gnomonlearn
Docs/RFC-0049/Design Q&A

RFC-0049 · Design Q&A

Versioned draft snapshot

RFCs, reference catalogs, and evidence reflect Spec 3b2e3e7a114fb826b1445d39a48756b8fff629cd. Counts and pending/status notes belong to that snapshot. Read the dated project update for later progress.

Source repositories are private during staging. Links marked “repository access required” need access to the linked repository.

← Read the authored RFC

Q: What is the identity of a gnomon after field projection or rename?

Status: Open

Source: RFC-0049 §§3.2, 5, 6, 14 invariant 9 Issue: fields and rename produce new subgraphs with altered data but retain the underlying gnomon identities. Union and intersection are set operations, yet the RFC does not define equality or conflict resolution when two differently shaped copies of the same gnomon appear in one composition. Recommendation: Separate immutable gnomon identity from projection view shape, and define canonical set equality and merge behavior for every composition operator.

Q: How do filtering and lifting preserve edge-endpoint closure without widening access?

Status: Open

Source: RFC-0049 §§3.2, 4, 5, 9.3 Issue: Selecting an edge while filtering out one endpoint violates closure; auto-adding its endpoint can reintroduce a gnomon that failed authorization or selection. byId also claims a singleton with zero edges for any gnomon, which cannot represent an edge alone. Recommendation: Specify edge-drop or endpoint-retention semantics for each operator, and require authorization before closure adds any endpoint.

Q: What is the meaning of union across different bitemporal snapshots?

Status: Open

Source: RFC-0049 §§3.4, 7.1, 14 invariant 4 Issue: Union of subgraphs at different qualifiers chooses the latest effective time, but the union still contains facts from both snapshots. The result may not be a valid subgraph of the carrier at that latest time, contradicting the claimed temporal meaning and algebraic identities. Recommendation: Require matching qualifiers for ordinary set operations or represent mixed-time results as explicitly tagged comparison data outside the closed algebra.

Q: How is user authorization enforced for service and system projections at every output path?

Status: Open

Source: RFC-0049 §§4.1, 9.3, 10.4, 11 Issue: Service and system projections bypass caller filtering; §10.4 only says they must be wrapped by a query that re-applies filtering. Other materialization, caching, or view paths can expose the elevated subgraph, and a service-level aggregate can leak information without revealing individual gnomons. Recommendation: Make declassification an explicit, audited capability with typed output policy and forbid direct materialization of elevated projections to caller-visible surfaces.

Q: Which aggregate and cross operators are part of the normative projection grammar?

Status: Open

Source: RFC-0049 §§6.2, 8, 10.1–10.2 Issue: cross, count, sum, avg, and groupBy are described and used in examples, but the normative projection-expr grammar has no production for them. cross also says it yields a subgraph without synthetic pairs, leaving its cartesian-product semantics unspecified. Recommendation: Add typed grammar and evaluation rules for each supported operator, or remove unsupported operators and examples from the normative surface.

Q: How is an empty aggregate materialized without emitting forbidden null?

Status: Open

Source: RFC-0049 §§8.1, 9.2; RFC-0006 §4.A Issue: Empty avg, min, and max return null, while the type and wire contracts forbid null and say optional fields are omitted. A scalar result has no field to omit, so its result type and error behavior are undefined. Recommendation: Use an explicit Option or result variant in the source type and wire schema, with a deterministic empty case.

Q: Which provenance statements are guaranteed for ordinary runtime mutations?

Status: Open

Source: RFC-0049 §§12.2, 13, 15.4; RFC-0039 §6.5 Issue: The mutation text and core.provenance example assume template.instance_of, template.with, and caller lineage statements, but RFC-0039’s execution result and RFC-0041’s install provenance do not define those runtime writes. The cited RFC-F is not part of the current corpus. Recommendation: Define the creation/provenance contract in an owned RFC, including statement identities, rollback behavior, and migration of legacy graph writes, before claiming every gnomon has lineage.