First up for myself is a SRFI to specify a general interface to existing persistence engines, probably relational, at roughly the level of for example Java Database Connectivity (JDBC).  It would not attempt any generalizing of SQL language specifics, but might add connection pooling to its remit.