The W3 Model of Information

An information system must have an abstract model of what information is. The reader uses this model when exploring the information space, so it must be a comprehensible model.

The W3 information space is conceptually simple, but is sufficiently flexible to be able to incorporate existing information systems as subsets of the space. The reverse is not generally possible, because W3 has probably the most flexible model of current systems.

Document objects

The information universe is compsed of objects which we call "documents" to match user interface parlance. In hypertext terms they are nodes. The format and type of data contained within an object is not constrained. In practice most objects are text files or hypertext files, but any numeric, graphical or abstract type could equally well be a document.

Links

The hypertext model allows documents to be linked. To be precise, an "anchor" may be a whole or part of a document, and a directional "link" joins two anchors. In principle one anchor may have several links, but in practice for simplicity we do not exploit this.

Links are in principle typed with information conveying the semantics of their existence, though again we have not used this feature yet. The topology of the "web" of links and documents is unconstrained, although we can envisage applying constraints to links with particular types.

User operations

To navigate web, the use has two basic operations. One is to follow a link. This is typically done by clicking with a mouse on the anchor in one document, causing the document of the destination anchor to be presented, and the destination anchor highlighted.

Links in practice include handmade links which the authors of documents have included so as to convey an association of ideas to the reader, and links generated automatically by programs which produce a hypertext view of some multiply connected system of data.

Indexes

There is a common case, though, in which following of links is not sufficiently powerful. That is in the case of an index. It is not practical to present, for example, the entire telephone book to a user, and expect him or her to browse through it. Instead, he or she would prefer to use the many powerful techniques of information retrieval to find the required entry.

Therefore, some documents are flagged as "indexes". This is a broad term, indicating that behind them lies some search which the user can request. The visible view of the index (its "cover page") specifies what sort of a search will be done, and what paparmeter it requires. In the case of an index document, the user is allowed to specify a query. This is the second operation of navigation.

The web does not impose any constraints on the variety of serach engine, and the protocols are transparent to the communication between the user and the search engine. In practice we find that real users tend to go for very simple one or two word queries.

Search results

When a user queries a search engine, the result is a hypertext document which points to other anchors which match the search criteria. Two points are important here.

The first is that this document is just like any other hypertext document from te point of viewof the user interface. This means that no extra user inferface metaphor is required, and the same interaction pattern may be used for browsing it as would be used for any other document.

The second is that the search result, like any other document, has a name. This is composed of the name of the index plus the search criteria. In a more sophisticated version it would be a functional expression in some genral functional language for the algorithm which generated it.

The user may therefore refer to the search result just like a document, and make links to it. Evry time the link is followed (subject to caching), the search is reevaluated.