Gordon Bell [Bell:95a] described the natural snappy evolution of computing to a universal parallel or distributed system of commodity networks (such as ATM), and commodity PC compute nodes. Here, we build on his extrapolation, and describe our vision of corresponding software environment at both the application and operating system levels. The current HPCC (High Performance Computing and Communications) activities can interact with this evolving world in two distinct ways--firstly, the World Wide system of several hundred million PC's and set top boxes is a remarkable parallel computer to which we can apply our expertise in parallel algorithms and methodology. Secondly, it is natural to build parallel computing environments in terms of pervasive World Wide Web (WWW) technologies. This bottom-up approach illustrated in Figure 1, should be contrasted with the more typical top-down HPCC approach. As an example, it is certainly laudable and useful to port HPF (High Performance Fortran) to run on a SNAP network of PC's; this will allow the (small) HPCC community access to such hardware, but it will not persuade the PC community to use HPF. Our bottom-up philosophy would be illustrated by the development of parallel versions of Java---the new language of the Web---optimized for MPPs (Massively Parallel Processors).
