Since their inception over forty years ago, L-systems have proven to be a useful conceptual and programming framework for modeling the development of plants at different levels of abstraction and different spatial scales. Formally, L-systems offer a means of defining cell complexes with changing topology and geometry. Associated with these complexes are self-configuring systems of equations that represent functional aspects of the models. The close coupling of topology, geometry and computation constitutes a computing paradigm inspired by nature, termed developmental computing. We analyze distinctive features of this paradigm within and outside the realm of biological models.