Fixes C-1/C-2. Incremental change-driven maintenance (notify->delta) is
primary; full rebuild is a rare, envelope-respecting, concurrent fallback
(not required cheap). Equivalence via blocking/LSH candidate-gen + verify +
incremental maintenance, replacing O(N^2). Index is derived, per-tenant.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Fixes bug B-2. States the guarantee (read-your-writes for journal-owned
coordination-canonical state; causal across the derived tier; eventual+
freshness-labelled for sharded inputs). Conflict detection+representation =
core mechanism, resolution = policy. Overlay-apply-under-drift semantics
(fast-forward / three-way / refuse+re-present) and journal ordering.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Fixes bug B-1: page identity is a stable assigned handle (survives edits),
not a content fingerprint; fingerprints identify versions/content for the
equivalence mechanism. Chain: identity -> placements -> equivalence. (§7.2)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The optimal architecture synthesised from INTENT + the full research arc:
- Thesis: canonical at the edges, derived in the middle (orchestrator not engine)
- Dual narrow waist: adapter contract (15 capability spectra) + page model
- 6 layers + provenance/capability rails; L4 union/projection is a rebuildable cache
- Federation-model taxonomy (plural/composable); two-axis projection model;
moldable view registry; identity != placement; computational content in scope
as page-model+projection, out as execution platform
- Concrete src/ module layout with downward-only dependency rule
- Canonical data flows; policy surface; tradeoffs; full traceability to INTENT/UCs
References ArchitectureBlueprint.md as the L5 authorization sub-blueprint.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
