This session was moderated by Rachel Heery of UKOLN (http://www.ukoln.ac.uk).

From the outset it was decided that we were not looking for differing definitions of a registry, but rather the functions that it might support. Discussions centered around the following questions:

There needs to be a distinction between describing and prescribing terms, need some interaction between a registry and a dictionary. There is also the issue of prescribing use at differing levels e.g. DC, MARC, at the profile level

A registry should also provide some cataloging rules, to inform non-experts, these should be kept simple.

A registry should link services, with the aim of providing some form of quality control of the metadata.

It should be a tool for managing linguistic diversity within and between domains (in the same language)

A registry should disclose domain information such as which schemas are in use and what the element sets are. It should also clarify what the relationships between various schemas are.

Should there be a taxonomy of schemas -a structured view with links to purpose e.g. Z39.50 profile?

A registry should provide metadata for both display and search functions as well as catering for extensions to schemas.

The level of prescription will depend on the purpose of use, as will the richness of the links to other information. The level of granularity required will depend on the domain/s that the registry covers.

There are some parallels with a dictionary such as provision of a style guide, rules for usage, cross-references etc., Note there is a whole host of different dictionaries covering specialist domains such as architecture to specific audiences such as children.

Does "register" imply that the information given out is in some sense "authoritative"?

Discussions here were largely concerned with the issues related to a distributed versus a centralised registry.

It would be necessary to identify the community, the scope of the registry and the audience (financial systems, librarians, museums, industry etc).

It would also be necessary to somehow migrate legacy data.

Is a registry merely a data dictionary? If not, how does it differ? Need to consider syntax, semantic and lexical levels.

A registry should provide guidance on element usage for different target audiences.

The registry should provide links to people, projects and applications at differing levels of granularity. How distributed should the information be?

Should the registry provide mappings between schemas?

Should it list language codes, country codes etc?