As mentioned previously, it is proposed to restrict the class of structures to those that possess a normalising function, and to use that function, possibly expressed as a Markov algorithm, as a description. The idea of a unique path naming strategy is inspired by [9]. It describes a system whereby the nodes of structures are given a name depending on the position of the node. This simplifies the implementation of the data structure for a SPMD machine. Additionally, aliased nodes (where nodes can be reached by multiple paths) are dealt with by maintaining a set of alternative names. This contrasts with our approach in that we attempt to study structures where the aliasing can be determined at compile time and the multiple names resolved uniquely. In [9], Gupta does also consider the idea of improving the efficiency of the implementation by analysing the programs.