any kind of so-called information resource, should be "slash" URIs -- in other words, should not contain a fragment identifier, whereas a URI used to identify a concept or abstract resource should be a "hash" URI using a fragment identifier.

For example: http://www.example.org/catalogue/widgets.html would both identify and locate a web page (maybe providing some human-readable description of the widgets sold by Silly Widgets, Inc.) whereas http://www.example.org/ontology#Widget