Summary
Lock in the v18 read-model decision that full graph materialization is a diagnostic/export/checkpoint compatibility operation, not the ordinary application read substrate.
This issue tracks the design for holographic slicing, TraversalOptic, NeighborhoodOptic, and streaming checkpoint basis construction.
Decision To Capture
- Full materialization remains legal for explicit plumber jobs such as checkpoint/export/diagnostic work.
- Application reads should move toward bounded holographic slices over worldline/coordinate bases.
- Checkpoints should manufacture CAS-backed basis artifacts by streaming admitted graph facts into shards/tries instead of requiring a heap-resident full WarpState.
- Traversal should be expressed as bounded neighborhood expansion plus an explicit traversal cursor, not as a hidden graph-wide read model.
- Missing basis, excessive tail, unbounded traversal, or global-scan requirements should fail closed with typed results/errors.
Expected Design Doc
docs/design/0271-v18-holographic-slicing-checkpoint-basis/v18-holographic-slicing-checkpoint-basis.md
Proof Direction
The resulting implementation work must include executable behavior tests that prove bounded reads, streaming checkpoint basis construction, no materialization fallback, cursorized traversal completeness, and machine-readable witness output. Design-doc assertions alone are not acceptance proof.
Summary
Lock in the v18 read-model decision that full graph materialization is a diagnostic/export/checkpoint compatibility operation, not the ordinary application read substrate.
This issue tracks the design for holographic slicing, TraversalOptic, NeighborhoodOptic, and streaming checkpoint basis construction.
Decision To Capture
Expected Design Doc
docs/design/0271-v18-holographic-slicing-checkpoint-basis/v18-holographic-slicing-checkpoint-basis.mdProof Direction
The resulting implementation work must include executable behavior tests that prove bounded reads, streaming checkpoint basis construction, no materialization fallback, cursorized traversal completeness, and machine-readable witness output. Design-doc assertions alone are not acceptance proof.