E.17.0:4.8 - Keep selection for one describing use separate

For one current describing use, always name the use. Add one singular viewpointRef : U.ViewpointRef only when selecting P changes what the use reads or checks or what a relying use may conclude; otherwise omit the reference. When present, resolve it under the effective reference scheme to exact P. ViewpointId is P’s designator; designator, reference, episteme, and describing use remain different objects.

When the viewpoint matters, that describing use selects P for itself only. The selection does not establish conformance or U.View membership, give E a new C.2.1 identity, reidentify E, or create a universal selection relation, legacy context tuple, bounded-context object, or generic model-use identity field. Another use may select another P while E remains unchanged. A use needing several viewpoints first identifies the C.13 collection and its exact membership; it does not overload viewpointRef with a collection value.

The architecture therefore keeps exactly two positive dependent-kind rules—P as U.Viewpoint by its fixed self-contained content about an exact target kind or, conditionally, by fixed content about selected S; and E as U.View by obtaining conformance—and two direct relation kinds: viewpoint-convention dependence and E/P conformance. D remains optional for a named A.22 use; the two local explicit-result ValueKinds remain optional for named evaluation consumers. Families carry exact U.ViewpointRef values only when a named use needs viewpoint selection. C.2.1 identity, MethodDescription, A.6.3 construction, E.24.PUB publication, C.29 representation, and unrelated interfaces retain their separate identities and rules.