An important function of an engineering database management system is to support incremental and cooperative design in an orderly way. This calls for a version-control mechanism of some generality if it is to be appropriate to the design of complex objects. The authors consider the nature of design objects, and develop some functional requirements for version control of such objects. They outline the basic object-oriented database model, which they take as the foundation for their proposal. They then discuss the relationship between versions and the objects they exemplify, describe the mechanisms for the creation of versioned objects and the conversion of non-versioned objects to versioned form, study the impact on referencing and type checking in the object model, and arrive at a flexible notion of a context containing rules for version selection. Their implementation of some of these concepts is also described.>

There is a requirement for a stronger treatment of intentional concepts and more general inferential ability within database systems. A framework for achieving this will be described in terms of extensions to an object data model.

Abstract Accepting the coexistence of a variety of computer languages as a fact of life, this paper seeks to make a virtue of necessity by proposing a modular framework within which these languages can be used. Examples of the types of languages considered are command languages, programming languages, data manipulation and definition languages, data communication languages and text editor languages. Aspects of syntactic modularity and semantic modularity are distinguished, and some answers to difficult questions are suggested, leading to a proposal for specific RUN and EXECUTE constructs that should be made widely available. Discussion is mainly at the level of the visible and conceptual concerns of the end user and the system architect, but the actual implementation of interconnection mechanisms is not totally ignored. A tutorial outline is offered as an indication of how the underlying ideas might be introduced to non‐specialists, and might even become a part of elementary education.

In order to define the semantics of PL/I in a form which is both precise and readable, a method of definition has been developed which employs an abstract machine operating on tree-structured data. The classes of trees involved are defined by formal grammars, while the algorithms governing the behaviour of the machine are expressed semi-formally in prose. This algorithmic metalanguage is itself intended to have largely intuitive semantics, but these are made more precise by an indication of how they could be supported at a lower level of detail within the abstract machine.