Just came across this link from Brian Lamb's weblog...
Very interesting article, which basically boils down to something like "Technical interoperability is not the same as semantic interoperability" and goes on to discuss some interesting issues involving the granularity of a learning object, and the impact that has on its value and reusability.