In this presentation, we explore and question the concept of motion safety, i.e. the ability to avoid collision, for robots sharing their workspace with people. We establish that absolute motion safety, in the sense that no collision between the robot and the people will ever take place, is impossible to guarantee (hence the open nature of the motion safety problem). We then discuss the choices that are available: mere risk minimization or what we call weaker motion safety, i.e. types of motion safety that are weaker than absolute motion safety but that can actually be guaranteed. In all cases, we argue that if robots are ever to be deployed among people, it is important to characterize the level of motion safety that can be achieved and to specify the conditions under which it can be guaranteed.