- Shopping Bag ( 0 items )
In its simplest definition, a domain is a collection of related products. Relatedness of products is a very broad and manifold concept. Products can be related if they perform similar tasks, if they use similar technologies, or if they are used in similar contexts. Domain analysis is the analysis of products in a given domain. Domain engineering is the production of products in a given domain. The distinction between domain analysis and domain engineering is not very sharp, because analysis and production are closely related. While the distinction exists in literature, most people now refer to a single concept: domain analysis and engineering (DA&E). This book describes the theory and the application of Sherlock, a methodology for DA&E.
Traditionally, the motivation for doing DA&E has been to foster software reuse. Reuse means that software artifacts-like requirement and design documents, code, documentation, and so forth-are used in several products, which is the scope of DA&E. The reuse approach requires (1) the creation of reusable artifacts and (2) the integration of those artifacts into final systems. In current practice, firms perform both the creation and the integration. The proportion between creation and integration varies, covering the full range between pure creators (e.g., library producers) and pure integrators.
Software reuse tackles the issues of productivity in software engineering and is seen as one of the most promising approaches to software production. Its potential benefits range from decreased development time and increased product quality to improved reliability and decreased maintenance costs. As an investment, reuse works only if the software fits a variety of situations and is actually reused many times. The larger the variety, the greater the potential for reuse and to reap the benefits of that reuse. However, dealing with variety bears some risks, like the fact that unknown or unforeseen requirements can challenge the reusability of software, or that quality can decrease as a consequence of adaptations. Variety is generally the result of variability, a leitmotiv of this book.
Variability comes from many sources: customers' unique needs, differences between versions, mutability of the environment, localization, and so on. All of these can potentially disrupt the efforts of reuse. A fundamental task of DA&E and Sherlock is to explore all the sources of variability within a domain. Sherlock is the recognition that variability is a characteristic of modern software that every software firm needs to confront. In this way, variability is no longer a source of trouble and unpleasant surprises. Rather, from a competitive viewpoint, it is a feature to exploit more effectively than your competitors.
Variability is a crucial part of decision-making in all software projects. Variability shows that a project can be undertaken in different waysadopting one of several alternative solutions, or implementing different strategies. Decisions determine the purpose and shape of the final products, and thus determine their impact on the market. Decisions can be made on a range of possible alternatives-which derive, for instance, from the market positioning, the technology, and the software architecture. Sherlock finds the benefits and the costs of the alternatives and provides the tools to evaluate these alternatives objectively. Moreover, it finds the interactions between the various aspects of variability-market, technology, architecture, and so on-which are often critical but also very difficult to dominate. In this way, Sherlock gives the DA&E practitioner the tools and confidence to make clear-cut decisions about software product strategies.
Decisions are made at all levels. Using a top-down approach, the following is a list of areas where decisions are encountered:
1. Strategic positioning in the market. Decisions concern the role of the firm and its products in the market. The elements to consider are the competitive relations in the market. The alternatives to choose from are those that allow the DA&E practitioner to achieve its long-term goals.
2. Definition of the strategic positioning in terms of product lines. Product lines are the ultimate implementation of a strategy. A product line is the combination of products that embeds a strategic idea. A product line has the breadth to support a strategy over a long period of time. Decisions here should determine the shape of the product lines, considering the relations between them and competing product lines. Decisions should also make clear how many product lines are necessary and, in case there is more than one, what different objectives they address.
3. Definition of product lines in terms of products. This is the DA&E practitioner's tactic. Here, decisions are made with regard to the articulation of each product line in terms of products. Decisions should evaluate the best options to ensure that products are cohesive in terms of the product line's strategy. Products should also support each other, exploiting the synergies (or network externalities) between them and other products. The result of these decisions can be a plan for the development and coordination of the product line's products.
4. Development of products. The decisions should determine which design and implementation solutions best support the development of the products. Development is conducted exploiting the benefits-like synergies and reuse-of working in the larger context of product lines.
By taming variability, Sherlock also makes it more friendly to reuse. More precisely, it formalizes variability into manageable schema through which reuse is possible. A major difficulty in the practice of software reuse is that reusability of a software artifact is difficult to assess. Reusability can only be proved in the future: A piece of software is not reusable until it is actually reused! A piece of software may seem reusable in the current situation, but, if the assumptions of the current situation change, reusability can be compromised. In stable domains, assumptions rarely change, which results in predictable levels of reusability. However, not all domains are stable.
Instability may come from several sources, such as immature technology or competition between firms pulling the domain in different directions. The goal of Sherlock is to foster software reuse in domains with high variability. On one hand, this is a matter of tracking the evolution of the domain, so that changes in the assumptions do not come as an unpleasant surprise...
|2||Domain analysis and engineering: the definition||9|
|3||An overview of Sherlock||15|
|8||Domain framework development||143|
|9||Case study 1: Sherlock applied to WaveRider Communications, Inc||177|
|10||Case study 2: Sherlock in a neural network domain||241|
|11||Case study 3: Sherlock applied to Novatel Wireless, Inc||277|
|12||Case study 4: Sherlock in a military domain||311|
|App. A||UML basics||363|
|App. B||Review of existing methodologies for domain analysis and engineering||373|
|About the Authors||381|