PAMELA: A Proto-pattern for Rapidly Delivered, Runtime Extensible Systems

The paper describes a proto-pattern for a system architecture which facilitates very rapid development and extension of an application. The overall pattern has three major components: Overall Architecture, Process Patterns for standard editing, viewing and navigation, and Meta Data patterns for knowledge storage. The architecture level pattern uses meta-data held explicitly in a persistent store and accessible at run time to customize process patterns implemented in an interpreted language, rendering them executable. The meta-data structure in turn uses patterns which facilitate extensible data structures and relationships. The use of the proto-pattern in the design of a commercial knowledge management product illustrates its potential. Benefits are highlighted and areas for further work indicated. 1. Problem Statement Business pressures and unclear or rapidly changing requirements require systems that can be developed very rapidly and extended with minimal delay. Conventional development techniques appear to be inadequate in addressing this problem. Object oriented techniques have long held much promise and have provided benefits, but many organizations are still frustrated by slow, expensive development which simply does not meet their business needs. [Taylor, 1995]. The advent of the Internet and e-Commerce has simply exacerbated the position. Applications are increasingly mission critical, having to operate 24 hours per day, 365 days a year in a global economy that does not pause. There is thus a need for very high reliability and availability in the systems produced. 2. Previous Approaches Some of the approaches already used to speed development are listed below, with a discussion of their effectiveness. High level languages assist by allowing developers to work at a higher level of abstraction. Figures from [Capers Jones, 1997] indicate that language level is a major determinant of developer productivity. Some researchers, e.g. Taylor [1995] have proposed the ideal of a language at a business domain level, where users can specify requirements in their own terms Use of class libraries has certainly helped the developer community to reuse code and save effort in the implementation of commonly required facilities. This has achieved much success in the areas of graphical user interfaces (GUIs), relational database and complex internal data structures (e.g. Collections). The success of Java in the marketplace is largely due to the provision of an increasingly rich environment, delivered in the form of Java classes. Class library proliferation to support business objects has been less successful, although the [OMG] and [IBM] have made significant attempts in this direction Components have been very successful at the developer level, in much the same way as class libraries. There are vast component libraries available for C++, Visual Basic, Delphi, Java and Smalltalk. Again, these primarily address technical aspects, such as GUI, collections, communications, database access and the like Application Generators and CASE Tools aim to take models and turn them rapidly into reliable executable code. While there have been some seminal successes, the general experience is that these