Domain Concept

An approach to integrate these two aspects can be based on the concept of domains. Its purpose is to provide a general means for grouping and structuring HTML pages. A domain is defined as a set of related HTML documents which together form one document. It is possible for a single HTML page to be part of several domains and it is possible that a domain is itself an element of another domain.

The following information which will be necessary for WebMap to support domains should not be part of the document pages' bodies but could be made available as meta-information stored with a ``starting'' document page (domain entry page):

1. The definition of the domain, i.e., the set of pages.

2. A set of structure types available for this domain, e.g., linear list, circle, array, hierarchy.
3. For each type in : the structure information defining the relationships between the pages ( topology of this domain).

4. A set of domain traversal strategies a user should be able to choose from to get navigation support for the domain (see section 4.2).

As a first step towards realization of this concept and to support what already exists, WebMap derives graphical document structure maps from special ``table of contents''/index pages which often are provided by authors or document conversion tools. WebMap parses such pages using a simple heuristic: it searches for the first nested list, interprets it as a hierarchy and creates a node for each anchor it finds.

As proposed in section 2.1, it would be nice if an author provided a graphical navigation map along with a large document. To protect authors from learning a complicated map description format, the logical consequence is to extend WebMap's capabilities towards a simple authoring tool. Then an author could interactively create a map reflecting the document structure. In this case WebMap would not only store this map, but also generate a set of HTML document frames derived from an author-provided HTML template and enriched with structure information.

The next step to implement the domain concept will be the support of the HTML <LINK> element together with the <REL> attributes. Using this element, simple disjoint domain topologies may be created (resp. generated by WebMap).

As the final goal it appears most promising to keep additional document structure information, like navigation maps, completely separate from the documents themselves. As already proposed, it could be stored as meta-information together with one ``domain entry page'' instead of spreading it over several documents.

This approach has several advantages:

• It leads to higher performance. Using only <LINK>-elements would cause WebMap to request every single page belonging to a domain from a HTTP-server to be able to provide the user with a navigation map for this domain.
• The maintenance of the structure is easier and inconsistencies are less probable.
• Multiple topologies may be defined for a domain. Due to its dynamic software structure, WebMap is capable to handle several topologies for the same domain (domain structures) simultaneously-analogous to the support of multiple layout strategies for a single topology.

• Structure information need not necessarily be defined by the author of the document. It could as well be generated by, e.g., an information server, or it could be based on document affinity (calculated content dependent document relationships, see the satellite tool [4]).

Of course, each page belonging to a domain should contain a <LINK> element in its HEAD that points to its domain entry page. This way, a user who reaches a page belonging to a new domain can immediately be provided with a complete map of his new environment.