Network Working Group Peggy M. Karp Request for Comments: 321 MITRE NIC #9608 24 March 1972 Related RFCs: 313 Categories: F CBI Networking Activity at MITRE This RFC is in response to Tom O'Sullivan's probe for comments on Computer Based Instruction. MITRE is currently doing some pre- liminary work in the area,at the present time limited to use of CAI- related network resources. The most expedient way for me to respond is by attaching excerpts taken from a recent status report prepared for internal MITRE use, intended for staff not generally familiar with the ARPANET. Comments directed at specific items in Tom's attachment will follow in a future RFC. We welcome comments on our activities and would like to partici- pate in any CBI dialogs that take place. [Page 1] STATUS OF IR&D TASK 9780 ARPA NETWORK STUDIES AND EXPERIMENTS 1.0 OBJECTIVES MITRE/Washington became a node in the ARPA computer network in September 1971 when a Terminal Interface Message Processor (TIP) was installed. Since that time MITRE's use of the network has largely been supported by IR&D funds. The objectives of the IR&D Network Studies are: (a) to promote the use of the network resources in an operational mode within MITRE to increase the computer capabilities available for supporting sponsor's work; (b) to use the ARPANET itself as a research tool for conducting computer networking experiments; and (c) to demonstrate the use of ARPANET resources to extend the capabilities of existing systems. Specifically, the proposed objectives are: (1) to select ARPANET resources that can be used to support identified MITRE projects and to demonstrate how these resources can be used; [Page 2] (2) to use the ARPANET as a research tool to conduct data sharing experiments to study techniques for data handling on a computer network; and (3) to demonstrate the feasibility of using remote resources on the ARPANET to augment the capabilities of the TICCIT* system. *The principle use of the TICCIT (Time-Shared Interactive Computer Controlled Information Television) system has been for Computer Assisted Instruction (CAI). A TICCIT/CAI system is currently being developed by MITRE under a National Science Foundation (NSF) grant. [Page 3] 2.0 APPROACH The IR&D Network Studies and Experiments project requires soft- ware development on selected ARPANET Host computers and extensive exercising of network resources. The approach taken to fulfill the project objectives involves effort in three areas. Task area I is designed to build up ARPANET usage by MITRE. Task II involves conducting data sharing experiments on the network utilizing software being built by MITRE at several network sites. Task III is intended to demonstrate the use of CAI-related network resources to augment the TICCIT/CAI system. The specific activities involved in each task area are discussed below. 2.3 Task III - Demonstrate Extensions to TICCIT System Capabilities --------------------------------------------------------------- That Are Possibly by Drawing on ARPANET Resources ------------------------------------------------- The structure of the ARPANET is such that it is technically possible to combine resources on the network to form a "distributed system." A first step in this direction would be to use network resources to augment a special-purpose system connected at one of the nodes. The special-purpose system would serve as the vertex of the distributed system, dynamically drawing on network resources to perform specific functions. [Page 4] The TICCIT/CAI system is a good candidate for demonstrating the feasibility of the concept of a distributed system. However, rather than initially connecting TICCIT to the network, resources on the network will first be used to demonstrate how they could augment the capabilities of TICCIT Task III is organized as three subtasks to demonstrate that the range of curriculum material can be extended, to demonstrate that the TICCIT/CAI system can be supplemented with specialized CAI systems that use AI techniques, and to demonstrate that TICCIT system resources and capabilities can be augmented by using computing and storage resources available on the network. (a) Demonstrate that the range of curriculum material can be extended. The range of curriculum material offered by the TICCIT/CAI system can be extended by using network resources. Two areas that were selected for the demonstration are mathematics and computer science. The Culler-Fried On-Line Graphics-Oriented system available on the IBM 360/75 at the University of California at Santa Barbara (UCSB) was selected for use in mathematics courses. A collection of diverse systems and languages will be selected to demonstrate how they can be incorporated in a computer science curriculum. The following tasks are necessary for demonstrating that the range of curriculum material can be extended: * learn the use and structure of the Culler Fried system. * use the Culler-Fried system over the network using an [Page 5] IMLAC graphics display terminal. * design sample courses for demonstration of the use of the Culler-Fried system in a mathematics curriculum. * develop software to incorporate the use of the Culler- Fried system in the sample courses and demonstrate. * prepare a sample computer science curriculum and demonstrate how the wide range of systems on the network can be utilized in a typical computer science curriculum. * determine and document the technical specifications required for