Current DBMS's assume a time line starting at 1 A.D. or later and
consisting of days or seconds, up to 9999 A.D.. One
difficulty is that there are several definitions of second and of day. Another difficulty is that such a limited
time line is of little use to many potential users of a temporal
database, such as geologists, archæologists, anthropologists, and
astronomers. Such a time line doesn't even include all of recorded
history, and so doesn't fully support historians. Expanding the time
line back to the creation of the universe (approximately 15 billion
years ago), raises other definitional questions. For example, a solar
year in the time of the dinosaurs was 400 days long. A year is
difficult to define more than 6 billion years ago, before the earth
was formed.
What is needed is an application-independent identification of one or
more physical clocks
that cover all of past time (15 billion years) and all of the
foreseeable future. This definition of a physical time line should be
convertible to other definitions that might be useful. A representation as a
time-stamp data structure is also needed, with a precise semantics, i.e., a
correspondence with a particular time of this physical clock for each
valid bit pattern.
Decisions need to be made about treating events as infinitely
small points in time or as chronons of finite but nondecomposable
length, closed or open representations for intervals,
granularity, discrete versus continuous time, bounded versus unbounded
time, and linear versus branching time.