A distributed digital library architecture incorporating different index styles

The New Zealand Digital Library offers several collections of information over the World Wide Web. Although full-text indexing is the primary access mechanism, musical collections can also be accessed through a novel melody retrieval system. In offering this service over a three-year period, we have had to face many practical challenges in building, maintaining and administering diverse collections of different kinds of information, involving different search and retrieval systems, with different user interfaces. This paper describes the design of the software we have built to support the service. Interface server programs provide a uniform interface between the search engine and the client, irrespective of the nature of the collection. Search engines that embody completely different index styles operate under a single distributed framework-we describe as examples MG (Managing Gigabytes), a full-text retrieval system, and the MR (Melody Retrieval) system. A flexible protocol for communicating between an interface server and a search engine is defined. The resulting architecture simplifies library administration and the creation of new collections by providing a unified framework under which vastly different user interfaces and search engines can co-exist in a distributed computing environment.