a TICCIT/CAI system interface. (b) Demonstrate that the TICCIT/CAI system can be supplemented with specialized CAI systems that use AI techniques. There are two network resources utilizing AI techniques that have been applied to computer assisted instruction tasks: SCHOLAR and LOGO. SCHOLAR is a mixed-initiative rather than a frame-oriented CAI system. LOGO is a LISP-based programming language designed to study whether notions and skills of formal reasoning and problem-solving can be taught. We will investigate if these systems can be used to supple- ment the TICCIT system and, if possible, will demonstrate their use. The following tasks are required: * investigate the use of SCHOLAR on the TENEX system at BBN over the network. * investigate the use of LOGO and the "turtle" over the network from the PDP-10 at MIT/AI and/or from TENEX at BBN. * demonstrate the use of the systems over the network. [Page 6] * determine and document the requirements for an interface with the TICCIT system. (c) Demonstrate that TICCIT system resources and capabilities can be augmented by using computing and storage resources available on the network. It should be possible for the TICCIT/CAI system to substantially in- crease its capacity by using large data storage devices on the ARPANET for storing student records and curriculum and by using some Host computer to run complex statistical analysis programs to analyze and evaluate student progress and course content. In order to demonstrate that such a scheme is possible, the following tasks are necessary: * determine format and requirements for curriculum material and student records. * develop a method to store and retrieve data on a demand basis. * select a Host on the network and develop software for a demonstration. * determine and document requirements for an interface with the TICCIT system. 3.0 PROGRESS OF WORK In this section the progress of the work associated with each task is presented. Both achievements and problems encountered are discussed. The manpower expended thus far for each subtask is given in graphic form, together with a proposed schedule for completing [Page 7] the work. 3.3 Task III - Demonstrate Extensions to TICCIT System Capabilities --------------------------------------------------------------- That Are Possible by Drawing on ARPANET Resources ------------------------------------------------- Many of the resources on the ARPANET were designed as stand alone systems not intended for use over the network nor for facile interfacing with other systems. Thus we have found that more work than initially was estimated is needed to use the selected CAI-related systems over the network. Therefore we plan to fulfill the goals of each subtask by demonstrating the use of selected resources over the network but will not attempt to interconnect the systems in any fashion during FY1972. A proposal to investigate the notion of a distributed CAI system using ARPANET resources is under preparation. The status of our progress in using CAI-related resources over the network is presented below. (a) Demonstrate that the range of curriculum material can be extended. The Culler-Fried (C-F) On-Line Graphics-Oriented system at UCSB was selected to demonstrate its use in mathematics curriculum. The system has been used successfully in a variety of courses at UCSB, including chemistry, mathematics, and economics(30,31,32,33,35). The Culler-Fried system's normal mode of operation is with two keyboards and a Tektronics graphics display device. The keys on the second keyboard are "function" keys; for example, by pressing a single key the user can initiate complex actions such as displaying a plot [Page 8] of a convolution function(35). The system can also be used to provide online access to the 360/75 system to remotely control the execution of programs. The latter service is currently available on the network through TELNET. Access to the system was made possible by defining a mapping from ASCII characters (sent as the "network virtual terminal") to characters expected by the OLS. Thus it is often necessary to type a sequence of characters on a TTY-type device to invoke the action of a single function key. Under the current implementation, graphics output to the network is suppressed. We plan to demonstrate the use of the C-F system on an IMLAC graphics device attached to our TIP. During the first phase of our implementation, we will not simulate the C-F function keyboard but will enter our graphics input using the procedures defined for use via TELNET. For output, UCSB agreed to provide a new output processor which will no longer suppress graphics output, but will map it into the remote graphics capabilities as provided by IMLAC (Figure 3). Such a system was implemented in early February but due to the character-oriented nature of the IMLAC device available to us, the resolution was unsatisfactory. We have now defined a low-level graphics protocol. UCSB has agreed to send graphics output to us using this protocol. We have had to program new processes for the IMLAC and are now debugging the new programs. Due to the limited core available on the particular IMLAC that we are using (4K 16 bit words) which limits the size of internal display lists, we will not [Page 9] be able to plot very complex graphs. However, the software being developed is general and can be moved to an IMLAC with larger core capacity. We currently send alphanumeric input from a TTY or 3300 