We identify principles characterizing Solomonoff Induction by demands on an agents external behaviour. Key concepts are rationality, computability, indifference and time consistency. Furthermore, we discuss extensions to the full AI case to derive AIXI.