In recent years novel distributed computing environments termed grids have emerged. Superficially, grids are considered successors to, and more sophisticated and powerful versions of, conventional distributed environments. This paper investigates the intrinsic differences between grids and other distributed environments. From this analysis it is concluded that minimally, grids must support user and resource abstraction, and these features make grids semantically different from other distributed environments. Due to such semantical differences, grids are not simply advanced versions of conventional systems; rather, they are oriented towards supporting a new paradigm of distributed computation.