Exploiting Structured Data in Wide-Area Information Systems

摘要

Information produced by outside parties, such as the World Wide Web, isincreasingly important in many software applications. Effective use of this information requires the ability to exploit its semantic structure. Unfortunately, existing wide-area information systems force data to be distributed in either a lowest-common-denominator form, or in a form meaningful only to programs designed around a particular application. The former results in significant loss of information, while the latter severely limits the information available. One solution is to develop a way to describe information types with rich semantics, allow new type descriptions to be easily added and related to existing ones, and construct agents that can distribute and interpret this type information. The extra weight of general-purpose network object systems like CORBA and OLE is not required. Providers can create and manage their data and data types as they wish, and clients can adapt them to their needs. The Typed Object Model (TOM) implements such a solution. It treats information as typed abstract data objects, with encodings defined for compatibility with existing infosystems, such as the Web. Mediator agents register and relate new data types, and locate agents that can interpret or convert unfamiliar data formats.