Discussion

We first discuss the concept; then we discuss the name.

In the adopted definition, ``bi'' refers to the existence of exactly two times. An alternative definition states that a bitemporal relation has one or more system supported valid times and one or more system supported transaction times. In this definition, ``bi'' refers to the existence of exactly two types of times.

Most relations involving both valid and transaction time are bitemporal according to both definitions. Being the most restrictive, the adopted definition is the most desirable: It is the tightest fit, giving the most precise characterization (+E9).

The definition of bitemporal is used as the basis for applying bitemporal as a modifier to other concepts such as ``query language.'' This adds more important reasons for preferring the adopted definition.

Independently of the precise definition of bitemporal, a query language is bitemporal if and only if it supports any bitemporal relation (+E1), see Section . With the adopted definition, most query languages involving both valid and transaction time may be characterized as bitemporal. With the alternative definition, query languages that are bitemporal under the adopted definition are no longer bitemporal. This is a serious drawback of the alternative definition. It excludes the possibility of naming languages that may be precisely named using the adopted definition. With the alternative definition, those query languages have no (precise) name. What we get is a concept and name (bitemporal query language) for which there is currently little or no use.

Also, note that a query language that is bitemporal with the alternative definition is also bitemporal with regard to the adopted definition (but the adopted definition does not provide a precise characterization of this query language). Thus, the restrictive definition of a bitemporal relation results in a non-restrictive definition of bitemporal query language (and vice-versa).

The name ``temporal relation'' is commonly used. However, it is also used in a generic and less strict sense, simply meaning any relation with some time aspect. It will not be possible to change the generic use of the term (-E7), and since using it with two meanings causes ambiguity (-E9), it is rejected as a name for bitemporal relations. In this respect ``temporal relation'' is similar to ``historical relation.''

Next, the term ``fully temporal relation'' was proposed because a bitemporal relation is capable of modeling both the intrinsic and the extrinsic time aspects of facts, thus providing the ``full story.'' However, caution dictates that we avoid names that are absolute (-E6). What are we going to name a relation more general than a temporal relation?

The name ``valid-time and transaction-time relation'' is precise and consistent with the other names, but it is too cumbersome to be practical (-E2). Also, it may cause ambiguity. For example, the sentence ``the topic of this paper is valid-time and transaction-time relations'' is ambiguous.

We choose to name relations as opposed to databases because a database may contain several types of relations. Thus, naming relations is a more general approach.