A systematic approach to designing a WWW application

I n this sidebar, we demonstrate the systematic design principles and methodologies advocated by many of this special section’s authors for a World-Wide Web (WWW) application. We used these approaches to reimplement ACM SIGLINK’s LINKBase, containing abstracts from conferences and other hypermedia-related “events.” We followed the seven steps of the Relationship Management (RM) methodology [3]. In the first step, entity-relationship design, we identified five entities (Organization, Event, Event_item, Person, and Publication) and eight relationships among them. The second step, entity design, involved slicing an entity’s attributes into different (overlapping) subsets, with a head slice connected to the others by structural links [2]. While the RM methodology prescribes guidelines for slicing a given entity, it does not discuss combining slices from different entities for presentation. We added the concept of a cross-entity slice combining elements from different entities to be presented in a single window. For example, to display details about a conference event adequately, we included attributes from the Organization, Person, and Publication entities. We also added the concept of a minimum slice—the minimal set of an entity’s attributes to include (as anchors) in slices from other entities. This concept is similar to identifying objects and groups of objects suggested by [1]. These anchors are the starting points of applicative or reference links, while anchors belonging to the same entity (for example, description in Figure 1) are the starting points of structural links [2]. Slicing helped us determine the appropriate chunks of information to be displayed without fragmentation, thus increasing local coherence [5]. The third step, navigation design, produced a Relationship Management Data (RMD) diagram featuring major entry points, indexes, guided tours, and other access structures. This step forced us to plan access structures carefully enabling us to