URNs are useful for the same reason that ISBNs, ISSNs, etc. are useful. They weren't intended to replace URLs; they were intended to cover a case for which DNS-based and protocol-based URLs have repeatedly been demonstrated to work poorly -- specifically, as identifiers which have a high probability of being stably bound to the same resource over very long periods of time (decades or more). The assumption is that for some purposes a stable association between the resource name and the resource is more important than being able to access that resource using that name from anywhere at any time.
--Keith Moore on the xml-dev mailing list, Wed, 17 Apr 2002 16
(CafeConLeche?) - Quotes about XML in 2002