underwent a major makeover — adopting a new name, logo and motto, “People Committed to You.” The bank, then less than $250 million in assets, now $370 million, committed to a new two-year sales and product training program for its retail staff. Its management, hoping to boost revenue (noninterest income was 10.8 percent of net operating revenue in 1998), also sought to explore new profitgenerating product lines. To help assess opportunities in the insurance and annuity market, in 1998 PeoplesBank executives commissioned Michael White Associates (MWA), an ICBA Preferred Services Provider for insurance consulting, to map out a suitable strategy. The company produced a fiveyear economic assessment report. Its verdict: The bank’s prospects were good for selling annuities and life insurance offerings through a dedicated, full-time life insurance specialist. The assessment viewed property-casualty insurance as offering the bank fewer opportunities. Personal lines auto and homeowners insurance were least attractive. Michael White Associates also performed a strategic evaluation and financial valuation of a leading PeopleBank’s acquisition target — a property-casualty insurance agency. MWA recommended that the bank not purchase the agency. Unable to recruit suitable life insurance specialists, the bank and MWA agreed that the use of licensed bank employees would produce the forecasted annuity volume. PeoplesBank moved ahead with a recb y M ic h a e l D . W h it e S Plan of ACTION
Combinations of C++ features, like inheritance, templates, and class nesting, allow for the expression of powerful component patterns. In particular, research has demonstrated that, using C++ mixin classes, one can express layered component-based designs concisely with efficient implementations. In this paper, we discuss pragmatic issues related to component-based programming using C++ mixins. We explain surprising interactions of C++ features and policies that sometimes complicate mixin implementations, while other times enable additional functionality without extra effort.
Multi-paradigm design is a metaparadigm: it enables to select the appropriate paradigm among those supported by a programming language for a feature being modeled in a process called transformational analysis. A paradigm model is a basis for multi-paradigm design. Feature modeling appears to be appropriate to represent a paradigm model. Such a model is proposed here for AspectJ language upon the confrontation of multi-paradigm design and feature modeling. Subsequently, the new transformational analysis is discussed.
Product instantiation is one of the less frequently studied activities in the domain of software product lines. In this paper, we present the results of a case study at Axis Communication AB on product instantiation in an industrial product line, i.e. five problems and three issues. The problems are concerned the insufficiency of functional commonality, features spanning multiple components, the exclusion of unwanted features, the evolution of product line components and the handling of initialization code. The issues discuss architectural compliance versus product instantiation effort, quick-fixes versus properly engineered extensions and component instantiation support versus product instantiation effort. The identified problems and issues are based on the case study, but have been generalized to apply to a wider context.
An application framework is a collection of classes implementing the shared architecture of a family of applications. A technique is proposed for defining the specialization interface of a framework in such a way that it can be used to automatically produce a task-driven programming environment for guiding the application development process. Using the environment, the application developer can incrementally construct an application that follows the conventions implied by the framework architecture. The environment provides specialization instructions adapting automatically to the application-specific context, and an integrated source code editor which responds to actions that conflict with the given specialization interface. The main characteristics and implementation principles of the tool are explained.
In this paper, we describe an XML-based language, called XVCL, for managing variants in component-based product families. Using XVCL, we can organize product family assets and instrument them to accommodate variants. A tool that interprets XVCL and provides semi-automatic support for asset customization is also introduced. In our projects, we applied XVCL to manage variants in UML domain models and in generic architectures for product families. We have achieved simple forms of separation of concerns (in both models and architectures) and we are investigating advanced forms in current work. We plan to compare XVCL to other emerging techniques that lead to separating of concerns in software models, documents, architectures and code.
Software reuse and especially the paradigm of software components are promising approaches to increase the efficiency of software development. One of the basic problems of software reuse, however, is the tradeoff between the abstraction from project-specific aspects and the actual contribution of a reusable artifact. Stringent resource constraints further complicate the application of these approaches in domains, where efficient and therefore specialized solutions are required, e.g. in the domain of embedded systems. Generic components - designed to be adaptable to new application scenarios - allow to overcome these limitations, esp. if they automate the essential modifications. This paper presents a concept of generic components that has been developed to facilitate the construction of highly specialized embedded operating systems. Besides the illustration of the underlying concept, the paper discusses the external representation of generic components and the internal realization of the required variability and reflects some of our experiences in constructing generic components.
With Aspect-Oriented Programming (AOP) a new type of system units is introduced (aspects). One observed characteristic of AOP is that it results in a large number of additional (coarse-grained to fine-grained) system units (aspects) ready to be composed to the final application. With this growing number of system units the dependencies between them become vast and tangling. This results in the necessity of an improved management of the dependencies between these system units. Our paper investigates this problem, proposes a more general model (version model) to capture different facettes of AOP as well as a partial solution towards unit consistency based on versions.
In this paper we describe our experiences with the development of a platform in the medical imaging domain. Three important ingredients of this platform are components, interfaces and information models. We will explain the requirements for the platform, why these three ingredients have been chosen, and our experiences when using this approach.
In order to compose components, we have to adapt them. Therefore, we pursue a transformational approach focusing on the communication view. We show how to separate the definition of communication from the definition of other system aspects, how to extract this definition from existing systems, and how to weave it back into the system. Our main concern is the reconfiguration of this aspect.
This paper presents a concrete application of template meta-programming techniques to provide a generic, efficient and type-safe implementation of statistical estimator computation using visitors.
