If I understand the Manifesto correctly, the metadata db can be any SQL database but the client needs to be DuckDB + DuckLake extension no ?
*for now. The principle in the client side (especially read only) should be the same with Iceberg. Ideally there's an Iceberg adapter for clients.
Good point. I think - any ducklake implementation for any SQL compliant database will work.
Of course, the performance will depend on the database.