At the meeting at ALA in January 2002 there was discussion of the identifier
element:

"Decision: Identifier is required. A globally unique identifier is
preferred (with scheme identified), but in the absence of that, a
constructed or derived identifier of a local nature can be supplied,
including a specification of the institution or local organization
supplying it.
        Reason: It is desirable to have some required elements, and there
is always a need to identify the resource in some way."  [from Rebecca's
notes]

This would make both Title and Identifier mandatory in *all* DC
Lib-compliant records.

A far better approach, in my opinion, is to follow standard library practice
and make standard identifiers mandatory-if-applicable, and local identifiers
optional-if-desired.

Let me categorize identifiers into 4 classes for the sake of discussion:

1. Standard identifiers: Typically these are identifiers compliant with a
published scheme and/or registered with a maintenance agency, DNS server,
etc. (e.g., ISBNs, URLs, etc.) -- and can often be "resolved" in varied
settings to a copy of the resource.  Standard identifiers routinely have
widely-recognized value as important elements for identification and access
across agencies and domains.

2. Unique-resource identifiers: Accession numbers and similar "local"
numbers assigned to one-of-a-kind resources (such as items in a museum
collection) for all intent and purposes take on many of the characteristics
of standard identifiers, often appearing in catalogs, citations, photo
captions, etc. produced by third parties.

3. Local library agency identifiers: Typically these are accession numbers,
bar code numbers, call numbers assigned by a library to aid
browsing/retrieval by local users and support local administrative tasks.
Given that libraries primarily collect published materials, local library
identifiers are rarely meaningful for identification or retrieval purposes
beyond a local context -- in the absence of a standard identifier, users
outside the local context will generally rely on a combination of
non-identifier elements in the record (e.g., title, agent, date) to
establish identity, and most outside users will probably pursue retrieval
through channels that will find no utility in the local library identifier.

4. Generated identifiers: For third party agencies that create or publish
metadata to DC Lib specs for items they don't own, it will be necessary to
provide an identifier (even if it's arbitraty and meaningless) if identifier
is mandatory under DC Lib.

Observations/opinion:
- Standard identifiers (#1) are routinely mandatory-if-applicable in library
cataloging practice. The other classes (#2-4) are not to my knowledge. It
strikes me that on principle we would not want DC Lib's requirements to
exceed the requirements of standard library practice.

- Unique-resource identifiers (#2) are also likely to have value and
probably will be routinely, voluntarily made part of DC Lib records if they
apply.

- Local library identifiers have no appreciable value (or significant
contribution to identification of the resource) outside the library's
context, and in many libraries some identifiers such as call number may not
even be unique within the library. It seems best to leave it to the judgment
of the library whether to include a local identifier in a DC Lib record.
Additionally, it should be noted that if DC Lib records include local
identifiers routinely, there is some risk that these identifiers will become
index "noise" if DC Lib records are collected from multiple sources,
converted to DC simple and added to a common database (a realistic scenario
given OAI harvesting).
--- Local identifier scheme discussion: As discussed at ALA, conceivably the
local identifier could have some measure of "global" identification value if
one could encode the local agency identifier in a manner that makes it
unique and traceable to the local agency, perhaps in the form of
[name]-[identifier]. While some existing name/code registries (e.g., OCLC
Symbol, MARC Organization codes) might reasonably be adopted to underlie
such a scheme, no single registry has universal reach.  Making such a scheme
genuinely workable on a global scale seems unrealistic to me. Yet without
the use of such a scheme local library identifiers have minimal value for
identification outside of the narrow confines of the library. This argues
for making local identifiers optional in my opinion.

- Generating identifiers: Generating strings for the sake populating an
identifier element seems counterproductive and unnecessarily burdensome for
all parties, yet this is exactly what would be required in some contexts
(e.g., third parties creating/publishing metadata for resources they don't
own) if DC Lib makes identifier mandatory under all circumstances.

Recommendations:

So I would favor something along the lines of the following as practice for
the DC Lib identifier element:

Mandatory-if-applicable:
-Standard identifier: Provide least one standard identifier from a standard
scheme (e.g., URL, ISBN, etc.) if one or more standard identifiers have been
assigned to the resource and are known to the metadata agency.

Recommended:
-Unique-resource identifier: Provide applicable identifiers assigned to
one-of-a-kind resources (such as accession numbers assigned to items in a
museum collection) if one or more of this class of identifiers have been
assigned to the resource and are known to the metadata agency.
Best practice: If the structure of the identifier (e.g., all numerals) is
potentially ambiguous it is recommended that the associated agencies' name
be included as part of the identifier element.
-Citation: Provide a citation if no standard identifier is assigned and a
formal citation is a common means of identify the resource being described
(e.g., a journal article).

Optional:
-Agency-specific identifiers
-Other identifiers

Added thought:
Further I would add that DC Lib adopt the classes of standard identifier and
citation (even if abridged to a title, or name-title) for Source, Relation,
and any other elements that require an identifier value.

Regards,

Eric

Eric Childress
Consulting Product Manager
OCLC Metadata Services Division
OCLC Online Computer Library Center, Inc.
6565 Frantz Rd., Dublin, OH 43017 USA
US: (800) 848-5878 or (614) 764-6000
Fax: (614) 718-7361 email: [log in to unmask]