Chapter 2½ Levels and Abstractions 15

its most aggregate behavior. It then consists of central processors, core memories, tapes, disks, input/output processors, communication lines, printers, tape controllers, buses, Teletypes, graphics terminals, etc. The system is viewed as processing a medium- information-which can be measured in bits (or digits, characters, words, or the like). Thus the components have capacities and flow rates as their operating characteristics. All details of the program are suppressed, although many gross distinctions of encoding and information type remain, depending on the analysis. Thus one may distinguish program from data, or file space from resident monitor. One may remain concerned with the fact that input data are in alphameric and must be converted into binary, or are bit-serial and must be converted to bit-parallel.

We might characterize this level as the "chemical engineering view of a digital computer," which likens it more to a continuous- process petroleum-distilling plant than to a place where complex FORTRAN programs are applied to matrices of data. Indeed, this system level is more nearly an abstraction from the logic level than from the program level, since it returns to a simultaneously operating flow system. Figure 10 illustrates a PMS diagram for a dual-processor UNIVAC 1108.

The PMS descriptive system is meant to provide a notation for the top level of computer systems. Figure 10 is given in this notation. On the surface it is largely self-explanatory, given the mnemonics of P for processor, M for memory, S for switch, T for transducer (hence also terminal), and K for control (since C is for computer). There is also L for link, but in most computer structures it is unnecessary to distinguish a separate link component, except to show connectivity. (It does become appropriate if communication delays exist.)

There is an issue about whether this small set of components is an appropriate set of primitives, but the issue is not of major proportions. The real issues in the development of the notation come from the stress of two opposite forces. On the one hand, one wants extremely compact notations for expressing computer systems. The systems are large in any event, and if there is much extra notational freight in the way of fixed formats or forced writing of what is already known and assumed, then the notation