Due to the semantic conflict between what the user wants to query (such as books from book store or publisher suppliers) and the capability of the data sources, only four of the data sources can actually contribute to the answer of Q. They are Source 2, 4, 6, 8. The first problem is the overhead of contacting the rest of the data sources that do not contribute to the query. For instance, Source 1 will be contacted by most of the simple keyword-based search tools, even though it is actually not able to contribute to the answer of the query Q, because Source 1 contains reviews for books published during 1970 and 1980 whereas the query Q asks for books published in 1996.
The second problem is the risk of receiving too many irrelevant data items in the query result. For instance, Source 5 will be contacted and used to answer the query Q, because it is an online book store database. The results returned from Source 5 would be all reviews about books on subject Cancer, published after 1970, even though the query Q only asks for those books published in 1996. This is simply because Source 5 does not accept a particular year as the input of external queries, nor provides the year as the output parameter for external queries.
To avoid or reduce the overhead of contacting irrelevant sources at early stage of the query processing, and to guarantee semantically correct answers, it is quite clear that there is a need of a metadata specification language that can be used to describe the content and capability of data sources and the scope and context of user queries. There is also a need of transformation methods that can establish a dynamic interconnection between information consumers and producers.