I'm considering implementing geometric semantics perhaps as an extension to Boost.Geometry. In essence, it attempts to deal with changes in reference frames in a general way, which is something Boost.Geometry does not capture yet. Sone of the conceptual equivalents of the OGC specification's approach are described in the papers referenced by an author here:
There is an existing implementation of their specification, but it isn't generic in the way boost.geometry is. An implementation may also involve boost.graph, I'm not sure yet. Would there be any interest in this? While the linked documents are about robotics the underlying concepts apply to many other domains including GIS, graphics, and simulations.