Massachusetts Institute of Technology, Laboratory for Computer Science research records

PROJECT HISTORIES

Kludge

Graphics research in the 1960s grew out of a desire to improve communications between computers and their users. Early researchers focused on the goal of getting a graphic image on a screen and making user interaction with the image fast and convenient. Ivan Sutherland's MIT Ph.D. research, which created a graphics interactive device called Sketchpad in 1963, aroused a great deal of interest.

The MIT Electronic Systems Laboratory Computer Applications Group worked on graphics in 1963. John Ward and Doug Ross looked for ways of reducing the computational load required for real-time graphics display. Project MAC shared the cost of this work with ESL. These efforts produced the ESL Display Console, nicknamed the Kludge, which provided faster, more direct, computer-controlled display capabilities in 2-D with rotation and allowed interaction by means of switches, knobs, a light pen, a keyboard, and a globe input device. Use of incremental digital interpretation for line generation reduced the computational load and increased speed. The Kludge was installed in 1964 on the CTSS and was the first graphic display capable of operating on a time-sharing system.

Many researchers from various departments at MIT found the Kludge useful in their work. Additional users came from Harvard.

The Kludge contained a Digital Equipment Corporation Type 330 Incrementing Display made to MIT specifications. The Incrementing Display consisted of a magnetic-deflection cathode ray tube with deflection amplifiers, digital-to-analog converters, digital registers containing beam position coordinates, and various controls. The active display surface was 9 3/8 inches on each side and had 1024 horizontal positions and the same number of vertical positions, allowing the user to specify over 1,000,000 points on the screen. Lines could be drawn by producing strings of pulses into the horizontal and vertical registers.

A great deal of attention focused on the Kludge in the 1960s and early 1970s. Improvements and additions included a stereo routine and 3-D drawing programs. It was used for displaying stress contours, creating models of molecules, modelling the vocal tract for speech analysis, designing highway interchanges, naval architecture, and many other forms of graphical input and output.

ARDS

The Advanced Remote Display Station (ARDS) console was developed at Project MAC between 1965 and 1968 by R. H. Stotz, T. B. Cheek, and others in the Display Group of the Graphics Research division. The research addressed a need for low cost display equipment that could be widely installed to extend the range of potential applications of dynamic graphics capability. The ARDS console developed at MAC provided high-speed alphanumerics and full graphical capabilities and could be used over an ordinary voice-grade telephone line. This development occurred in two phases: ARDS-I was the breadboard model and ARDS-II contained a full console with a full-screen vector generator and a character generator capable of reproducing the entire ASCII symbol set.

ARDS used a direct view storage tube. Its development was hampered somewhat by the unavailability of such tubes of a sufficient size to allow adequate picture resolution. The Tektronix Type 611, which came on the market in 1967, eliminated this problem.

The main impetus behind ARDS development was to create a low cost console for dynamic graphics. To this end ARDS was made with integrated circuits for the digital portion of its logic, an integrated diode array for the read only memory of the character generator, and new technology for conversion of digital line information to analog voltages.

Stotz and Cheek left the Display Group in 1968 to form a company to market ARDS. Project MAC and other groups at MIT purchased many of the units for use on CTSS.

CTSS and Multics

The "MAC" in Project MAC stood for "Machine-Aided Cognition" as well as "Men and Computers." The idea behind machine-aided cognition was that a system could and should be devised to enable a human and a computer to work together in a real-time collaboration focused on solving a problem. Such intimate collaboration would, in a practical sense, require a computing system that could be used simultaneously by many humans and would be physically and intellectually convenient to use. John McCarthy had proposed a multiple-access computer system along these lines in 1961, and his opinion was shared by an MIT committee whose mission was to assess the future computing needs of the Institute. Under the leadership of Fernando Corbato, the MIT Computation Center created the Compatible Time-Sharing System (CTSS) and demonstrated its possibilities near the end of 1961. CTSS continued to evolve throughout 1962 and 1963. When Project MAC started in 1963 as an interdisciplinary effort funded by DARPA its main purpose was to expand on the successes of CTSS to create a MAC Computer System with more elaborate time-sharing capabilities. To this end Project MAC proceeded to search in the Fall of 1963 for the right computer and the right computer manufacturer, one sufficiently interested in time-sharing systems to collaborate with MAC in developing significant modifications and additions. MAC chose General Electric and the G.E. 635 computer, later modified into the G.E. 645.

MAC's hardware specification phase was complete by 1965. In 1965 Bell Telephone Laboratories decided to participate in MAC's software development effort and acquired the same hardware installation from G.E. MAC, G.E., and Bell became partners in the Multiplexed Computing and Information Service (Multics) Project in 1965. The project continued for longer than expected, and was more expensive in terms of money and personnel than had been anticipated. When Multics was sufficiently advanced to replace CTSS, the older system was switched off and Multics became the MAC time-sharing system. Development and improvement of time-sharing systems was the main focus of Project MAC research for seven years. Multics became operational in November 1969.

