February 1995
The problem with an ever-expanding meta-document is the semantics of the relationships between document nodes. Technically, it is easy to make, and continue to make, links, but at some point this becomes unmanageable in a practical sense.
The problem, it seems to me, is that of context.
Links are created within a specific context and are valid only within that context. Thus, although a document may have an ever expanding number of links, the user should only see those valid for their current context. Otherwise, we end up with the ridiculous situation of everything being linked to everything else.
We have done some research on this and have come up with the simple ER model as follows:
is_linked_to
------->
doc context
component
------->
is_linked_from
The two relationships are needed because a document is both linked to and linked from one or many contexts (the normal many-to-many relationship in database terms). This model enables the user to specify the reason for creating a link (ie. associated with the content of the link), as well as determining who is linked to the current document, a function not supported in SGML/HTML systems that I have seen and yet shown, by our research, as being very important in managing the lost-in-space syndrome. Interestingly enough, the actual document from which the component is derived, is itself an instance of a context in which the component is used.
This simple model also then enables the reader to restructure a set of document components to form new 'virtual' documents, again shown as being important by our research. The problem of object re-use then becomes important, especially in the light of anaphoric references etc.
bobj
morphing editors