- From: Larry Masinter <LMM@acm.org>
- Date: Fri, 12 Sep 2003 08:44:55 -0700
- To: uri@w3.org
Perhaps there's another way to think about the issues (for 'doi' and 'info') without name-calling. Their definitions are tied up with the service guarantee of a registration organization. The definition of "info" depends on the service guarantee of NISO, since NISO is maintaining the registry of schemes. The definition of "doi" depends on the service guarantee of IDF, since IDF is maintaining the registry of dois. (This isn't different from URNs, where the meaning of a URN and the promise of uniqueness depends on the service guarantee of the namespace holder.) Whenever there is a service guarantee and a namespace, the question still arises: What happens if something happens to the organization that is making the service guarantee? Organizations change. They lose interest in providing a service, decide there is no interest, lose funding, are acquired, merge, go out of business. Some organizations have more stability than others. With most assignment-registry namespaces in URIs, the issue then becomes: what would happen with the deployed names if there is an organizational change? With DNS based schemes (http, ftp, mailto), the answer may not be pretty but it is understandable: you get either a DNS failure (the host name lookup fails), or a protocol error: TCP error, 404 not found, mail bounce, etc. With the proposed "doi" and "info" the question is still open: what happens if the organizations that is the root of the meaning of the name fails? Certainly any URN namespace faces this as well, but at least it is a requirements of a URN namespace application that there be a credible assertion about permanence of the name assigments. It seems like it is a requirement of assigment-registry is some assertion about permanence. Part of the reason for the process of URI scheme naming, and asking for "demonstrated utility" (that it adds value over other existing URI schemes) was to avoid putting IANA into the business of being overloaded as the IETF's own assignment-registry service. I think there is a problem, though, in that we now have non-profit or for-profit organizations who may have operational plans based on getting the IETF to accept the URI schemes named "doi" or "handle" or "info". And the reasons for IETF not cooperating aren't clear. Some suggestions to consider: * Redefining 'urn' to be more specifically reserved for 'assignment-registry' services (whether for fee, for profit, non-profit, etc.), where 'permanence' is a requested attribute rather than a definitional quality. * Clarifying what is requested by 'demonstrated utility' in the URI registration process in a way that would dissuade assignment-registry services asking for additional URI names. * Encouraging (or even forcing) those who still insist on a separate scheme to use 'vnd-' or 'prs-'. * For 'doi' -- because the group pushing them has had some amount of marketing success -- allow the publication of the 'doi' draft as Informational, but with an IESG cover note that would discourage similar namespaces. This might resolve the difficulty without necessarily giving the IDF their desired imprimatur. * For 'info', ask NISO to instead assist its sub-namespace authorities to register URN schemes, or to use urn:us-niso: (to distinguish NISO from the National Irish Safety Organization and Norske Idrettsutøveres Sentralorganisasjon). Larry -- http://larry.masinter.net
Received on Friday, 12 September 2003 11:47:03 UTC