Bell Labs withdrew from the partnership early in 1969. MAC and Honeywell Information Systems, which had bought the computer division of G.E., continued to collaborate on Multics improvement until 1977. Honeywell started offering Multics as a commercial product in 1973.

MACSYMA and Mathlab

MACSYMA is an acronym meaning MAC's Symbolic Manipulation System. It is a large computer program written in LISP, running on an operating system called ITS, or Incompatible Time-sharing System. It was developed by the Mathlab Group at Project MAC for performing symbolic as well as numerical mathematical manipulations. With MACSYMA the user is able to differentiate, integrate, take limits, solve systems of linear or polynomial equations, factor polynomials, expand functions in Laurent or Taylor series, solve differential equations (using direct or transform methods), compute Poisson series, plot curves, and manipulate matrices and tensors. MACSYMA has a language similar to ALGOL-60 to permit users to write their own programs for transforming symbolic expressions.

William A. Martin's Ph.D. dissertation at MIT, "Symbolic Mathematical Laboratory" (1967), describes a large computer program he developed to aid applied mathematicians in the solution of problems in non-numerical analysis involving tedious manipulations of mathematical expressions. It was designed to solve sample problems and analyze problems of input, transformation, and display.

Joel Moses' Ph.D. dissertation at MIT, "Symbolic Integration" (1967), describes how he used the rational function package of Carl Engelman's Mathlab in writing a program for symbol integration, SIN, which was more powerful than the best previous program, SAINT, created by James Slagle for his MIT Ph.D. dissertation in 1961. Moses' thesis discusses SIN (Symbolic Integrator program) and SOLDIER (Solution of Differential Equations Routine) and projects a future mathematical online laboratory.

Martin's thesis and the related work done by Moses at MIT and Engelman at MITRE Corporation gave Martin the idea of creating a large, powerful symbolic manipulation system, combining various techniques and representations. He was the leader of this development between 1968 and 1972. Actual development of MACSYMA started in July 1969 in the Mathlab Group at Project MAC. An operational system was in place by July 1970. The system had its first users in 1971 and evolved as additional contributions were made to it. Of particular importance in this regard was the research completed by Richard Fateman and Paul Wang in the early 1970s.

Mathlab was originally a system for computation developed by Carl Engelman. The name then attached to a group at Project MAC, whose principal research was on the system eventually called MACSYMA. At times in early records of the MACSYMA Project, the terms Mathlab and MACSYMA are used interchangeably.

A MACSYMA consortium was formed in 1976. MACSYMA was the subject of important "users' conferences" in 1977, 1979, and 1984. Many additional published and unpublished papers illustrate its usefulness in various fields. MACSYMA proved to be very successful and heavily used, not only in Cambridge, but also by off-site users who accessed it through ARPANET. MIT licensed MACSYMA to Symbolics, Inc., in 1983, and MACSYMA, Inc., acquired the rights in 1992.

Network

ARPA-funded time-sharing developments such as CTSS and Multics at Project MAC illustrated the feasibility and benefits of sharing computer resources. L. G. Roberts at ARPA and others were planning a wide network as early as 1967. Computer time-sharing experiments had given researchers insights that proved very helpful in the design of the network that came to be known as the ARPA network or ARPANET (the ancestor of the Internet).

The ARPANET was designed by an Interface Message Processor Group composed of representatives from many ARPA contractors as well as ARPA itself. The network was implemented by Bolt Beranek and Newman (BBN). It started operating with four nodes in 1969 and had expanded to include Project MAC and 23 other nodes by 1972. It was developed as an experimental system that would allow ARPA to link together dissimilar computers at dispersed sites (universities doing defense-related research, defense contractors, and military installations). The idea was that such a network would decrease duplication of expensive hardware and permit sharing of software, promote communication by allowing computers at different sites to have conversations at a high data rate, and facilitate decisions and coordinate actions by means of electronic mail. The implemented network had two basic functions: as a research project to test ways of sharing computer resources among different kinds of computers and operating systems; and as a facility for providing services to a large group of computer users. Dynamic rerouting was an important aspect of ARPANET: if an enemy attack knocked out a Network link, the traffic on it could automatically be rerouted to other links.

The network's basic structure called for a subnet of small computers called Interface Message Processors (IMPs) created by interconnecting them through wideband communication lines. The communications subnetwork was a packet-switching system (also known as a store-and-forward message switching system). Message switching did not require a dedicated path between computers that wanted to communicate. Such computers sent instead a sequence of messages, each of which had an address, and each of which might travel through several network nodes en route from sender to receiver, being stored at one node until picked up by another. IMPs controlled traffic flow and routed messages. Host computers at the various military sites, universities, and manufacturers were connected to the IMPs, and communicated via them, to form the ARPANET.

Additional ARPA-funded research focused on network security measures to prevent unauthorized intrusion. Some nodes split off in 1984 because of security concerns and formed the MILNET. However, because of the success of security programs developed in the 1970s, MILNET was able to maintain important connections to ARPANET services.