Data lineage, provenance and rights are related—but they are not the same claim
A pipeline can show where data moved without proving who owned it. Provenance can show where an object came from without proving that every statement about it is true. A rights record can authorize use without changing the object's historical origin. Keeping those questions separate prevents metadata from becoming accidental authority.
Lineage
Data lineage usually describes movement and transformation: dataset A entered process B, which generated table C, which was used by model or report D. It is extremely useful for debugging, governance and impact analysis, but the existence of that chain does not by itself establish ownership, licence scope or external truth.
Provenance
Provenance adds attribution and historical context: which source supplied an object, which actor controlled a relevant event, which evidence supports the record, and which transformations led to the current state. ENTITY makes source, controller and rights-holder relationships explicit rather than collapsing them into one owner field.
Rights
Rights answer what an actor may do with a governed object—access, use, license, transfer, derive, settle or perform another authorized action under applicable terms. A provenance relationship alone does not create that authorization.
| Concept | Primary question | Does it automatically prove ownership? |
|---|---|---|
| Lineage | Where did the data move and transform? | No |
| Provenance | Where did it come from and what events led here? | No |
| Evidence | What supports a claim? | No |
| Rights | What is an actor authorized or entitled to do? | Only within the governing rights/legal context; the record is not itself a universal legal judgment. |
| Economic participation | What explicit terms create payment/participation consequence? | No; it requires explicit terms and applicable authority. |
Derived data makes the distinction more important
A source dataset can contribute to a derived output without automatically creating a royalty or ownership interest in that output. ENTITY's lifecycle can preserve the chain from source through usage and derivation into an economic event, but economic entitlement remains explicit rather than inferred from proximity in the graph.
BTDU topology does not change the rule
BTDU can represent reusable atoms, bonds, compounds and governed relationships efficiently. A bond can be computationally or semantically meaningful without becoming a property right. ENTITY remains the layer that governs authority, rights, provenance and explicit economic state.
Why this matters for AI and analytics
Training corpora, retrieval systems, analytics pipelines and agent workflows often combine many sources. A useful governance system should be able to preserve lineage/provenance while refusing to infer title, licence or causal economic entitlement merely because an input appears somewhere upstream.
External review question
Can you find any path in the released implementation where a lineage/provenance/topology relationship silently upgrades itself into ownership, permission or economic entitlement? If so, publish the exact counterexample and evidence.