Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

0029: Content-format declarations and recorded registry guidance

Context

Authors need to distinguish a malformed complete declaration from valid spelling whose registration or meaning is unknown. RFC 9880 Appendix A note 2 references RFC 9193 Section 6, whose decimal grammar has no digit ceiling; Section 3 separately gives the Content-Format identifier range. A small convenient registry subset would turn known dated positives into misleading unfamiliar results. Parameter and payload meaning require work beyond declaration checking.

Decision

An explicit fixed ten-area content-format assessment adds complete grammar and dated registry guidance to the immutable initial inventory, following ADR 0003: Check findings and completeness and ADR 0013: Shared workflow and report guarantees. FullBaseline joins both declaration dimensions into M024 and its pattern/format area while preserving producing condition identities. Final effective Full coordinates define the observation population, following ADR 0020: Owned effective views and nearest origins and ADR 0024: Effective alternatives and nullable characteristics. Overrides, deletions, critical-feature gates and Direct donor uncertainty retain their established meanings.

Whole borrowed grammar classification precedes decimal range checking under recorded interpretation I-CF-RANGE-1. Dated guidance uses all relevant positives from pinned CoAP, media-type and HTTP-coding editions, with canonical identities and temporary, obsolete and deprecated flags preserved. A positive-table miss leaves current registration unknown; actual table use does not imply completed lookup or current status. This follows ADR 0002: Recorded standards baseline.

UseRecorded and WithholdRecorded share the existing evidence-policy and borrowed metadata seam established by ADR 0027: Unit names and recorded registry guidance. Withholding is an embedding’s evidence trust choice. It reports needed unavailable evidence without pretending the compiled table is physically absent. Parameters have explicit excluded-meaning Notices in ExpandedContentFormat. FullBaseline additionally uses ADR 0033: Content-format model meaning for bounded parameter, registered tuple and ordered coding meaning; reserved identity remains interpretation unknown. No AST, payload decoder, protocol binding or refresh machinery is added.

The content-format contract records the exact grammar, extracted identities, dates, digests, status distinctions and conditions. Mandatory aggregates and actual-use evidence survive complete optional omission. Shared source-bound ownership, two complete DTO generations, old-result fallback and independent output growth fit the unchanged admission reserves. Every new semantic visit uses the original finite operation allowance, following ADR 0009: Bounded processing and reporting.

A complete grammar/range traversal precedes registry replay over the same borrowed derived coordinates. Operational completion remains separate from unknown semantic premises; replay never clears earlier uncertainty or duplicates findings and observations. A finished population with no reliable query needs omits replay. This preserves later independent declaration errors when lookup exhausts work, while deliberately spending additional traversal/classification credit and moving first consultation later under a tight capability. The unit module’s ordering and the original allocation/work allowances remain unchanged.

Considered options

A short popular-format allowlist was rejected because unfamiliar misses would omit available dated positive knowledge. Runtime XML or registry refresh was rejected because it would change reproducibility, evidence trust and capability. A public parser AST or payload validation was rejected because callers need declaration conclusions and preserved effective locations, not decoder machinery. Giving declaration checking its own ad hoc Full membership was rejected because it cannot establish a complete model inventory. The separate immutable model inventory now supplies that relation. Parameter semantics within the expanded inventory was rejected because syntax and registry evidence cannot establish parameter-specific rules or payload compatibility.

Interleaving classification and lookup was rejected because a lookup stop can leave later declarations unvisited behind a false completed syntax area. Marking both areas unfinished on every lookup stop is a truthful smaller alternative, but loses useful independent syntax completion and later spelling/range findings. Retaining classifications for later lookup was rejected because it introduces a growing population owner where charged borrowed replay suffices.

Consequences

Callers opt into added checking without altering the initial meaning of scalar getters or the old inventory. Reliable syntax errors remain blocking errors, while missing registration evidence and unsupported meaning are incomplete. Recorded positives and dated flags are Notices, with no current-status claim. Tight finite capabilities can leave requested work unfinished; detail omission alone cannot change semantic status. Table, backend, header or reporting graph changes require renewed source-bound qualification.