Standards
9 min

Stardog vs GraphDB vs TopQuadrant: A Governance Comparison

Key takeaways

  • Two of the three are databases that can hold a vocabulary. One is a layer built to govern it. A demo will not show the difference.
  • Stardog governs write access to named graphs, not the meaning of a definition. Schema axioms stay readable regardless of named-graph permissions.
  • GraphDB custom roles are documented as markers with no inherent rights, and fine-grained access control needs an Enterprise licence.
  • PoolParty has real editorial governance, gated at Enterprise tier, and its approval workflow does not cover deletion.
In this article
Share

A taxonomy shortlist that contains Stardog, GraphDB and TopQuadrant is comparing three different product shapes against one requirement. Two of them are databases that can hold a vocabulary. One of them is a layer built to govern it. The demo will not make that obvious, because all three can show you a hierarchy of terms on a screen within ten minutes.

The separation matters once the vocabulary has more than one owner. This comparison runs the three against the capabilities that decide whether an agreed term stays agreed, and every claim below is taken from the vendor's own published documentation so you can check every one of them.

Stardog and GraphDB are RDF databases with real strengths. Neither documents an approval step that requires a second person before a model change takes effect, and that is the capability a governed taxonomy is built on.

What are you actually comparing?

Three products with three different jobs, which is why a feature grid laid flat across them misleads.

The three shapes

  • Stardog. An RDF database with a set of applications on top: Designer, Studio, Explorer, Knowledge Catalog and Voicebox. Platform documentation is at version 12.2.1. In June 2026 the company announced a new brand identity introducing the term "Semantic Control Plane". The product and company names did not change.
  • GraphDB. An RDF triplestore, described by its own documentation as a "Semantic Graph Database", compliant with RDF 1.1 and SPARQL 1.1. It is now sold under Graphwise, the merged entity that also owns PoolParty. The product documentation still lives on ontotext.com.
  • TopQuadrant. A governed layer that holds definitions and their approval history above whatever stores the data.

One note on scope before the comparison. Graphwise also sells PoolParty, which is a genuine taxonomy manager and is a much closer comparison to us than GraphDB is. It gets its own section below, because leaving it out would make this page easy to dismiss.

What governance does Stardog document?

Access control on where the model is written, and little above that level.

Four findings from Stardog's documentation

  • Permissions govern the named graph and stop there. Publishing from Designer requires write access to "each named graph Designer publishes to, including the data model named graph". That governs who may write. It does not govern what a definition means or who agreed it.
  • The model is readable even when the data is not. Their security documentation states: "Named graph permissions do not affect the schema axioms used in reasoning, and every reasoning query will use the same schema axioms. This is true even though some users might not have been granted explicit read access to schema graphs." The same page adds the qualifier, which is worth quoting in full: "However, non-schema axioms in those named graphs would not be visible to users without authorization." The data stays protected. The model does not.
  • There is no propose or approve action in the security model. The complete action list is read, write, create, delete, grant, revoke, execute, and "all", glossed as "Special action type that permits all previous actions over a resource." Nothing in that list separates the person who suggests a change from the person who accepts it.
  • The audit log has no approval event. Audit logging is real and thorough, described as "a superset of the access log", with published event types including SetOption, glossed "Sets a database's configuration option (metadata).", plus DataAdd, DataRemoveContext, UserPermissionGrant and RolePermissionRevoke. The published table contains no approval event type, because there is no approval step to record.

On versioning, stated precisely

This one is widely misreported and the precise version is more useful than the loose one.

Stardog 10 introduced schema versioning: with reasoning.schema.versioning.enabled set to true, "Stardog will automatically compute a 64-bit hash from the contents of the reasoning schema graphs." The documented operation is checking the current hash to determine whether the schema has changed. It tells you that something moved. It does not tell you what moved, who moved it, or what it said before.

Pinning does exist, and it is worth being accurate about. Schema multi-tenancy lets a query bind to a named schema, and Stardog's own motivating example is version pinning: "There are two different versions of a schema that evolved over time. Older legacy applications need to use the previous version of the schema, whereas the newer applications need to use the newer version." The qualification is that you maintain the old version yourself as separate named graphs. Nothing captures a prior version for you, and it is unconnected to the versioning hash.

Separately, Stardog 7.2.1 published a versioning API, com.complexible.stardog.api.versioning, with history navigation. That package is absent from the Stardog 12.2.1 package list and class index. We cannot say when it went, only that it is in the one and not the other.

What governance does GraphDB document?

Less than the word semantic in the product name suggests, and their documentation is candid about it.

