Browsing through a hyperdocument often leads to orientation problems:" you don’t know where you are, you are lost in hyperspace”. Although some of these problems may be diminished by specific browsing tools, such as fish-eye views and history mechanisms, most problems are embedded in the structure of the hyperdocument. In order to detect and analyze potential problems in the use of a hyperdocument, we reduce different types of browsing actions to a unified and well know formalism: the Petri net.

In this paper we use the object-oriented model from [3] to describe hyperdocuments. This model, unlike other reference models [5, 6, 9], describes objects using a number of different description layers: these layers are not fixed by the model, and none of them are excluded from the model. In order to use Petri nets to describe browsing, the operations dealing with the hierarchical structure of objects (composites, towers and cities) are translated into operations that are equivalent to the simple follow-link. Unlike Stotts, Furuta and Ruiz [14] we combine the operations offered by the hypertext system (or presentation layer) with the link structure of the document, in order to characterize the browsing semantics of document and system as a unity. By using existing analysis tools for (different kinds of) Petri nets one can easily detect potential browsing problems like unreachable nodes and links, dead ends, unlimited numbers of simultaneously displayed nodes, infinite loops, etc. and analyze the consequences. Such analysis is most valuable to hypertext authors, whose biggest problem is the creation of the most appropriate links.