URIs are, prima facie, are a simple idea. But their simplicity and generality give rise to a whole range of possible uses that lead to endless debate about the true nature of URIs. From a protocol design perspective, URIs are often viewed as addresses -- labels that are used by a protocol to select network elements that participate in a transaction. From a document (or information representation) perspective, they are viewed as names that reference other documents or information. URIs serve both of these purposes, sometimes both at the same time, which gives rise to "philosopher's elephant" situations in which different parties see different facets of the same underlying ideas.
Comments to Graham Klyne. Document updated 3/16/01.