The systematic study of software self-adaptation has emerged as one of the key areas of software engineering. The challenges and the ontology relevant to this area are still being formulated. I take this opportunity to present some of my observations on compliance, context, and uncertainty as they pertain to adaptation. I also argue that requirements engineering, and to a large extent software engineering, takes a centralized perspective of systems, and therefore cannot model let alone enable reasoning about adaptation in multiagent systems.