Partially shared views: a scheme for communicating among groups that use different type hierarchies

Reviewer: Craig Partridge

Converting from one data representation to another is critical to distributed systems that support heterogeneity. Different hosts, let alone applications written in different programming languages, cannot communicate effectively unless the data conversion problem is addressed. This paper examines the data conversion problem in an object-oriented environment that supports object type inheritance. The principal application examined is an object-oriented electronic mail system in which users can define their own message object types. The paper exhaustively enumerates the possible conversion schemes and then endorses a hybrid solution that combines object inheritance with translation rules. In the solution, systems share a limited set of common object definitions. When systems exchange an object not in this common set, the receiver of the object can (1) import the type definition; (2) translate the object into a local object type using locally defined translation rules; or (3) coerce the object into a type that is shared by both systems and is an ancestor of the type being exchanged. Unfortunately, this paper makes no reference to the standard literature on data representation conversion. As a result, it is almost impossible to determine if the paper makes a contribution. In particular, the authors do not address the m × n argument [1], which argues that given m systems sending data to n distinct systems, writing m × n translation rules is a nuisance and, more important, every time an additional system is added, another m translation rules must be implemented. Thus, the argument runs, peer-to-peer translation schemes such as the paper proposes are undesirable from a software engineering point of view. Also, although I suspect that the paper's notion of object type inheritance in data conversion is more sophisticated than those proposed previously, the paper appears to have been written without knowledge of prior work, such as that in network management [2].