Glover, Steven James (2000) Inherently flexible software. Doctoral thesis, Durham University.
Software evolution is an important and expensive consequence of software. As Lehman's First Law of Program Evolution states, software must be changed to satisfy new user requirements or become progressively less useful to the stakeholders of the software. Software evolution is difficult for a multitude of different reasons, most notably because of an inherent lack of evolveability of software, design decisions and existing requirements which are difficult to change and conflicts between new requirements and existing assumptions and requirements. Software engineering has traditionally focussed on improvements in software development techniques, with little conscious regard for their effects on software evolution. The thesis emphasises design for change, a philosophy that stems from ideas in preventive maintenance and places the ease of software evolution more at the centre of the design of software systems than it is at present. The approach involves exploring issues of evolveability, such as adaptability, flexibility and extensibility with respect to existing software languages, models and architectures. A software model, SEvEn, is proposed which improves on the evolveability of these existing software models by improving on their adaptability, flexibility and extensibility, and provides a way to determine the ripple effects of changes by providing a reflective model of a software system. The main conclusion is that, whilst software evolveability can be improved, complete adaptability, flexibility and extensibility of a software system is not possible, hi addition, ripple effects can't be completely eradicated because assumptions will always persist in a software system and new requirements may conflict with existing requirements. However, the proposed reflective model of software (which consists of a set of software entities, or abstractions, with the characteristic of increased evolveability) provides trace-ability of ripple effects because it explicitly models the dependencies that exist between software entities, determines how software entities can change, ascertains the adaptability of software entities to changes in other software entities on which they depend and determines how changes to software entities affect those software entities that depend on them.
|Item Type:||Thesis (Doctoral)|
|Award:||Doctor of Philosophy|
|Copyright:||Copyright of this thesis is held by the author|
|Deposited On:||13 Sep 2012 15:46|