I'm a little worried about repository identifiers, when it seems to me we don't really know what a repository is. Maybe you can finesse this; X is a repository if Y who owns/controls X says it is a repository. JISC had the working definition: "a repository is a managed store of content that enables sharing of that content, so the key words are ‘managed’, ’sharing’ and ‘content’. If a filestore (etc) does this adequately for the intended users then fine". On that basis, for example, an eJournal collection managed by OJS is a repository. This is fine, but unambiguous identifiers for ambiguous entities somehow represent a problem for me!

I suspect that your worry is indicative of something more fundamental, applying as much to 'repositories and preservation'. Perhaps this is the place to discuss it as identification requires us to know what we are identifying, as you state. I detect a lot of self-obsession with repositories. I confess to having shared at least five minutes of that obsession (in a managed, content-focused way of course) when (a) working out the essential features of Jorum, the Depot and ShareGeo, in contrast to Digimap, NewsFilmOnline or even Suncat (say) - more about which, as you know, is at edina.ac.uk . These all manage a store of content had allows others to use that content. The terms and conditions of licence apply for onward sharing, but that surely is orthogonal to the definition of repository. IMHO, a repository is a repository if and only if it supports a minimum of two use communities, by running two (and maybe three) basic services: an ingest service for a use community of depositors of content; an access service for a use community of 'users' of that content. The third service is some sort of keep-safe service, which minimally warrants the relationship between what is deposited and what can be extracted for use. Of course any such repository is likely run (owned/controlled) by an organisation, and it is through that organisation that any given repository gains other qualities. And of course, it is possible for that responsibility for a given repository (if named, scoped and identified) could pass from one organisation to another. As analogy, a journal is a managed store of content, the responsibility for which can pass from one organisation (typically a learned society, university or commercial publisher).

But to return to what I think is the larger confusion that we should try to avoid, with identifiers and with other matters such as preservation: that is, there is life outside the repository, and even of the virtual walls of repository space. Typically, although by no means exclusively, I am referring to the world of scholarly communication & publication. Here I think we have a choice. Do we (who live inside the university and research world) attempt a universal scheme for defining / naming / identifying objects being manufactured and exchanged within the university and research world or (maybe and/or) base that upon the existing domain-specific identifiers - including the Internet-related adaptations? Thinking also of audio-visual material, and some of the industry surrounding that, they too have domain-specific schemes for defining / naming / identifying objects.

My view remains constant:
1. any universal centrally-registered unique ID scheme is bound to fail under its own weight
2. therefore we shall have a pleethora of schemes for each object
3. the problem is to identify each object uniquely and unambiguously (and thus allowing disambiguation)
4. the only way to do this is based on values of attributes of that object (basic relational theory)
5. this means choosing appropriate attributes to give maximum chance of uniqueness (like differentiating species and individuals in biology)
6. and producing apporpriate software to disambiguate then give the object a unique id for purpose P and time T (both purposes and time change and may affect the relative weight given to disambiguating attribute values)
Keith

@Keith, perhaps the RDF lesson, given the truth of your proposition 1, is that everything is declared relative to some namespace. That's how the Shibboleth stuff works as well (in a different way). My CV asserts that various CRs who could be (or maybe once were) identifiable at various institutions, as past student, past employee, current employee, current member etc. Unfortunately much of that information & assertions are private. But my identity as an author in D-Lib Magazine, Ariadne, IJDC, ERA and whatever the Glasgow repository is called these days, is public, and we ought to be able to build on that. I'm guessing that's what item 5 in the plan above is saying!

New report on identifiers from the RLG/OCLC: http://www.oclc.org/programs/news/2009-05-01.htm
From the press release:
This report identifies the necessary components of a "Cooperative Identities Hub" that would address the problem space in the research community and have the most impact across different target audiences.

The fifteen members of the RLG Partnership Networking Names Advisory Group developed fourteen use case scenarios around academic libraries and scholars, archivists and archival users, and institutional repositories that provide the context in which different communities would benefit from aggregating information about persons and organizations, corporate and government bodies, and families, and making it available on a network level.

The report summarizes the group's recommendations on the functions and attributes needed to support the use case scenarios.