The composition of software units has been one of the main research topics in computer science. This paper addresses the composition validation problem evolving in this context. It focuses on the composition for a certain kind of units called aspects. Aspects are a new concept which is introduced by aspect-oriented programming aiming at a better separation of concerns. Cross-cutting code is captured and localised in these aspects. Some of the cross-cutting features which are expressed in aspects cannot be woven with other features into the same application since two features could be mutually exclusive. With a growing number of aspects, manual control of these dependencies becomes error-prone or even impossible. We show how assertions can be useful in this respect to support the software developer.
In component-based software development, the mechanism by which the software components which best satisfy a user’s query are retrieved is indispensable. However, conventional retrieval methods cannot evaluate the total characteristics of a component, because they consider a single aspect of the component or require an additional description. In this paper, we propose a new distance metric, “directed replaceability distance” (DRD), which represents how different two components are in detail from the viewpoint of structure, behavior, and granularity. We have developed a retrieval system which uses DRD as a measure of the difference between a user’s query prototype component and the components stored in a repository. In this paper, we outline the concept of DRD and the usefulness of our retrieval system.
Despite advances in software engineering and process methodologies over the past decades, not many IT projects seem particularly well adapted to today's fast-paced world. Software developers must start to acknowledge change and even uncertainty as a given, rather than the exception that should be studiously avoided, and they must adapt their techniques accordingly.Some business domains have seen attempts to address this situation. Several workflow vendors, for instance, have been marketing change and end-user programmability as major assets of their products. But, in general, they have been strangely ignorant of (good) modern software engineering practices, and the results have not really lived up to the claims. But we may expect a revival on a grander scale: the ability to (re) define the business logic on the fly is becoming a crucial asset when businesses re-align their core processes around the Internet.The Internet is transforming the way we envision and design applications. While we could build yesterday's simple Web applications with, let's face it, primitive techniques, this is simply no longer true. High-volume databases, long-term transactions, interoperability, distributed objects, re-use, these are some of the technical issues that must be dealt with. But the real challenge will be to leverage all this technology: we must empower the user to set up, maintain and change his applications more easily.We need dynamic systems, where applications can be changed at run-time in a high-level way, preferably by end-users. Above all, we need appropriate architectural techniques. In this paper we explore the use of dynamic object models. It turns out that the basic concepts are fairly simple. As for the difficulties, we can borrow solutions from many disciplines in computer science. If we do it right, we can even make the system work for itself.
Aspect-Oriented Programming on the one hand supports a separate treatment of different concerns in software development. On the other hand it provides "weaving" technologies for knitting together such individual concerns in software systems. Since the aspect-oriented approach is an extension of the classical OO-paradigm it requires an enhancement of well-known language constructs on the meta level. Although some general-purpose aspect languages (GPAL) are available in the meantime, no commonly accepted object model has yet been proposed. Consequently a common terminology is still not available what substantially hinders the spread of good and useful concepts. This paper proposes an object model, which represents the foundation of the GPAL Sally. We compare our model in respect to AspectJ which is by far the most popular and well-established aspect language and, therefore used by a wide community.
It is well known that modifying software by hand, especially someone else’s, is not only slow and tedious but so notoriously error-prone that we strive for components that never need changing — black-box building blocks. Unfortunately, we lack the omniscience required to engineer such parts, especially when our problem domains are ill-defined. A promising alternative is to make each component adaptive — a simple mark-up language (e.g., frame technology) converts each component’s properties (i.e., granularities that are meaningful within a given domain) into a default value that can be overridden or extended by other components in a rapid yet reliable manner. Adaptive components constitute a gray-box strategy — black when their properties fit the context as is, otherwise white. The approach works by automating the tedious and error-prone aspects of construction and modification, while localizing all the unique properties of an object or program in its root component. Each root exercises complete control over the sructure and content of the code that its hierarchy of components emits. Such custom roots constitute only 5-15
Class hierarchies, though theoretically reusable, have generally not seen much practical reuse in applications, due in part to the inflexibility of the inheritance relationship. We present a technique, base class injection, that allows the creation of generative class hierarchies that may be adapted by adding new methods, data members, and ancestor classes without modification to the class library code; an implementation of this technique in the C++ language is given.
The C++ language has only a limited runtime type information system, which doesn’t provide full reflection capabilities. We present a general mechanism to support reflection, exploiting template metaprogramming techniques. Two solutions are presented: a static one where metaclass information is only available at compile time to produce class specific code; and a dynamic one where metaclass objects exist at runtime. As a case study of technique we show how to build an object interface to relational database tables. By just annotating a class definition with meta information, such as storage attributes or index properties of fields, a programmer can define objects that can be stored, fetched or searched in a database table. This approach has been used in building a high-performance, full text search engine.
Template techniques in C++ allow a modest degree of generative programming: creating specialized code for specialized problems. This use of templates has been controversial; indeed, one of the oft-cited reasons for migrating to Java is that it provides a simpler language, free of complexities such as templates. The essence of generative programming in C++ is not templates - the language feature - but rather the underlying algorithms in the compiler (template instantiation) which unintentionally resemble an optimization called partial evaluation [12,18]. By devising a partial evaluator for Java, we reproduce some of the generative programming aspects of C++ templates, without extending the Java language. The prototype compiler, called Lunar, is capable of doing "expression templates" in Java to optimize numerical array objects.
We propose a standard problem to evaluate product-line methodologies. It relies on common knowledge from Computer Science, so that domain-knowledge can be easily acquired, and it is complex enough to expose the fundamental concepts of product-line methodologies. As a reference point, we present a solution to this problem using the GenVoca design methodology. We explain a series of modeling, implementation, and benchmarking issues that we encountered, so that others can understand and compare our solution with theirs.
