Process modelling for information system description

Stanczyk, Stefan K (1987). Process modelling for information system description. PhD thesis The Open University.



My previous experiences and some preliminary studies of the relevant technical literature allowed me to identify several reasons for which the current state of the database theory seemed unsatisfactory and required further research. These reasons included: insufficient formalism of data semantics, misinterpretation of NULL values, inconsistencies in the concept of the universal relation, certain ambiguities in domain definition, and inadequate representation of facts and constraints.

The commonly accepted ’sequentiality’ principle in most of the current system design methodologies imposes strong restrictions on the processes that a target system is composed of. They must be algorithmic and must not be interrupted during execution; neither may they have any parallel subprocesses as their own components. This principle can no longer be considered acceptable. In very many existing systems multiple processors perform many concurrent actions that can interact with each other.

The overconcentration on data models is another disadvantage of the majority of system design methods. Many techniques pay little (or no) attention to process definition. They assume that the model of the Real World consists only of data elements and relationships among them. However, the way the processes are related to each other (in terms of precedence relation) may have considerable impact on the data model.

It has been assumed that the Real World is discretisable, i.e. it may be modelled by a structure of objects. The word object is to be interpreted in a wide sense so it can mean anything within the boundaries of this part of the Real World that is to be represented in the target system. An object may then denote a fact or a physical or abstract entity, or relationships between any of these, or relationships between relationships, or even a still more complex structure.

The fundamental hypothesis was formulated stating the necessity of considering the three aspects of modelling - syntax, semantics and behaviour, and these to be considered integrally.

A syntactic representation of an object within a target system is called a construct A construct which cannot be decomposed further (either syntactically or semantically) is defined to be an atom. Any construct is a result of the following production rules: construct ::= atom I function construct; function ::= atom I construct. This syntax forms a sentential notation.

The sentential notation allows for extensive use of denotational semantics. The meaning of a construct may be defined as a function mapping from a set of syntactic constructs to the appropriate semantic domains; these in turn appear to be sets of functions since a construct may have a meaning in more than one class of objects. Because of its functional form the meaning of a construct may be derived from the meaning of its components.

The issue of system behaviour needed further investigation and a revision of the conventional model of computing. The sequentiality principle has been rejected, concurrency being regarded as a natural property of processes. A postulate has been formulated that any potential parallelism should be constructively used for data/process design and that the process structure would affect the data model. An important distinction has been made between a process declaration - considered as a form of data or an abstraction of knowledge - and a process application that corresponds to a physical action performed by a processor, according to a specific process declaration. In principle, a process may be applied to any construct - including its own representation - and it is a matter of semantics to state whether or not it is sensible to do so. The process application mechanism has been explained in terms of formal systems theory by introducing an abstract machine with two input and two output types of channels.

The system behaviour has been described by defining a process calculus. It is based on logical and functional properties of a discrete time model and provides a means to handle expressions composed of process-variables connected by logical functors. Basic terms of the calculus are: constructs and operations (equivalence, approximation, precedence, incidence, free-parallelism, strict-parallelism). Certain properties of these operations (e.g. associativity or transitivity) allow for handling large expressions. Rules for decomposing/integrating process applications, analogous in some sense to those forming the basis for structured programming, have been derived.

Viewing alternatives

Download history


Public Attention

Altmetrics from Altmetric

Number of Citations

Citations from Dimensions

Item Actions