Long term dynamics of farming systems (FS) are closely linked to farms life-cycles. Farming systems are also influenced by events of their context. Dynamics of FS might be of two orders: (i) evolution corresponding to incremental changes of the FS inside a frame of action and (ii) transition implying radical change of FS moving from a frame of action to another. Conversion from conventional to organic agriculture might promote such a change of frame, asking for specific experience and knowledge for farmers. In this abstract, we will present how agronomic research deals with farming system dynamics on time steps longer than the agricultural campaign. We will then propose a way to formalise transition of FS. French authors formalised farming systems as two sub-systems interacting in a global context: a human and a biotechnical sub-systems. Interactions between the sub-systems were formalised by agricultural practices. Those researches, mainly involving agronomic and management sciences, focused on the farmer seen as a manager in a static farming system. Most of long term approaches of FS don't give interest to dynamics of change. They focus on new forms of FS potentially able to face future challenges. Systemic approaches in crop and livestock sciences gave interest to adaptive capacities of FS. "Room for manœuvre" defines the capacity of systems to adapt in a defined plan where as "flexibility" defines the capacity of a system to cope with determined fluctuations of its environment. Thoses approaches relie on biotechnical sub-systems represented as static or cyclic over the long term. Human sub-system has been represented (i) mostly following the theory of rational action, farmers acting accordingly to determined plan, and (ii) more recently, following the theory of situated action, farmers adjusting practices to fluctuations of their environment. Few researches analyse transitions and evolutions of FS facing unanticipated context. They ask for a focus on changing processes involved in the system. Co-evolution of the human and the biotechnical sub-systems are promoted by dynamics of farmer activity through learning. Evolution and transition are generated by, respectively, simple and double loop learning. Approaching dynamics of farming systems through learning processes involves (i) reconsidering French FS model in order to integrate experiment and knowledge coming from productive tasks of farmers through action creativity (ii) questioning the nature of the frame of action in FS