Skip to content

Section-level ledger granularity: two proven rediscovery limitations #13

Description

@ryandmonk

Recorded from dspack-studio Phase 2 production verification (2026-08-03/04), as observed — no workaround applied.

1. Prop-level rediscovery cannot safely refresh within a human-owned components section. Once any component entry is enriched (e.g. a label prop authored — required, since discovery is variant-centric), the whole section is human-owned by hash mismatch, and fresh extraction facts about other entries (a new cva variant in source) cannot be applied: regenerateSections has no three-way merge base at sub-section granularity, and pretending otherwise would risk clobbering enrichment. Documented as a deliberate scope boundary in the 0.4.0 release.

2. Deliberately removed top-level components may be restored by pure-addition rediscovery. Proven in fresh-clone testing: an author restructured flat discoveries (info-card-header/body/footer) into composition.subComponents of info-card, deleting them as top-level entries; rediscovery re-added them, because at section granularity "deliberately removed" is indistinguishable from "never discovered."

In both cases no authored content is destroyed and every addition is reported — the no-force invariant holds. These are limitations of granularity, not correctness bugs.

Resolution is deliberately parked behind an explicit per-component ledger decision (see the forthcoming decision proposal); this issue is the durable record of the evidence.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions