Abstract. Building a repository of proof-checked mathematical knowledge is without any doubt a lot of work, and besides the actual formalization process there is also the task of maintaining the repository. Thus it seems obvious to keep a repository as small as possible, in particular each piece of mathematical knowledge should be formalized only once. In this paper, however, we claim that it might be reasonable or even necessary to duplicate knowledge in a mathematical repository. We analyze different situations and reasons for doing so, provide a number of examples supporting our thesis and discuss some implications for building mathematical repositories.