A central goal of conventional relational database design
database schema consisting of a set of relation schemas. In normalization theory, normal forms constitute attempts
schemas, and a wide variety of normal forms has been proposed, the most prominent being third normal form and Boyce-Codd normal form. An extensive theory has been developed to provide a solid formal footing for relational database design, and most database textbooks expose their readers to the core of this theory.
In temporal databases, there is an even greater need for database design guidelines. However,the conventional normalization concepts are not applicable to temporal relational data models because these models employ relational structures different from conventional relations. New temporal normal forms and underlying concepts that may serve as guidelines during temporal database design are needed. In response to this need, a range of temporal normalization concepts have been proposed,
normal forms. Considering that the different representations of the Checked Out relation model the same mini world and are capable of recording the same information, it may reasonably be assumed that these different representations would satisfy the same dependencies. At any point in time, a customer may have checked out several tapes. In contrast, a tape can only be checked out by a single customer at a single
determines Customer ID, but the reverse does not hold.
If we consider the information contents of a temporal relation, independent of its actual format, to be the set of conventional snapshot relations it logically comprises, we achieve
normalization theory that leads to a temporal theory, which naturally generalizes conventional dependencies and may be applied to dependencies other than functional. Specifically, a temporal relation satisfies a temporal dependency if all its snapshots satisfy the corresponding
notion of temporal dependency based on snapshots, a temporal normalization theory may be built that parallels conventional normalization theory and that is independent of any particular representation of a temporal relation. However, the resulting theory, while temporal in that it applies to temporal databases, is actually atemporal, in that it applies to each snapshot of a temporal relation in isolation. This theory therefore fails to account for ?temporal? aspects of data. Temporal data models generally define time slice operators, which may be used to determine the snapshots contained
temporal relation as their argument and a time point as their parameter, these operators return the snapshot of the relation corresponding to the specified time point.
Another approach to taking the temporal aspects of data into account during database design is to introduce new concepts that capture the temporal aspects of data and may form the basis for new database design guidelines .
Perhaps most prominently, time patterns may be used for capturing when the values of an attribute for an entity change in the modeled reality and in the database. For example, the set of tapes checked out by a customer may be expected to change substantially more frequently than the customer?s address, meaning that the addresses of customers and their checked out video tapes should be stored in separate relations.
Next, the concept of lifespan, that captures when an attribute of an entity has values, also has implications for database design. Specifically, if the lifespan of two attributes differ, null values of the unattractive ?do not exist? variety result unless the attributes are stored in separate relations. Assuming that the temporal data model used timestamps tuples, attributes should also be stored separately when different temporal aspects need to be captured for them or when the temporal aspects are captured with differing precisions (resulting in different timestamp granularities).