Using document (or other resource) signatures rather than IP addresses to identify resources on a network. Decoupling content identity from storage location. If the same resource is available from multiple locations, and has a network address that may resolve correctly any of those locations, then there are multiple (non-unique) addresses. Imagine a document named fooDocument with numerous equally valid copies scattered around the internet. If there were a URL that could be used to find any valid copy - find://fooDocument - then that URL would be a non-unique address. Resource discovery is a related topic. See directed vs. undirected routing.