propagation-is-topology-complete-and-inconsistency-safe

IN derived (depth 4)

Truth propagation is both topologically complete — reaching all transitively dependent nodes including those connected through outlist entries, not just antecedent references — and safe under graph inconsistency, skipping dangling dependent references with structured warnings rather than crashing, ensuring correct cascading even in networks with imperfect structural integrity such as nodes deleted without full reference cleanup.

Summary

When a belief changes, the ripple effect correctly reaches every downstream belief that needs updating, even ones connected through less obvious dependency paths. At the same time, if the dependency graph has broken links — like references to nodes that were deleted without cleaning up all pointers — the system logs a warning and keeps going instead of crashing. This means propagation is both thorough and robust, so you can trust cascading updates even if the graph has gotten a bit messy over time.

Justifications

SL — Completeness and inconsistency-safety are independent properties that together make propagation production-ready for real-world networks

Antecedents (all must be IN):

  • incremental-propagation-is-fully-complete — Incremental truth propagation reaches every node whose truth value should change — including nodes that depend via outlist entries — without requiring periodic full recomputation, because safe terminating BFS traversal operates over a dependents index that tracks both antecedent and outlist relationships.
  • propagation-is-safe-under-graph-inconsistency — Truth propagation achieves correctness even when the dependency graph contains dangling references: missing nodes are skipped with structured warnings rather than crashing, dangling IDs are excluded from both the changed and visited sets, and this graceful degradation composes with the underlying termination and lifecycle-awareness guarantees for all reachable nodes.

Dependents

These beliefs depend on this one:

Details