Three constraints from GraphDB's own pages

  • Custom roles carry no rights. Their documentation states that custom roles "have no inherent rights or restrictions associated with them, and are instead just markers that can be freely granted to any number of users." A marker on a user is not a control on an action.
  • Shape validation is a decision made before you have a model. "SHACL support in a given repository must be enabled when that repository is created. You cannot modify an already existing repository by enabling the validation afterwards." Get that wrong on day one and the repository cannot be validated later.
  • Fine-grained access control is licensed separately. "This feature requires a GraphDB Enterprise license."

None of this makes GraphDB a poor database. It makes it a database, which is what it says it is. The error is the shortlist that treats it as a taxonomy governance platform.

What about PoolParty, the taxonomy product?

PoolParty is the strongest editorial governance of anything on this page other than us, and pretending otherwise would be dishonest.

It has changeset history at project and concept level, a four-state workflow moving concepts through Draft, Suggested, Approved and Deprecated, a workflow audit that records state, assignee, comment and date, and a propose-versus-approve split expressed as two distinct permissions, Workflow Assign and Workflow Actions. If you are managing SKOS thesauri and you need a review step, it does that work competently.

Two things qualify it, both from their own documentation.

Two documented limits

  • The workflow is an edition gate. "The approval workflow functionality is only available from PoolParty Enterprise Server level. It is available as an additional module for PoolParty Advanced Server." The capability that makes the product governable is not in the product you were first quoted.
  • Approval does not cover deletion. "Deletion of concepts are not covered by an approval process at the moment. This means a deleted concept can not be rejected to bring back a deleted concept." An approved taxonomy can lose a concept with no review and no documented way to reverse it.

That second point deserves weight. Removing a term from an agreed vocabulary is the highest-risk edit available, and it is the one edit the workflow does not gate.

How do they compare on the capabilities that matter?

Scored against what a governed taxonomy needs. A flat feature list would compare different things.

CapabilityStardogGraphDBTopQuadrant
Model authoring in the productYes, via DesignerNo, it is a storeYes
Change history on a definitionExport to external version control is the documented pathNone publishedIncluded
Retrieve what a term meant on a past dateNo documented retrieval of a historical versionNoIncluded
Propose and approve as separate permissionsNot in the action listCustom roles are markers onlyIncluded
Approval recorded with author, date and commentNo approval event type in the audit logNone publishedIncluded
Deletion governed by approvalNo approval step to applyNoGoverned like any other change
Model hidden from unauthorised readersSchema axioms readable regardless of named-graph permissionsFine-grained access control is an Enterprise licenceIncluded
Open W3C standards throughoutYesYesYes

The last row is not a differentiator and we are not claiming it as one. All three are built on RDF and SPARQL, and that is the right foundation for all three.

When is Stardog or GraphDB the right choice?

A comparison that answers "never" is marketing, so plainly.

If the job is storing and querying a large RDF estate at scale, both are capable engines with real track records, and you should evaluate them on that basis. This page is the wrong place for it. Stardog's reasoning and virtual graph capabilities are genuinely strong, and if your requirement is query-time inference across federated sources it belongs on the list. GraphDB has a long operational history and a large installed base. Teams whose vocabulary has one owner, whose model changes rarely, and who do not need to answer audit questions about it, will find either sufficient and should not pay for more.

One argument of theirs deserves a straight answer, so here it is. Stardog runs a portability case of its own, positioning against being locked into a proprietary engine. That argument is sound and it is not ours alone. Standards-based storage is a real advantage they have over property graph vendors, and our claim is not that they lack portability. It is that portability of storage and governance of meaning are different problems.

Where does TopQuadrant fit?

We built for the case where the agreement is the asset and the store underneath it is an implementation detail.

TopQuadrant holds the terms, their hierarchy, their mappings, their owners and scope, the record of who approved each change and what it replaced, and the version a downstream system is reading. Those capabilities ship in the product. No tier gates them. Deletion is governed on the same terms as any other change, because removing a term is a change.

The difference shows up in an audit. A demo will not surface it. When someone asks which definition was authoritative in March, who agreed it, and what it replaced, the system answers. Nobody has to remember who was in the room. For the category argument see taxonomy management, taxonomy governance for how a review process is actually run, and taxonomy and ontology platforms compared for the same evaluation applied across a wider field.

Every claim on this page is quoted from published documentation. Run the same test against TopQuadrant. We would rather be checked than believed.

Give your AI the context it's been missing

See how the TQ Data Foundation turns your enterprise knowledge into trusted, Al-ready context.