* Beam spot constraints, prior information, ghost tracks, vertex suppression field, and other "extra information" needs to be respected in the interfaces, without bloating the API. 

 * Zvres and a special-purpose MultiVertexSeeder for the application of b-tagging are desirable future developments. 

 * Gaussian sum filters will be introduced, allowing for non-Gaussian track models. (Think e.g. modelling bremsstrahlung electrons here)