We argue that the time has come to make major investments in the devel
opment of massively parallel applications, where the historic distinct
ion between SIMD and MIMD parallel computers should not be the main co
ncern. A classification scheme of applications based on the structure
of data, communication and program control is presented and exemplifie
d with recent applications. The discussion on strategies for developin
g massively parallel applications is closed with a brief look at the f
uture.