Some time ago we began a portfolio project called Metadata Switch. We have recently written up a high level overview of some of this activity in the catchily titled Metadata switch: thinking about some metadata management and knowledge organization in the changing research and learning landscape [pdf].
This was a response to a rapidly changing environment of metadata creation and use. Libraries are potentially creating metadata in several different editing environments (a library management system, a digital asset management system, ...), in different formats (MARC, DC, VRA, ...), with different controlled vocabularies, and within different data input regimes (AACR2, ...). There is also a growing interest in pulling metadata together from different sources, whether by harvesting or federating search.
What we wondered was what types of services would be useful in this environment. We have been working on metadata harvesting (see the list of activities involving OAI), on schema transformation, on terminology and name authority services, and on other related issues. These individual projects have now developed into signficant activities in their own right.
We have been doing much of this work in a web services environment, creating modular, 'pluggable' services which can be invoked at the point of need.