I include this paper to show an example of a functional language based composition system. Theoretically, a performance engine should be powerful enough to support incorporation of a wide variety of compositional systems.
The paper describes a sound synthesis system based on writing lisp like functions.
Note that the output of the system is audio. The question of integration becomes what signals need to pass between the "control/sequencing" system and the compositional one to integrate it into the performance? Is control transport adequate as an interface between these two types of modules.
parameter modification through cue and timelines
tweaking parameters manually
allocation/deallocation of resources (render agents, media sources)
Note: Can resources be managed like java beans? In other words, can they be described in a way that generalizes their attributes? For example if the system was originally designed to take input from a trackball, could this be described as a "pointing device interface" and replaced by another pointing device like a mouse? The system could "know" this through a description of the interface to the composition engine and a way to define resources in the system. (this is similar to the idea of a driver only on the input/resources rather than control side of things).
Is there a general theory about how to bring compositional agents into an integrated performance system. What different types of agents are there?
2. understanding (sensor)