Message-Id: <9306071928.AA07332@wilma.cs.utk.edu>
From: Keith Moore <moore@cs.utk.edu>
To: uri@bunyip.com
Subject: dereferencing URNs
Date: Mon, 07 Jun 1993 15:28:24 -0400
I have always assumed that there would be a service that maps URNs to URLs
and other information, and that publishers that hand out URNs would be
required to make sure those URNs are registered with that service.
Such a service might work as follows:
1) take the 'naming authority' (urn prefix) and append to it some fixed
string. For instance a URN "URN:foo::xxx" has naming authority "foo",
to which one would append ".urn.org", giving "foo.urn.org"
2) do a DNS lookup on "foo.urn.org", which would return one or more
TXT records containing URL server information. Each TXT record
would contain (a) domain names of servers and (b) information about
constraints on the use of those servers (cost, who can use them, etc.)
3) Narrow down the list of servers based on the constraints. Then pick
one of the remaining ones at random and send the query to it (protocol
TBD).
Note that nothing prevents the naming authority from being hierarchically
structured, so it could be "foo.bar", rather than "foo" and thus point to a
different set of servers.
Part of what makes this work is that the 'naming authority' has some
responsibilities in addition to assigning names. For instance, it must be
willing to maintain an authoritative 'citation record' for each of the
objects that it gives a name to, and it must be willing to supply URLs for
those objects if they are still 'in print'. So a 'naming authority' has a
similar function to a 'publisher' for print media.
There may be other locations, and therefore other URLs, for an object,
available through other means: e.g. library card catalogs or an OCLC-type
database. But this scheme ensures that there is at least one path to finding
an object and information about that object if you have its URN.
Comments? Stones?
Keith