The ubiquity of the World Wide Web offers an ideal opportunity for the depl
oyment of highly distributed applications. Now that connectivity is no long
er an issue, the attention has turned to providing a middleware infrastruct
ure that will sustain data sharing among Web-accessible databases. We prese
nt a dynamic architecture and system for describing, locating, and accessin
g data from Web-accessible databases. We propose the use of flexible organi
zational constructs service links and coalitions to facilitate data organiz
ation, discovery, and sharing among Internet-accessible databases. A langua
ge is also proposed to support the definition and manipulation of these con
structs. The implementation combines Java, CORBA, database API (JDBC), agen
t, and database technologies to support a scalable and portable architectur
e interconnecting large networks of heterogeneous and autonomous: databases
. We report on an experiment to provide uniform access to a Web of healthca
re-related databases.