often contains information that can be used to
verify an input (selection) condition or fill in an output parameter
of a query. Consider the query in Example 1. Source 2
does not have information in its content description
that directly matches to the supplier attribute of Q or its
synonyms, but using the source category description
=Retail Book Store Company, we can easily
infer that the source name ``Barner & Nobel" of Source 2 is
a book supplier. A similar analysis applies to the Source 4.
We model the content of an information source in terms of
the object types and the object access constraints that
the source objects must satisfy. Each source object type is described by a
unary relation. Each access constraint is described using
a conjunction of built-in comparison atoms of the form
where
a is an attribute of a source type and v is a constant
drawn from a domain that is compatible to the domain of a.
Formally, given a source
of k relations, the content description of
,
denoted as
,
is described by a set of content records:
(i=1, ..., k)
are relations in
and
are access constraints over relations
.
We may view a source content description as a
collection of views defined over the source.
Each
describes one type of source objects.
Every object in
satisfies the insertion constraint
.