In trying to provide information on the web ourselves, we encountered some limits on the way that information can be organised. For example even slight variations in content can only be implemented (in HTML) by duplicating the documents and then changing the information in one of those documents.
This approach has two main disadvantages; one, lots of information is duplicated, two, new organizational structures have to be built around the documents to replicate this variation, i.e. documents have to be organised in different file trees or given complicated extensions.
Sometimes it would be much easier to put all the variations into one document and let the parser decide which of these is to be presented to the user. We want a document that changes its content to fit the context it is viewed under.
Another problem is, that the hierarchical structure that is created so easily with HTML is not well suited for random browsing (an overused mode of access in the web). But sometimes one has no choice but to browse randomly, when one encouters a web space for the first time for example. In these cases it would be better for the user to be provided a linear path through the structures, a path introducing the key features of what is being offered.
Sumarizing, what we need is a way to put more information into a document than is shown at one time and of course a way to access it.