Thanks for the example. I can see how a temporal/stochastic dimension can be important for some domains, but it seems like a crime to overload the relational FK semantics to support it. Just add one more layer of data abstraction (e.g. Org/Person <- temporal Customer profile <- Sale, or temporal Attributes -> Customer <- Sale), the RDBMS is made for this after all... Or use a DB with proper temporal support.
Comments
Thanks for the example. I can see how a temporal/stochastic dimension can be important for some domains, but it seems like a crime to overload the relational FK semantics to support it. Just add one more layer of data abstraction (e.g. Org/Person <- temporal Customer profile <- Sale, or temporal Attributes -> Customer <- Sale), the RDBMS is made for this after all... Or use a DB with proper temporal support.