Structuring programs for multiprogram time-sharing on-line applications

Abstract

The modern art of computation has developed from plugboard programming through the stored machine instruction programs controlled by the users on the consoles, then to problem-oriented symbolic programs computed in the batch mode, towards the on-line computing during which the users have a large amount of control over their programs. The lower cost per computation and flexibilities of a large capacity high-speed computer naturally lead us to consider the provision of on-line computing service to several users on a single high-performance machine in a time-sharing mode, rather than several smaller machines, one for each individual. To maximize the efficiency of a man-machine team working in an on-line computing mode, it is desirable to let the man choose the language---say English---for communication and to let the machine do the translation. This idealistic goal is not impossible, but is currently impractical. A good compromise is to select as the user language a formal language such as ALGOL, FORTRAN or LISP which has a set of explicit syntactical rules and a small set of basic vocabulary. The user then may extend the vocabulary by declarative statements and communicate with the machine in the extended vocabulary. Due to frequent message exchanges between the man and the machine during on-line computing, the machine representation of users' programs must be easy to modify at the source language level. The technological trend towards large random access memory suggests the retention of several users' programs in core simultaneously, hence mutual memory protection must be ensured.