Really good explanations recently on Grahame Grieve's weblog recently describing HL7 FHIR (Fast Healthcare Interoperability Resources) from the perspective of v2, CDA and v3 implementation.
FHIR for v3 and CDA implementers
FHIR for v2 implementers
One thing I am still trying to understand is the issue of aggregation, which is accomplished using Atom feeds, which in itself makes a lot of sense, and may have all sorts of side benefits, though it is a little frustrating if your browser (ie Opera) tries to handle the aggregation as though it were a blog feed... but there are two kinds of aggregation we must consider here one when we aggregate a number of like resources, for instance as the result of a non-deterministic search, and the other when multiple types of resource are aggregation into a Message (for v3) or Document (for CDA) Resource.
I find this confusing because it appears that position is used within the Atom syndication to determine which Resource to use as the transport wrapper. Still trying to wrap my head around this. Should prove to get interesting very quickly.