In reply to messages of Tom, Andy and Pete,

I agree with you on most points but I think that I need to be more
clear about the background of my proposal and that has to do with the
use of protocols like OAI and SRU. I expect that more and more often
users will access "unknown" services with unknown metadata schemas. Like
in TEL we can obtain collection/service descriptions from different
places containing a baseURL of the service and transform that into a
target for searching. A second expectation is that applications may
become more madatadriven (opposite to reqest driven) and that users may
want to add functionality based on metadata they receive but which were
previously unknown to the application.

Now suppose we access a system for which we got the base-URL in some
way. We do not know which metadata schemas are available and even when
we know we do not know what they mean. So we can only request DC. Now
suppose that a system has a schema (or application profile) called XYZ
which consists of DC and some extra elements. Even when our application
knows that the system can provide XYZ, our application will never ask
for it because we do not know what it means and we will never be aware
of the richness of the extra metadata that might be available in that
schema.

Now suppose that that system would say: we have DC, XYZ and DCX. DCX is
the same as XYZ, but we do not have to know that. We know that DCX means
that our application can understand the qualified DC terms from the
response and might ignore the rest. But when we (user, developer) see
that it contains interesting terms that might be useful we change our
application to use those extra terms.

So DCX is just an extra name for schemas that are qualified DC plus
something extra. Asking for DCX we get the extra metadata unsollicited.
The response may indicate that it is DCX or it may give the real schema
name (and we pray it actually was DCX :-). I do not know how this can be
achieved in another way. Instead of DCX we could also use another name
like DCAP but maybe we should reserve DCAP for real DCMI registered
application profiles.

Now the dcx namespace proposal. When we need to create new terms we
create a private namespace for it. However in some cases these terms
could very well be shared. On those cases it is a pity when one creates
his owbn namespace for it. Having a general namespace for this purpose
makes it possible to express the fact that these terms are not meant to
be kept private but may be shared and it avoids inflation of namespaces.
It also expresses the fact that these terms are complimentary to DC and
dcterms and other DCMI controlled namespaces. It remains encouraged
though to register these terms in a registration system that does not
have to be a DCMI controlled one. It is an efficient way to develop
applications and expose new terms. DCMI will not need to have control.

Theo


>>> [log in to unmask] 9/12/03 2:36:26 nm >>>
Theo said:

> Attached you find a proposal for a concept named as DCX. This
> proposal can also be found at
> http://krait.kb.nl/coop/tel/DCX_proposal.doc. There has
> already been some discussion on this subject before it was
> decided to send it to this list, soI copied and pasted the
> most recent messages below, assuming they also cover previous
> parts of the discussion.

As the comments from Tom and Andy note, I think there are two separate
aspects to this, which I think need to be separated out:

(1) a more generic abstract model for a DC metadata record that
includes
the use of non-DC-owned terms, and provides a label for instances of
this type of metadata;
(2) mechanisms for improving the disclosure of the semantics of
non-DC-owned terms

I agree with Tom and Andy's comments that the first of these seems to
correspond to what some people have called a "DC Application Profile",
though stopping short of providing a specific label for this entire
class. (I'm not sure how useful it is to have such a generic label.)

For (2), the suggested solution is for DCMI to support the use of a
"catch all" "dcx namespace" that anyone (?) can use as the basis for
the
URIs for any (?) of their (non-DC) terms.

I wonder whether this suggestion slightly mixes up two things, or at
least suggests a stronger relationship between then than is actually
the
case:

- the "ownership" of a URI; and
- the ability of someone other than the owner to discover information
about the meaning of a metadata term identified by that URI.

I'd argue that it makes no difference whether a term has a URI of
http://purl.org/dc/dcx/myterm or http://dublincore.org/dcx/myterm or
http://purl.org/tel/terms/myterm or
http://www.europeanlibrary.org/terms/myterm as regards my ability to
find out what the term means.

My ability to find information about the meaning of any term
defined/owned by someone else depends on people following shared
social/administrative conventions about disclosing that information,
whether that involves

- the owner/creator of the URI making that URI "dereferenceable" (for
the full "lifetime" of the term) and ensuring that dereferencing it
generates some useful information; and/or

- the owner/creator of the URI "depositing" information about the term
it denotes with a global or community-wide service (e.g. a registry)
that I know about and can access and use (It's no help to me if the
information is registered with a service I don't know about or can't
access)

(If you regard the Web itself as the registry, then maybe the first is
just a special case of the second.)

I just don't think that information about terms with URIs beginning
http://purl.org/dc/dcx is somehow _intrinsically_ more "discoverable"
than information about terms with URIs beginning
http://www.europeanlibrary.org/ or http://myproject.org/. It's the use
of shared procedures and conventions (including considerations about
dereferencing and persistence) that make that information
available/accessible to other people. DCMI has explicitly addressed
this
for the URIs it owns/creates in the namespace policy document (which I
think would be better titled a _URI_ policy document, but that's by
the
by!) and in DCMI Usage Board implementation of that policy in its
procedures.

To address (2), it seems to me we need

(a) improved understanding of the _responsibilities_ associated with
declaring metadata terms and assigning URIs to them (I don't deny
that,
as Andy said in one of his messages, there may be some diffficult
issues
in terms of persistence if such URIs are being created in the context
of
short-term projects);

(b) shared, consistent practices for the _disclosure_ of information
about those terms and URIs

I don't think either (a) or (b) requires the use of a catch-all "dcx
namespace", and indeed creating a "dcx namespace" for people to use as
the basis of their URIs wouldn't in itself address the problem: it's
the
conventions and policies that do that.

In terms of disclosure "services", one practical measure for DCMI to
consider might be whether the DCMI registry (or a DCMI-provided
registry
implementation) indexes data for terms from sources other than DCMI
(though I'd suggest there would need to be some control on what was
indexed.)

Pete