When making modes, it is a constitutive question, how a system can be described and where the borderlines between a system and its environment can be drawn. The epistemic position of system realism (there are systems as ontic entities) and system nominalism (systems are only descriptions) leads to the concept of natural surfaces on the one side and to the authorship of defined borders on the other side. A more mediated position, the system descriptionism, is unfolded here: it is shown that changing objects or processes, which are described as systems, require new system theoretical concepts. As one of them, the notion of non-classical system is introduced, and a classification thereof is proposed.