RDF & SPARQL Patterns Guide — Remediation Status

Unit ID: rdf-sparql-patterns-remediation State: ✅ Complete and closed. Executed on 2026-09-23. Several of its fixes were later found incomplete or wrong by a further review and corrected by iri-patterns-post-3866b21-remediation (see “Later corrections”, below). All three deferred items are now dispositioned (see Deferred, below). Nothing further is owed by this unit. Plan: rdf-sparql-patterns-remediation.md — superseded in execution by docs/developer/review/IRI-patterns-remediation.md, which the human designated authoritative when comparing it against this plan Sketch: None Governing review: docs/developer/review/IRI-patterns-remediation.md, executed against iri-identity-patterns.md/ADR-A82 as the identity authority Primary target document: rdf-sparql-patterns-guide.md

What changed and why this status supersedes the pre-execution plan below

The original plan (still preserved unmodified for its analysis) treated several findings as blocked on a human design decision (B1(b), B3, B8, B9, B11). The human’s execution instruction changed the resolution strategy for all of them: LATTICE does not mandate a single answer to a genuine deployment trade-off, so every one of those “open decisions” was resolved by adding a dal: property or class to ontology/persistence/spec/persistence.ttl (and, where the trade-off has a discouraged-but-valid option, a sh:severity sh:Warning shape in ontology/persistence/shapes/constraints.ttl) and repointing the guide’s prose at the configuration choice instead of a mandated fix. No design decision was made unilaterally on the framework’s behalf; each became a documented, adopter-facing choice.

Findings disposition (post-execution)

Later corrections (2026-09-23, iri-patterns-post-3866b21-remediation)

The rows above record what this unit did. Where a later review found the fix incomplete or wrong, the current position is:

Deliberate scope reduction

The guide’s worked examples use a 16-digit zero-padded revision-IRI width throughout as an illustrative simplification, while several places also claimed “19 digits in production” — a self-contradiction (C3/A-1). Mechanically regenerating every worked example in this ~3000-line guide to a consistent width was judged disproportionate to the defect: the actual bug is the guide asserting two widths are both normative. The fix applied throughout (§10.1, §19.4, Chapter 13) states the width is a per-deployment profile decision (dal:DigestScheme, iri-identity-patterns.md §10.2), fixed once and never mixed, and stops presenting 16 or 19 as competing recommendations. A follow-up slice can regenerate every example to a single width if a consistent visual convention is wanted; this pass did not do that. Done later by iri-patterns-post-3866b21-remediation (B14): every example is now 19 digits.

Deferred (not attempted in this pass)

1. Compiler wiring. Handed over to persistence-compiler-iri-sync, which owns it from here. Closed 2026-09-25: that unit’s Slices 1–5 are complete, 774/774 tests passing, human-validated; Slice 6 (documentation close-out) is its own final pass.
2. New TCK test bodies. Written as specifications in guide Chapter 27 by iri-patterns-post-3866b21-remediation (T-12 to T-17, the S-suite and the R-suite). The note above that “T-1 through T-13” existed was wrong: only T-1 to T-11 existed then. Executable TCK code belongs to the store SPI TCK (epic P0.5.6).
3. Fixture file for Worked example 4 (examples/identity-epoch-privacy-profile.ttl). Handed over to persistence-compiler-iri-sync and authored in its Slice 3 (2026-09-23); its privacy part is exercised by Slice 4.

Validation performed

All edits were verified with the editor’s get_errors check (Markdown lint only) after each batch, with no errors reported at any point. No SHACL, SPARQL, or build/test command was run, per this repository’s Default Mode (.github/copilot-instructions.md): build, test, and validation commands are for the human to run and report back on.

Commands to run

From the repository root:

mise run check:persistence

Expected: the Turtle in ontology/persistence/spec/persistence.ttl and the shapes in ontology/persistence/shapes/constraints.ttl parse and validate cleanly, and no naming collision is reported against the pre-existing dal: vocabulary. If a Markdown link checker or prose linter is part of the standard mise task set, run that too, since this pass added several new internal cross-references (iri-identity-patterns.md anchors, dal: shape names) that a linter would catch faster than manual review.

A pass is: no parse errors, no validation errors, no broken internal links. A failure in mise run check:persistence most likely means a dal: name used in the guide’s prose does not exactly match the vocabulary (this pass cross-checked every reference against persistence.ttl directly via grep_search and corrected three mismatches found that way — dal:HardDeleteReplay, dal:AbortAndRetry/dal:ExternalLockOrder/dal:SinglePartitionedWriter, and dal:PruneWithLiveHeadGuard/dal:NoPruning were all invented names not present in the vocabulary and have been corrected to the real ones, but a residual mismatch elsewhere is the most likely remaining failure mode).

Pre-execution plan (preserved for reference; superseded by the disposition above)

Findings disposition

Pre-execution blockers

Nothing else blocks starting the mechanical (Decision required: NO) fixes listed in the plan’s ledger, but Part D’s sequencing notes mean several mechanical and decision-gated fixes share the same source blocks (notably §19.1, §10.1, and §19.4) and should not be edited twice — implementation should wait for all decisions touching a given block before editing it once.

Next steps

1. Human resolves the five open design decisions in Part F items 1-5.
2. Implementer confirms the fnd:replacedBy question in ontology/foundation/ (Part F item 6).
3. Review’s author completes finding C3, or the human explicitly accepts this plan’s independent cross-check as sufficient to proceed on a stated width choice without it.
4. Once resolved, decompose the plan into one or more implementation slices per copilot-instructions’ mandatory slice shape (code/doc change, Validation Pack, traceability update, doc delta), following the sequencing order in the plan’s Part D.
5. Each slice’s Validation Pack must include at least one adversarial case per Blocking finding it closes, per the human validation gate’s mutation-check step, given several of these defects (B2, B8 in particular) are exactly the kind of “test passes while the system is corrupt” bug that discipline exists to catch.