A schema mapping M = (\sigma,\tau,\Sigma) consists of a source
schema \sigma (i.e., a finite set of relation symbols), a target schema
\tau, and a set \Sigma of logical formulas that (a) specify the
relationship between the source and the target and (b) specify particular
properties that the target should have.
Data exchange deals with the following problem: given a schema mapping
M and a source database S (i.e., a finite structure of vocabulary
\sigma), construct a database T over the target schema \tau that satisfies
the relationships and properties listed in \Sigma. Such a target instance
T is called a solution for S with respect to D. Perferably, in case that
solutions exist at all, one would like to find particular solutions that
reflect the given source data "as accurately as possible".
In this talk I want to give an introduction to the area of data exchange,
with a special emphasis on the question of which solutions can be
considered to be "good" solutions. I will concentrate on notions of "good"
solutions for query answering over target instances in the contexts of
open world semantics and closed world semantics.