This paper presents the preliminary design of a new database management system, POSTGRES, which is the successor to the INGRES relational database system. The main design goals of POSTGRES are to support complex objects, allow user extendibility for data types, operators, and access methods, provide facilities for active databases and inference, simplify crash recovery code, take advantage of new technologies like optical disks and VLSI chips, and make as few changes as possible to the relational model. The paper describes the query language, programming language interface, system architecture, query processing strategy, and storage system for POSTGRES. The first goal is to support complex objects, which are more dynamic and complex than business data. The second goal is to allow user-defined data types, operators, and access methods. The third goal is to support active databases and rules, such as alerters and triggers. The fourth goal is to reduce the amount of code in the DBMS for crash recovery. The fifth goal is to make use of new technologies like optical disks and VLSI chips. The sixth goal is to make as few changes as possible to the relational model. The query language, POSTQUEL, supports complex objects, user-defined data types, and time-varying data. It includes features like iteration queries, alerters, triggers, and rules. The programming language interface allows applications to access the database. The system architecture includes a process structure, query processing strategies, and storage system. The storage system supports time-varying data, including historical data and versions. The system architecture also includes a mechanism for handling active databases and inference. The paper concludes with a description of the system architecture and the implementation of the system.This paper presents the preliminary design of a new database management system, POSTGRES, which is the successor to the INGRES relational database system. The main design goals of POSTGRES are to support complex objects, allow user extendibility for data types, operators, and access methods, provide facilities for active databases and inference, simplify crash recovery code, take advantage of new technologies like optical disks and VLSI chips, and make as few changes as possible to the relational model. The paper describes the query language, programming language interface, system architecture, query processing strategy, and storage system for POSTGRES. The first goal is to support complex objects, which are more dynamic and complex than business data. The second goal is to allow user-defined data types, operators, and access methods. The third goal is to support active databases and rules, such as alerters and triggers. The fourth goal is to reduce the amount of code in the DBMS for crash recovery. The fifth goal is to make use of new technologies like optical disks and VLSI chips. The sixth goal is to make as few changes as possible to the relational model. The query language, POSTQUEL, supports complex objects, user-defined data types, and time-varying data. It includes features like iteration queries, alerters, triggers, and rules. The programming language interface allows applications to access the database. The system architecture includes a process structure, query processing strategies, and storage system. The storage system supports time-varying data, including historical data and versions. The system architecture also includes a mechanism for handling active databases and inference. The paper concludes with a description of the system architecture and the implementation of the system.