Open Data Infrastructure
The Catalog Wars: The 2026 Open Catalog Landscape
The 2026 open catalog landscape across Polaris, Unity Catalog, Nessie, and Gravitino, with the lock-in questions buyers should ask.
The catalog is where the next round of data platform lock-in is moving. The table files may be open, but the catalog decides who can find, govern, operate, and trust them.
The catalog landscape is fragmenting and maturing
Open table formats moved control closer to the data owner. Catalogs decide whether that control stays open. In 2026, the landscape includes Apache Polaris for Iceberg catalog services, Unity Catalog as an open catalog project with Databricks roots, Project Nessie for versioned catalog workflows, and Apache Gravitino for federated metadata across data systems.
These projects are not interchangeable. They make different bets about governance, federation, versioning, API compatibility, and operational ownership.
Compare responsibilities, not logos
Use a responsibility map. Which catalog handles table operations? Which one handles identity and access? Which one stores governance metadata? Which one connects multiple engines? Which one can be self-hosted? Which one can be replaced without rewriting every workload?
Polaris is important because it focuses on Iceberg catalog needs and open catalog interoperability. Nessie is important because versioned catalog state changes how teams manage data changes. Gravitino is important because metadata federation is a real enterprise pain. Unity Catalog is important because governance and discovery are now central platform concerns.
The lock-in moves into metadata
A catalog can become the new warehouse if every policy, privilege, lineage reference, table identifier, and operational workflow depends on private behavior. That is not a reason to avoid catalogs. It is a reason to evaluate them harder.
Core idea: open tables without open catalog strategy are only half an exit path.
Questions for 2026 buyers
- Does the catalog expose documented APIs for the engines we need?
- Can we export or recreate metadata, policies, and table registrations?
- Does governance work across engines or only inside one platform?
- Can the catalog fail in a way operators understand?
The catalog war is not a side show. It is the control-plane fight.
Sources to start with
These are the primary sources I would start from when checking the claims in this piece.