Temporal conversion functions for multitemporal relational databases

Abstract

Most research in the field of algebras for temporal relational databases concerns the study of operations within each of the formats transaction-time, valid-time, bitemporal and snapshot. Nevertheless, data of different temporal format (multitemporal relational data) should be allowed to interact. Two possible scenarios for this problem are: (1) federated databases, where the interaction is required of databases of different temporal format; (2) databases supporting selective versioning, where, due to the different update frequencies, the attributes are maintained in tables of different temporal formats. A solution to this interoperability problem is neither straightforward nor univocal, since the semantics differences among such data are quite deep. As suggested in [DGS94] for the multitemporal natural join, a first step for the definition of a multitemporal relational algebra can be the conversion to a common temporal format for the execution of operations. The semantic criteria at the basis of the temporal conversion functions in [DGS94] have been long debated for their safeness in the deduction or reconstruction of temporal information. In this paper we revise such criteria, discuss their problems and present several alternative solutions.