Shared abstract data types: An algebraic methodology for their specification

Abstract

In this paper a methodology for specifying data in a transaction system is proposed. We present the outline of a style of shared data definition, by which it is possible to avoid the unnecessary constraints on concurrency usually imposed by traditional transaction management approaches. The used method is based on the definition of the data as abstract data types in the well known algebraic method. The semantics of the data is given in two steps: the usual semantics of the operations and the specification of concurrency constraints among them. Uniformity in the semantics definition allows the user to employ semantic knowledge needed to exploit the maximum level of parallelism permitted by the expected notion of consistency on data.