...
Another option is to model one 'knowledge entity' which may have lots of constituent components as one single object in the repository ('compound model'). Each component is represented by a datastream. Each associated metadata record again becomes a datastream. A good example for this approach is the ARROW project (http://arrow.edu.au
).
The last approach is an 'atomistic model' in which all of the constituent compounds of a knowledge entity become Fedora objects, tied together by an aggregation object which acts as an anchor for all constituent parts. Thus, each knowledge entity is represented by a network of objects or a graph. The National Science Digital Library (NSDL, http://www.nsdl.org) follows this design principle.
...