Title:        Namespace Registration Issues
Author:       Stu Weibel
Date:         2004-03-12


I. Introduction

Registration and maintenance of terminology namespaces is
a key requirement for effective deployment of vocabularies
on the Web, for metadata in particular and for Semantic Web
activities in general.

This requirement has been recognized within the DCMI community
for a number of years, and this recognition has motivated
the development of a namespace registration facility within
our community, to the point of the development of actual
software and procedures to support namespace registration.
We have not deployed this infrastructure to date

II. Impediments to namespace registration:

The fact that DCMI has not deployed this infrastructure is
the result of the following factors:

1. Launching, testing, and operating this infrastructure
has required a significant amount of focussed attention on
the part of DCMI administrators, developers, and Usage Board
members to assure that the technology, the policy, and the
administrative apparatus is in place and working as intended.
The coordinated effort necessary to achieve this has not
been available, though significant effort has been expended
at various times toward meeting each of the three facets.

2. Partly as a result of the efforts that *have* been made,
it has become apparent that the problem is more difficult
than was first imagined.  I think it is significant that
DCMI is perhaps the first Web initiative that has tried to
cope with these issues in a systematic way.  That we have
backed away from full deployment on a couple of occasions
reflects an appropriate degree of conservatism and caution (as
well as the vagaries of coordinating a distributed effort of
great complexity).  Once begun, this task cannot be abondoned
without significant penalty for users who adopt it.

3. It has become evident that DCMI's problem is a subset of
a larger problem of the incompleteness of the "architecture
of identity" in Web architecture.  My belief is that it is
a risky and ill advised for DCMI to deploy a local solution
to a more global problem.  Still, the need is there, and the
question remains as to how to meet it.

Possible solutions:

1. Deploy the namespace registration facility that DCMI
has developed.

This is risky, but we're pioneers, and we've taken greater
risks.  The more important factor that mitigates against
it is that we cannot resource the activity with confidence,
and there is no imminent likelihood that this will improve.
If I am wrong about this, at this point in the discussion
someone will be stepping forward with assured resources and
committment to manage the task.

2. Adopt the "info" URI registration mechanism for namespace
registration

This path has the virtue of associating our problem with
others who share it, and hence with greater likelihood that a
broader community and institutional committment will emerge
to support it.  The prototype is in place and available,
and a number of namespaces have been registered in it to date.

Still, prototypes are not production systems, and success is
not assured.  There has been substantial pushback from within
the URI community on the prospects for "info" as an endorsed
part of the URI architecture, and my own best guess at this
time is that it will not become an IETF standard.  It is,
however, part of the OpenURL standard, and enthusiasm from
a number of other potential adopters may render the lack of
IETF adoption less important.

It is unclear as to whether "info" functionality is as closely
tailored to the functional requirements for the management
of terminology as we would like it.  It is probably as good
as we would achieve in a first round of our own registration
protocol, but we won't have as much control over change.

3. Develop (or participate in the development of)
terminology-specific identity schemes that help to fill in the
missing components of the Web "Identity Architecture".  This is
a substantial challenge and will require a lot of IETF and
W3C politicing. It is a middle term, not a short term, path.
I expect to be involved in some such effort, possibly as part
of a grant proposal, irrespective of the course DCMI takes.

4. Do nothing beyond encouraging namespace managers to declare
and maintain namespaces for their terminologies in a way that
is likely to be consistent.  This is actually a respectable
position -- low risk, and a suitable policy white paper on
the subject might have significant effect.

Hybrids of the above solutions are also possible.