The NASA Astrophysics Data System: Architecture

The powerful discovery capabilities available in the ADS bibliographic serv ices are possible thanks to the design of a flexible search and retrieval s ystem based on a relational database model. Bibliographic records are stare d as a corpus of structured documents containing fielded data and metadata, while discipline-specific knowledge is segregated in a set of files indepe ndent of the bibliographic data itself. This ancillary information is used by the database management software to compile field-specific index files u sed by the ADS search engine to resolve user queries into lists of relevant documents. The creation and management of links to both internal and external resource s associated with each bibliography in the database is made possible by rep resenting them as a set of document properties and their attributes. The re solution of links available from different locations has been generalized t o allow its control through a site- and user-specific preference database. To improve global access to the ADS data holdings, a number of mirror sites have been created by cloning the database contents and software on a varie ty of hardware and software platforms. The procedures used to create and manage the database and its mirrors have been written as a set of scripts that can be run in either an interactive o r unsupervised fashion. The modular approach we followed in software develo pment has allowed a high degree of freedom in prototyping and customization , malting our system rich of features and yet simple enough to be easily mo dified on a day-to-day basis, We conclude discussing the impact that new datasets, technologies and colla borations is expected to have on the ADS and its possible role in an integr ated environment of networked resources in astronomy. The ADS can be accessed at: http://adswww.harvard.edu.