Resource Discovery is the term commonly used to refer to the exercise of locating, accessing, retrieving, and managing relevant resources from widely distributed heterogeneous networks. The Resource Discovery Unit of the Research Data Network Cooperative Research Centre is actively working on tools and technologies which make these tasks easier in the Open Information Locator (OIL) project.

The OIL project takes a broad approach to solving the resource discovery problem. We have constructed a definition of resource discovery and are developing a conceptual model as a foundation for building and evaluating our work. Included in our definition are the following assumptions about resource discovery: resource discovery is a global problem; resource discovery systems must be scalable; resource discovery is, in its very nature, distributed; resource discovery does not imply any fixed structure or hierarchy of information.

A prototype resource discovery tool called HotOIL is currently under implementation. The HotOIL system can be viewed as a match-making service between a user's query and a vast range of information sources. It is a testbed for our research into a number of resource discovery issues including scalability, query routing, naming and meta-data, dynamic database access, information retrieval, and human computer interactions (HCI) . This paper outlines our research into each of these issues.

Scalability. Our research focuses on the scalability issues associated with finding and accessing a large and growing number of information providers. Similar scalability issues are being researched by the open distributed processing (ODP) community. We are investigating the ODP trader and interworking technologies as a solution to this problem, and implementing a resource discovery trader based on the X.500 directory service. Information providers register details of their service as a service offer within such a trader. Federation of these interworking traders provides a solution to scalability.

The federated network of resource discovery traders is fundamental to our solution for the problem of query routing. The trader will be used to determine what information sources to query by returning service offers from information sources that are relevant to a query. This solution also has meta-data implications as it is necessary to provide a characterisation of the information providers.

The effectiveness of resource discovery systems will rely on flexible and extensible naming and meta-data mechanisms as a key to accessing resources. Most meta-data work, including development of the Dublin Core meta-data set, is concerned with describing documents. However meta-data describing large-scale queriable collections of resources is important. The HotOIL system requires meta-data about the type of information served by an information provider, and information about how to access that data. Our work in this area includes the successful implementation of a URN resolution service which resolves IETF proposed URNs into URCs described by an enhanced Dublin Core meta-data set, and the development of meta-data for queriable resources.

Dynamically accessible databases, are important to support the needs of global resource discovery systems. We are applying the Z39.50 protocol to this problem. The meta-data supported by the Explain database in Z39.50 offers a resource discovery system the ability to dynamically discover not only information about the interface to an information provider, but also information about the kind of information served by the provider.

Information retrieval techniques are important to address many of the issues facing the resource discovery community. In our work, attention has been paid to improving precision in the query result. We have implemented a Query By Navigation (QBN) system that uses nonmonotonic inference. We plan to investigate QBN in the framework of a `global' hyperindex. This hyperindex will be automatically constructed from the meta-data of distributed resources. The resulting hyperindex can be considered as a conceptual schema of the contents of these resources.

While much attention is being paid to the more mechanical aspects of resource discovery, we view the HCI aspects of a resource discovery tool also to be vital to its success. Aspects of the formulation of a users information need is being investigated as part of the QBN system. We are investigating resource visualisation techniques for the presentation and control of large result sets.