In a perfect world, output from one program could be transparently used as input to another. The world of scientific computation is far from perfect, and its rich legacy of data and programs carries a major disadvantage: lack of interoperability. Business data processing largely solved this problem through common data models and shared databases, but current record-oriented database technology does not support scientific applications well. We believe object-oriented systems might alleviate such shortcomings, and are exploring object-oriented environments integrating diverse programs and data. Our approach constructs for each domain a unifying data model that creates a “plug-and-play” environment.
This paper outlines the utility of object data models and object databases, and describes three projects using them to model shared molecular structures. The first presents a conceptual model for chemical experiments, the second an object-oriented database to improve interoperability of ab initio chemistry programs, and the third a uniform object interface to diverse materials science datasets.