
This work is part of the DB
group's broader
research on
tools that facilitate the task of data integration on the Web.
The paper addresses a serious problem faced nowadays
by developers of applications that try to
access data from autonomous information sources
available on the Web.
Such information sources typically do not support all possible queries
against their data
due to commercial, privacy, or efficiency reasons.
For instance, Amazon.com allows applications to
retrieve all corresponding titles given an author name, but rejects
queries which list the entire book catalog. The
prevalent mechanism used for supporting only a
restricted set of queries is to package them as Web services. Past
works have shown how integration software can automatically
find ways to answer feasible queries, i.e., queries
that can be answered by combining and filtering the results of multiple
service calls.
The problem is that developers of integration applications need to
formulate feasible queries; infeasible queries were previously rejected
without
any explanatory feedback, trapping developers in a frustrating
trial-and-error cycle.