Unifying Temporal Data Models via a Conceptual Model

To add time support to the relational model, both rst normal form (1NF) and non-1NF approaches have been proposed. Each has associated diiculties. Remaining within 1NF when time support is added may introduce data redundancy. The non-1NF models may be incapable of directly using existing relational storage structures or query evaluation technologies. This paper describes a new, conceptual temporal data model that better captures the time-dependent semantics of the data, while permitting multiple data models at the representation level. This conceptual model eeectively moves the distinction between the various existing data models from a semantic basis to a physical, performance-relevant basis. We deene a conceptual notion of a bitemporal relation where tuples are stamped with sets of two-dimensional chronons in transaction-time/valid-time space. Next, we describe ve representation schemes that support both valid and transaction time; these representations include both 1NF and non-1NF models. We use snapshot equivalence to relate the representational data models with the bitemporal conceptual data model. We then consider querying within the two-level framework. To do so, we deene an algebra at the conceptual level. We then map this algebra to the representation level in such a way that new operators compute equivalent results for diierent representations of the same bitemporal conceptual relation. This demonstrates that all of these representations are faithful to the semantics of the conceptual data model, with many choices available that may be exploited to improve performance.