A self-replicating automaton capable of being programmed to construct a wide range of objects in addition to a replica of itself may be termed a self-replicating programmable constructing automaton (SRPC). An open problem in the field of self-replication is the design of a physical SRPC made from simple parts taken from a small set and using for its supply of parts (from which to construct things) a disorganized collection of parts distributed in a random fashion in its environment. A machine capable of doing this will need to contain a subsystem capable of taking an unknown part from its environment and determining which kind of part it is. The architecture of a discriminating subsystem of this kind is of course highly dependant on the precise nature and size of the set of parts used. The general principles of this type of discriminating system are discussed, and a specific design for such a system in an abstract discrete-space kinematic simulation environment is presented.