-STA-0050 supplies a partial order over States that a tie-break rule could use.

That is a repair a conformance rule can make. It is still not made. Until Q-G-0001 closes and PO-0003 is discharged, the determinism claim must not be repeated as fact — in documentation, in a demo, or to a customer.