Today's computing systems provide all kinds of media output to the users including pictures, sounds, lights and visual animations; but user input is mostly restricted to narrow, time-consuming input modes such as mouse, touch screen or keyboard, even if the system could sense this information implicitly from the human body. Physiological computing allows one to re-balance this information asymmetry by considering also physiological data of the users as input during on-line processing. Using such inputs, physiological computing systems are becoming able to monitor, diagnose and respond to the cognitive, emotional and physical states of persons in real time. In this paper we give an introduction to physiological computing, discuss the challenges that typically arise when developing physiological computing systems, and solutions that we found promising in the creation of case studies. More specifically, we discuss the concepts software frameworks should incorporate to provide guidance in the development process: component-orientation, data processing support, and distribution support were found useful when dealing with physiological computing systems, where real-time sensing, on-line data processing, actuator control, context-awareness and self-adaptation are involved. Furthermore, we discuss why agile software development methodologies seem appropriate to structure development efforts involving physiological computing. Finally, we cover the issues in validation and verification of physiological computing systems that arise from the characteristics of the application domain, and discuss how empirical validation through psychological experiments, software validation and verification can interact with each other in the domain of physiological computing.