Thinking about possible implementations we understand that many of its features and abilities (including late binding of multiply class implementations, "logical concurrency" mode, preliminary algorithm analysis, object level transaction serialisation) cannot be effectively implemented as a simple "syntax sugar" over existing DBMS and need some modification inside their cores. However we believe that these changes make a sense because they cover most important parts of offered approach and can vastly simplify the business system as a whole, make them faster, and ease access to business models from external programs with standard data access methods.