The Abstract Component Level

The representations within the abstract component level present the components that will be associated with one-another to form the hypertext. Within the context of this level, the representations are independent of each other - such associations will be made at lower levels of abstraction. The structure, perhaps a directed graph, is separated from the collection of contents that are to be displayed to the reader and the collection of "buttons" that will be selected by the reader when moving from location to location in the hypertext. Additionally, it may be the case that the view of the hypertext presented to the reader combines together independent content elements into an integrated whole. The presence (or absence) of such composition is also represented abstractly at this level.

The structure of the hypertext can be chosen arbitrarily. The Trellis system, by Stotts and Furuta, uses a Petri net structure, which provides automaton semantics as well as the network representation. Other graph-based structures are appropriate as well - for example deterministic finite automata or directed graphs, trees or lattices. The structure of the hypertext need not be limited to networks. Constraint-based descriptions may be used as well. Also, there is no requirement that, in the graph-based approach, the elements of the structure be fully connected. The structure merely provides "placeholders" that will be associated with the hypertext's content, and it describes relationships that exist among these placeholders.

The abstract content is of arbitrary form. It may include textual, graphical, animated, audio, video or other material. The content may be specified directly or may be the result of a computation. While it does not contain links, it may incorporate markers that define a collection of potential locations for the mappings of links and their presentations that occur in lower levels of the r-model. The content may be described in a form that is independent of the eventual characteristics of its display (SGML is such a generic description language), or it may be described in a form that is highly dependent on the eventual display. However, the flexibility of the mapping from content to structure in the next level suggests that a display-independent representation may be most appropriate.

The abstract buttons are abstractions of the ways in which the relationships among content elements can be displayed. Abstract buttons may themselves have content and an associated type. The content is provided to specify what will be shown when the button is displayed. The type is needed to specify how the button will be displayed, and may describe other characteristics of its behavior on display and selection. As with the content of the abstract content, the content of the abstract button may be statically defined or computed.

The abstract containers are an abstraction of how the pieces of the hypertext will be combined when shown to the reader (how the pieces will be aggregated and combined for display), and not of what is in the hypertext. For example, if several content elements are displayable, one possible presentation would be to show each element separately while another would be to combine the separate elements into a composite, which would be presented to the reader as a unit. In the first case, one could say that a separate container had been associated with each separate content element, while in the second case, one container would hold all content elements.