A library wants curated catalog facts under Unity Catalog lifecycle control and raw vendor MARC landings left in an external path. How should tables be chosen?

Library catalog architecture should split lifecycle ownership: managed Delta tables for curated catalog facts so Unity Catalog owns location and lifecycle, and unmanaged/external tables for raw vendor MARC landings that may remain on an external path the library does not want UC to fully lifecycle-manage. That split matches bronze landings versus curated gold under governance.

Using managed Parquet tables for curated catalog facts gets the external landing side right but fails curation because Parquet lacks Delta Lake transactional guarantees, versioning, and managed lifecycle optimizations that curated catalog facts require. Using unmanaged/external views for vendor MARC landings is wrong because views are logical queries without storage; raw MARC files left on an external path need unmanaged/external tables to register the landing location. Using foreign tables over JDBC for vendor MARC landings is wrong because landings are raw files on an external path, not live tables in a remote database to federate via JDBC.

Exam caveat: managed versus external is about who owns storage lifecycle and how data is registered—not a claim that landings are unimportant; landings often stay external while curated facts are managed. Operational check: register curated catalog facts as managed Delta in Unity Catalog, point vendor MARC landings at an external location as unmanaged/external tables, confirm DROP/lifecycle behavior differs as intended, and ensure librarians query curated managed tables for analytics rather than raw MARC paths.