The complexity of soft deletes is that they implicitly introduce the difficult semantics of bi-temporality into the data model, typically without the benefit of

The complexity of soft deletes is that they implicitly introduce the difficult semantics of bi-temporality into the data model, typically without the benefit of a formal specification that minimizes the number of edge cases that have to be dealt with. Mechanically, I've typically supported soft deletes with audit tables that shadow the primary table, with a bunch of automation in the database to make management mostly automagic. It isn't too bad in PostgreSQL.