An assessment of existing component-based software development methodologies and a holistic approach to CBSD

Software applications nowadays are getting ever larger and more complex. At the same time, users and sponsors of software applications have increasingly higher expectations: lower development cost, faster delivery time and higher quality of the products. This creates new challenges that the traditions of developing a software application from scratch every time a need arises, and of reusing code at a low-level of programming, are unable to address adequately. Component-Based Software Development (CBSD) is a strategic attempt to address these challenges by promoting extensive software reuse throughout the software development stages. This development strategy raises a spectrum of important issues. This research is primarily concerned with methodological issues such as system modelling, architecture and development process. This research validates two main hypotheses. The first hypothesis is that the theoretical basis of existing CBSD methods is weak. To test this hypothesis, existing CBSD methods need to be evaluated using an extensive and rigorous evaluation approach. This research identifies four publicly available CBSD methods and numerous approaches to evaluation of system development methods. These evaluation approaches are deemed unsuitable for the kind of evaluation envisaged by this research. Therefore, a new comprehensive framework for evaluating system development methods, called the MAP framework, is proposed. The existing CBSD methods are then evaluated using the MAP framework, which confirms the first hypothesis. The second hypothesis is that the limitations of the existing CBSD methods can be overcome. To test this hypothesis, various positive features of existing CBSD and non-CBSD methods that the MAP framework helps identify are synthesised, giving rise to NAVITA, a holistic CBSD method proposed by this research. The new CBSD method is then evaluated using the same criteria and rigour applied to existing methods. The evaluation confirms the second hypothesis. Furthermore, this research contributes to the application of Object-Oriented system development methods by proposing a set of principles that govern a rational allocation of class operations.

Name dc:type.qualificationname

phd

Level dc:type.qualificationlevel

doctoral

Grantor dc:publisher.institution

London Metropolitan University

Year dc:date.issued

2005

Author dc:creator

• Tun, Thein Than

• 000 Computer science, information & general works

Dc Identifier Grantnumber

N/A

OAI identifier oai:identifier

oai:repository.londonmet.ac.uk:7650