In a federated database environment, different constituents of the fed
eration may use different temporal models or physical representations
for the temporal information. This paper introduces a new concept, cal
led a temporal module, to resolve these mismatches among the constitue
nts. Intuitively, a temporal module hides the implementation details o
f a temporal relation by exposing its information only through two win
dowing functions: one function associates each time point to a set of
tuples and the other function links each tuple with a set of time poin
ts. A calculus-style language is given to form queries on temporal mod
ules. Temporal modules are then extended to resolve another type of mi
smatch among the constituents of a federation, namely, the mismatch in
volving different time units (e.g., month, week, and day) used in reco
rding temporal information. Our solution to this mismatch relies on ''
information conversions'' provided by each constituent. Specifically,
a temporal module is extended to provide several conversion functions;
each converts its information to a different time unit. The first ste
p to process a query addressed to the federation is to select suitable
conversion functions from the extended temporal modules. In order to
do so, time units are formally defined and studied. A federated tempor
al database model and its query language are proposed. The query langu
age is an extension of the calculus-style language above.