RFC-0030 · 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.
Q: How are 64-bit protobuf integers represented without precision loss?
Status: Open
Source: RFC-0030 §4; RFC-0006 Value types
Issue: int64 and uint64 map to Gnomon Integer, while the RFC leaves safe-range truncation to a runtime warning. It does not define whether Integer is exact across compiler, JSON, and runtime boundaries; a warning after rounding cannot recover an identifier or amount.
Recommendation: Define an exact 64-bit representation or reject out-of-range values before conversion, with wire-round-trip tests for signed and unsigned limits.
Q: How are protobuf enum numbers and unknown values preserved?
Status: Open
Source: RFC-0030 §4
Issue: Enums map to variants with Text names, but protobuf enum identity is numeric and future schemas may send values unknown to an older client. A name-only mapping cannot round-trip unknown numeric values.
Recommendation: Preserve the numeric value in a versioned enum representation, including an unknown-value arm, and specify name aliases and default zero behavior.
Q: Is a sub-millisecond google.protobuf.Duration allowed to lose precision?
Status: Open
Source: RFC-0030 §4.1 Issue: The adapter mandates truncation from nanoseconds to integer milliseconds and emits only a one-time note. This can turn a nonzero duration into zero and change deadlines or retry behavior. Recommendation: Preserve seconds and nanoseconds in a duration value or reject unrepresentable values at the boundary; do not silently truncate them.
Q: How is a protobuf schema pin reproducible across workspaces and offline builds?
Status: Open
Source: RFC-0030 §§7.2–8; RFC-0024 §6.3
Issue: The local specHash includes absolute file paths, so identical source trees checked out at different paths can hash differently. A Buf reference may omit a version, and the lockfile sample stores neither the descriptor bytes nor an immutable Buf revision, despite promising offline cache reuse.
Recommendation: Hash canonical descriptors with repository-relative source identities, pin an immutable Buf revision or digest, and store or reference the full verified descriptor/result snapshot through the canonical lockfile.