Let me speak in favor of draft-spinosa-urn-lex.
My understanding of the intention of the draft is essentially the
creation of a universal system for bibliographic citations of legal
works, using the syntax of URNs to express the citations. This
understanding colors how I think about the namespace proposal.
The intention seems to be that every document is assigned one or more
URNs by a governing jurisdiction, or by a publisher of a version of the
document. Note a point that I did not understand until recently, that
the uniqueness rule for URNs permits multiple URNs to refer to one same
resource. As the authors note, one use of this is providing references
to a document which are based on multiple natural languages.
The resolution process for assigned URNs is a typical indirection
system, where the URN is looked up in a database. However, the system
supports a usage that is much more like other library or bibliographic
systems -- the user can compose the information that he knows about the
document into an "approximate" URN which can be searched for in a
database using a "closest match" rule, which both converts the queried
URN into an assigned URN and provides information about the document
itself. In this usage, the user-composed URNs are fundamentally
equivalent to queries.
The current draft does not clearly separate the information about the
query/resolution process. It would be helpful if the resolution
information were in a separate section, as the registration template
format in RFC 8141 recommends. It would also be helpful if there was a
clear distinction between the parts of the resolution process that are
well-defined and fixed, and (IIRC) some information which has not been
firmly fixed.
However, the messy part is the user-created "query" URNs. To get these
within the formal rules for URN namespaces, we need to recognize them as
a special subclass whose definition does not come from assignment to a
document by an authority, but rather from the namespace definition
itself, in a clause somewhat like:
A URN that is not assigned to a document is defined to represent a
query into the universe of URNs that have been assigned to
documents, whose resolution is current set of assigned URNs that
most closely match the URN.
Clearly, there are some rough edges to that. But the meaning of a URN
is time-invariant as a query, despite that the results of the query
depend on the time. And when a URN is assigned to a document, the URN
is equivalently a query that returns just that one document (since it is
a perfect match to itself as an assigned URN).
Some of this reasoning is a stretch, but (a) I've written URN definitions
that aren't perfectly compatible with RFC 8141, and (b) this pattern of
definition and usage seems to be ideal for handling bibliographic
references.
So I think the overall concept is good and useful, though the
presentation needs to be cleaned up in the ways that we've been
discussing.
Dale
_______________________________________________
Gen-art mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/gen-art
