This deliverable discusses the management of the metadata associated with the
learning objects available to a SeLeNe network. This includes the following
tasks: description of learning objects in order to reflect the rich structure
and content of LOs; automatic generation of metadata when new LOs are
registered or when existing LOs are updated and finally support for search of
LOs according to the available metadata.

We assume that SeLeNe users can use operators to create new, "composite", LOs
from "atomic" LOs registered to the system. The integration of these operators
as part of the system's functionality has an impact on the definition of
metadata. Indeed, whereas the content of LOs falls beyond the scope of our
system, the structure of LOs can and must be reflected in the structure of the
metadata that describe them.

The main contribution of this deliverable is a semi-automatic mechanism to
infer the description of composite objects from the descriptions of their
parts. This mechanism can be summarized as follows.  First we assume that the
system relies on a common taxonomy of terms, together with a subsumption
relation.  The terminology is common to all members of the network.  Second,
we require users to "index" their atomic LOs with terms from the taxonomy.
Finally, whenever a composite LO is registered, the system is then in charge
of infering, bottom-up, the metadata for each node of the composition tree of
the new LO.