ABSTRACT: Many autonomous and heterogeneous information sources
are becoming increasingly available to the user
through the Internet -- especially through the World Wide Web.
The integration of Internet sources poses several challenges
which have not been sufficiently addressed. In particular, knowledge
of redundancy  can be used to reduce the number of source accesses
that have to be performed to retrieve the answer to the user query.
Moreover, probabilistic information about source overlap can help derive
efficient query plans for delivering partial answers to queries.