Great title! Especially interesting is the following section on the potential of a distributed XML Repository:

3.1.1 The XML Repository

We store pointers to published documents as XML records. By working with XML records, we gain the ability to change the information associated with their respective documents, which we cannot do with the proprietary databases. Indeed, the XML records should be seen more as dynamic objects rather than static documents. Not only do we gain the ability to change the original keywords and citation information from the respective documents, but also the ability to add annotations, links to other records, associations with other types of media (e.g. sound clips), etc. Furthermore, XML records can even have associated procedures to compute relevant algorithms. We can think of XML records as archival objects, “buckets” of pointers, links, data, and code, which are not affiliated with any one particular information resource, as defined by Nelson et al.

By transforming records from passive documents into active objects, we start our construction of the biologically motivated enabling substrate at the lowest level of information systems: the source data. This is an essential step to set up a distributed design. In centralized systems, documents can be passive since it will be up to a higher level program to decide if a certain document is relevant or not. In contrast, in distributed systems, much of the decision-making is off-loaded to lower-level components, which need to be endowed with computing capabilities. In this sense, records become active objects that store changing information, communicate with other components, and even perform actions (run code) on the information they store.

Here’s the paper:
Biologically Motivated Distributed Designs for Adaptive Knowledge Management

…and related material:
Active Recommendation Project