On 17-Mar-05, at 11:29 AM, Mark Jordan wrote:
> However, I'm not totally clear on the range of potential uses for the
> registry we're all describing, even though several
> people have explained what they want for a particular application. For
> example, do people want to query a registry and get
> back attributes of a database product (vendor, titles indexed, etc.)?
> What form would this query take? Do people simply want
> to use it as a "name authority" for products to assist collection
> development, authorization management, etc.?

What I'm looking for is the latter: a control number that could be used
to identify when eresources are the "same"  (defined as pretty much the
same content, regardless of interface). The DLF model described by
Peter would be more than adequate for my purposes.

I need something like this in order to merge lists of eresources
maintained by different organizations. This would be relevant for two
of my current projects:
1. A research database gateway for students taking courses from
multiple institutions
2. A combined list of resources licensed by several different library
consortia

Actually it isn't the initial merging of these lists that's so much at
issue, as the ongoing maintenance of them as various sites add & drop
resources. It would be really nice to be able to automate that.

> How would a
> registry be maintained?

Good question, and it would probably depend on the scale of the project.

It would be easy enough to do with a small number of agencies, assuming
that the agencies bought into the need for it. Whenever one of the
agencies added a resource to their collection they would check the
central registry to see if an ID was already there. If it was, they
would add it to a field in their local record. If it wasn't already in
the list, they would log into the central registry and add an entry for
it.

This could scale up to a point, but this model probably isn't all that
great for more than a few agencies and a few thousand resources. I
don't know ... is it possible to do distributed creation of control
numbers?

To generate a usable ID of the format
interface:resource/collection1/collection2
one would need to unambiguously identify the elements so that others
could be sure that the control number actually applied to a particular
resource. Not entirely sure how to do that, but here are my thoughts so
far:

The interface could probably be identified by some portion of the
URL(s) used to access it, eg you would need to include
'search.epnet.com' as part of the record for EBSCOhost. For CSA
Illumina  you might need all of  http://www.csa.com/csaillumina, since
www.csa.com is the vendor's home page, not the interface.

To clearly identify the resource, would it be enough to include the
title and the publisher (which may or may not be the same as the
interface provider) ? Again, this might not scale up past a certain
point.

If collections were defined as components belonging to a particular
resource, then their name would probably be enough to disambiguate
them.

Does any of this sound like it would be useful to anyone besides me?


John

--
John Durno
Project Coordinator
BC Electronic Library Network
~~~~~~~~~~~~~~~~~~~~
Phone: 604-268-7002
Fax: 604-291-3023
Email:   [log in to unmask]
Web: http://www.eln.bc.ca