How does Foundry handle schema evolution for datasets?

• A. Schema evolution allows adding, removing, or altering fields with backward/forward compatibility and versioned contracts.
• B. Schema evolution locks schemas forever.
• C. Schemas cannot change.
• D. Schema evolution only adds new fields with no compatibility.

Answer: (A)

Foundry treats dataset schemas as evolving contracts rather than fixed definitions. It uses versioned contracts to capture the exact structure of a dataset at each point in time, so you can change the schema—adding, removing, or altering fields—without breaking existing pipelines. When a schema change is made, a new contract version is created, and producers and consumers can declare or target the specific contract version they rely on. This enables compatibility between readers and writers across versions, supporting backward and forward compatibility depending on the nature of the change. In practice, old pipelines continue to operate against their established contract while newer ones can adopt the updated schema, with the system tracking and enforcing these versions.