Glossary · Backbone
An engineering backbone is the spine of typed entities and relationships that every other tool federates around. This page is the architecture reference, what the backbone owns vs what federates, the cost ledger of drift when there is no backbone, and how it differs from a data warehouse or a data lake despite the overlapping marketing language.
A backbone is not the system that holds everything. It is the system that holds the structure: typed entities, identifiers, relationships, lifecycle states, cryptographic baselines. The leaf payloads, CAD files, test rigs, MES traveler data - stay where they are, owned by the tools that produced them. The backbone holds the spine; everything else federates around it through four predictable patterns.
Typed identifiers · relationships · lifecycle · baselines · audit trail
Requirement.id, Component.partNumber, the bidirectional verifiedBy edge between them, the lifecycle state (Draft / Reviewed / Released / Obsolete), the cryptographic baseline that includes them at v1.2.
CAD geometry · firmware binaries · large test datasets
The CAD vault keeps the .STEP file. The backbone holds Component.geometryUri → vault://creo/rev/B412. When a baseline freezes, the URI is captured immutably; the file content is fetched from the vault on demand.
Computed rollups · coverage matrices · audit packs
Mass rollup, requirement coverage %, certification readiness, recomputed live on the graph, exposed as materialized read-only views. Engineers and AI agents query them; no one writes them by hand.
ERP MBOM ↔ engineering BOM · MES asBuilt ↔ asDesigned
When the backbone changes an EBOM line, an event publishes to your MES; the MBOM acknowledges. AsBuilt records flowing back become typed events the backbone records as state transitions.
Concrete failure modes from the patchwork-of-tools approach, with order-of-magnitude cost. Each is structurally impossible against a typed backbone.
Three data architectures get pitched as "single source of truth". They are not the same category. Only one carries the engineering structure.
Requirement.id, Component.partNumber, Test.testId, guaranteed unique, never recycled, never reassigned. The audit trail can resolve any reference 20 years later.
verifiedBy, derivesFrom, integratesInto, hasRiskControl, relationships are first-class with their own constraints. The schema can validate ‘a Requirement must have at least one verification’.
Draft → Reviewed → Released → Obsolete. State transitions require approvers, leave audit records, propagate to dependent entities. State is not a label; it is a contract.
Content-addressed snapshot of the typed graph at freeze time. Replay deterministic, tamper-evident, fork-able. The historical record of every engineering decision.
Coverage matrices, mass rollups, risk dashboards, certification readiness, computed on the live graph, materialized read-only. The view is always consistent with the source.