It is useful to express the location of a node in a semantic graph in terms of a sequence of knowledge-bearing arcs that lead to the node from a node that is used as a point of reference. If the arcs on which such "graph-based addressing expressions" depend disappear, the expressions become invalid. When an arc is reified as a node, the reified arc may or may not be removed. If it is not removed, the semantic graph may exhibit unnecessary complexity, ambiguity, and lack of parsimony. Alternatively, if the original arc is removed, the value of dependent graph-based addressing expressions may be lost. For systems intended to support collaborative knowledge aggregation, such as the Semantic Web, one way to resolve this dilemma is to disallow "lazy" reification, and, in effect, preemptively reify everything that may, at any future time, require reification. Preemptively reified nodes can be "virtual" until they are actually needed, thus effectively maintaining parsimony. The Reference Model of the Topic Maps paradigm, now under development in the ISO, shows how the need for "lazy" reification can be avoided by means of a fixed set of arc types and rules used to represent fully elaborated assertions.
