[OAI-implementers] Identifiers (catalog/entry)

Chris Hubick chrish@athabascau.ca
Thu, 18 Mar 2004 14:28:18 -0700

On Tue, 2004-03-16 at 06:32, Adam Cooper wrote:
> A number of IMS specs have played with this 2 part/1part problem. RDCEO 
> (Reusable Definition of Competences and Educational Objectives) talks about 
> URN scheme, and I think we had assumed that the catalog would map to the 
> NSS and the entry to the NID.

Hrm, that is what I have been doing up until now, but I am coming to
think I was wrong...

> Exactly what the significance of LOM catalog might be is probably another 
> question, and one that is intentionally open I think. Is it necessarily 
> more than an _indicator_ of the creator of the identifier?

I don't think it is open, the LOM spec says the catalog is "A namespace
scheme".  This would, at least to me, clearly indicate that the entry is
namespaced by it's catalog.  Any other interpretation would lead to much
greater problems, in that many people simply use increasing integer
numbers to identify their metadata records, and without those numbers
being namespaced by the catalog, we would have *many* collisions.

We basically have a 'three level' system.  It is up to all those people
sharing any particular LOM catalog to guarantee uniqueness within *that*
catalog.  For those (most) of us who share the 'URI' catalog, we have
the same uniqueness requirement, which we use the NID to satisfy.  This
pushes it down a level, where all those people sharing any particular
NID must also guarantee uniqueness within that NID.  A system like 'oai'
uses DNS names to do this.

IMHO, three levels is overly complex.  Yes, we could all run off and use
whatever format entries we like, namespaced by our catalog (which is
pretty much what people have done up until recently).  In an ideal
world, however, we would remove this extra level by all using the same
catalog, and partition within that ourselves.  The URI system gives us
that catalog, partitioned by NID.  All those using URI's have done the
extra work in agreeing to share a common syntax and associated
facilities for partioning the 'URI' namespace.  By not using a 'URI'
catalog, you basically mitigate that effort by making the fact you are
actually using URI format entries *opaque* to others from a LOM

I think LOM might have been better off to do as OAI has done and just
force everyone to uncode their id's as URI's, rather than having a
separate catalog field, but they didn't, so here we are.

Chris Hubick
phone:1-780-421-2533 (work)
phone:1-780-721-9932 (cell)

    This communication is intended for the use of the recipient to whom it
    is addressed, and may contain confidential, personal, and or privileged
    information. Please contact us immediately if you are not the intended
    recipient of this communication, and do not copy, distribute, or take
    action relying on it. Any communications received in error, or
    subsequent reply, should be deleted or destroyed.