From the description of the planned domain, the types of perturbation the input can suffer are identified, whether beneficial or harmful, intentional (adversarial) or not (sensor defects, environmental change), and mathematically describable (a blur kernel) or only exemplifiable (droplets on a lens shown through masks). Where the way a perturbed input is generated varies between people, the process must be explained. To assess robustness against a perturbation formally, a function describing its application is required; it must depend on one or more bounded parameters, each with a minimum and maximum variation set, one or more criteria of Clause 5 must be expressed on the domain, and a formal method should then evaluate those criteria over the admissible perturbations. Several perturbations are assessed by composing their functions, preferably commutatively.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.