Abstract
Due to a constantly changing environment, a poor understanding of the user's needs and preferences, as well as a lack of willingness to modify existing organizational structures and decision models, the full economic potential of Web Information Systems (WIS) has not been realized by now. The data object types of the presented meta model, the Extended World Wide Web Design Technique (eW3DT), provide hypertext designers with a conceptual, user-centric framework and graphical notation for the construction of both, reference and customized models, during the software development process of commercial WIS. A reference model as a normative concept represents an abstraction of a typical company, its functional units, or its (Web) Information Systems and is intended to streamline the design and implementation of complex applications at reduced costs. As precondition for pursuing a partial globalization strategy, eW3DT distinguishes between technical and content-specific responsibilities for designing, implementing, and maintaining WIS.
Introduction
Reference models are characterized by a high level of abstraction and provide designers of customized, company-specific models with a referential solution which merely has to be adapted according to individual corporate strategies and technical restrictions. In the case of WIS, they have to integrate a conceptual data and navigational model and - by choosing a system-specific optimal level of abstraction - should equally be applicable for structured as well as unstructured information. As one of the primary means of standardized communication regarding WIS architectures these models support cooperative efforts to design such systems. In order to serve as an efficient interface to people with very heterogeneous knowledge and expectations, diagrams have to include the essential information in an illustrative, clearly arranged, and comprehensible way.
Figure 1. WIS Analysis & Design
Results of research activities formulated in recipient-oriented reference models may be used by any organization to build up new WIS from scratch or to extend and adapt existing ones. The reference model together with the deduced conclusions represent a basic guideline for deducting customized models, streamlining the implementation of complex WIS at reduced costs (see Figure 1).
By avoiding structural inconsistencies and mistakes companies are able to improve the quality of their systems, usually at low costs in comparison with acquiring the necessary know-how from external commercial organizations. However, empirical validation of hypotheses concerning success factors of WIS in different lines of business remains vital since standard software metrics frequently fail to evaluate the structure of WIS adequately. Furthermore, no valid theory currently exists to "assess formally whether a conceptual model of a hypertext is clear to readers. The only feasible alternative is testing the hypertext in as real a situation as possible" [20].
Extended World Wide Web Design Technique (eW3DT)
In this section methods and tools for WIS analysis and design are described and compared. Due to the limitations found in many existing design concepts, Bichler and Nusser developed the World Wide Web Design Technique (W3DT) together with a working prototype called WebDesigner [HREF1] which supports the graphical, interactive design of complex WIS from a user's perspective [5]. Comparable academic or commercial modeling tools like the WebArchitect [HREF2], SchemaText [HREF3], or Microsoft's FrontPage [HREF4] to name just a few provide similar "authoring-in-the-large" [7] functionalities. However, most available products do not address responsibilities for content production, system implementation, or regular maintenance.
Other approaches like the widely accepted RMM methodology [15] or OOHDM [23;24] are especially suited for highly structured information domains [2]. In most cases they succeed traditional hypermedia design techniques like HDM [8;9] which themselves are under constant development. In contrast to these database-oriented concepts, W3DT was built from scratch to support the requirements of unstructured, hierarchical WIS and to visualize them from a recipient's perspective. Utilizing practical experiences in developing WIS, the graphical notation of the design tool was further refined and used to analyze a number of deployed applications. With special regard to reference modeling of commercial WIS, the Extended World Wide Web Design Technique (eW3DT) was developed [22].
The concept of hierarchy is essential for the architecture of any real-world WIS [25]. On the other hand - as far as database-centric hypermedia applications are concerned - there is no adequate substitute for entity-relationship or object-oriented approaches. In this sense both W3DT and eW3DT are not intended to replace existing hypertext design methodologies but to act as hierarchically oriented, complementary communication tools between researchers, system analysts, and the management responsible for the decision to implement WIS. For developing increasingly complex applications, therefore, it is recommended and necessary to integrate the user-centric eW3DT with technical, more database-centric meta models like RMDM, HDM, OOHDM, or any of the other approaches based on entity relations. This integration has to be carefully planned in accordance with existing corporate design and implementation guidelines and would ideally be supported by an integrated WIS analysis and design tool supporting each of the three model layers of Figure 1.
Developing Applications with eW3DT
The Dexter Hypertext Reference Model divides the structure of hypertext systems into three different layers: the run-time layer for accessing individual documents, the content-specific within-component layer, and the storage layer [11]. In the case of WIS, every reference model has to integrate a conceptual data and navigational model [14]. It should satisfy the need for modeling structured as well as unstructured information by choosing a system-specific optimal level of abstraction. Accordingly, the data object types of the presented, document-oriented meta model eW3DT focus on the storage layer of the Dexter model, describing the structure of hypertext systems as a finite set of hierarchically arranged and associatively linked elements.
The composition of the document itself, described by the within-component layer, is functionally dependent, outside of the hypertext model per se and therefore not part of the reference model. Since the simple run-time layer of the Dexter model, which relies on the fundamental concept of the instantiation of a component, is already predetermined from the perspective of the content provider, eW3DT merely has to provide an appropriate set of data and navigational components.
Besides its content, every hypertext document includes a logical and a layout structure [16]. The WIS architecture specifies a number of processes which are responsible for presenting documents in conformity with the specifications of other layers. The relative importance of all these components varies depending on the specific data object type and its instantiation.
By combining inductive and deductive methods the process of creating conceptual WIS representations is both top-down and bottom-up. As depicted in Figure 1, eW3DT provides a framework for the construction of both abstract reference and customized, company-specific models during the software development process of commercial WIS (Figure 2). Note the difference regarding the level of abstraction between meta models (eW3DT, RMDM, H