Evolving Relations

This paper presents a framework for evolving relation schemas that is based on conditional schema changes and tuple versioning. With each tuple a recorded schema and a conceptual schema is associated. This allows for a simple and semantically clean solution to the problem of schema mismatches that arise when the schema of a database is changed and some data no longer fits the schema. Specifically, no data needs to be migrated to the new schema, and no special null values are required. We precisely define evolving schemas in terms of schema segments and corresponding attribute mappings, present an algorithm to compute answers to queries over evolving schemas, and prove that the query answers consider the maximal set of schema segments consistent with the evolving schema.