and divert the graphics output to the IMLAC. When we can access the Culler-Fried system for graphics output successfully via the network, we plan to modify the IMLAC programs to permit alpha input from the IMLAC keyboard. We also plan to investi- gate the possibility of attaching a function keyboard to the IMLAC. Lastly, we intend to investigate the use of TICCIT display devices with the TIP. A sample mathematics course will be designed to assist in demonstration of the CAI applications of the Culler-Fried system. We are currently studying the past uses of the system in a CAI mode. We have not yet investigated computer science curriculums. We plan to get inputs from the ACM Committee on Computer Science Educa- tion(36,37) and to visit universities in the Washington area. This activity is scheduled to commence in April. (b) Demonstrate that the TICCIT/CAI system can be supplemented with specialized CAI systems that use AI techniques. The two network resources selected for demonstration are SCHOLAR and LOGO. Both systems are implemented in LISP and are currently avail- able over the network on the TENEX system at BBN. A version of LOGO with a "display turtle" is available on the PDP-10 at MIT/AI, however, they do not yet have their NCP implemented and thus are not currently accessible over the network. There is also a version of LOGO on the TENEX system at SRI/AI. They also are not currently connected to the [Page 10] network and we have not investigated their version. We used SCHOLAR over the network in January with a data base provided by BBN to review the knowledge of a student in the geography of South America(38,39). The LOGO system at MIT/AI is perhaps the most impressive system for use in a demonstration due to the availability of a "display turtle". The system is currently being used remotely by the Bridge School in Lexington, Massachusetts. We visited the school and observed a class in session. We also attended an undergraduate class in Applied Math at MIT that was learning turtle geometry. Seymour Papert of MIT expects the MIT/AI PDP-10 to be up on the network by the middle of April(40). He has agreed to modify his system to interface with our IMLAC via the network using the low-level graphics protocol that we specified. He has developed many courses and games using LOGO(41,42,43,44,45) that provide sufficient material for use in demonstrating the system. A real turtle may be available in early summer to run from our TIP. We have used the LOGO system at BBN via the network. Since the system is continually being modified, Dr. Feurzeig agreed to put a clean version on the RAND TENEX for our use. BBN's LOGO does not currently have a display turtle, however, they are willing to imple- ment one. Even without the display turtle, courseware that has been developed for teaching mathematics provides sufficient material for us to work from(46,47). A radio controlled turtle has been designed [Page 11] at BBN. It may be possible to obtain the specifications and have one built at MITRE to run from our TIP. (c) Demonstrate that TICCIT system resources and capabilities can be augmented by using computing and storage resources available on the network. Work has not begun on this subtask. However, much of the software developed for the data sharing experiments can be used to store and retrieve data on a demand basis. We have received preliminary curriculum material from TICCIT personnel. We expect to interact with them to determine more speci- fically the format and requirements for curriculum material and student records. [ This RFC was put into machine readable form for entry ] [ into the online RFC archives by BBN Corp. under the ] [ direction of Alex McKenzie. 12/96 ] [Page 12] REFERENCES 30. Ewig, C. S., Gerig, J. T., and Harris, D. P., "An Interactive On-Line Computing System and Its Use in Chemistry Education", Department of Chemistry, UCSB. 31. Howard, J. A., and Wood, R. C., "Computer Assisted Instruction in Engineering Using On-Line Computation", _Journal_of_Engineering_ _Education_. 32. Sullivan, J. J., "Computer Based Instruction in Economics: A Report on Facilities and Applications at UCSB", paper presented at a conference on Computers in Undergraduate Curricula, University of Iowa, Iowa City, Iowa, 1970. 33. Wood, R. C., and Bruch, J. C., Jr., "Teaching Complex Variable with an Interactive Computer System", article submitted for review and publication in _IEEE_Transactions_on_Education_, July 1970. 34. Wood, R. C., and Howard, J. A., "An Interactive Computer Class- room, _Educational_Research_and_Methods_Journal_, Vol. 2, No. 4, June 1970, pp. 29-31. 35. "UCSB On-Line System Manual", NIC #6502, September 1971. 36. "Curriculum 68", _Communications_of_the_ACM_. Vol. 11. No. 3. March 1968, pp. 151-197. 37. Teichroew, D., ed., "Education Related to the Use of Computers in Organizations", _Communications_of_the_ACM_, Vol. 14, No. 8, September 1971, pp. 573-588. 38. Carbonell, Jaime R., "AI in CAI: An Artificial Intelligence Approach to Computer-Assisted Instruction", _IEEE_Transactions_ _on_Man-Machine_Systems_, Vol. MMS-11, No. 4, December 1970, pp. 190-202. 39. Carbonell, Jaime R., "Mixed-Initiative Man-Computer Instructional Dialogues", BBN Report No. 1971, 31 May 1970. 40. Williams, R. W., "LOGO Manual", MIT/AI, Draft memo, 9 April 1971. 41. Papert, S., and Solomon, C., "Twenty Things To Do with a Computer", MIT Artificial Intelligence Laboratory internal report, June 1971. 42. Papert, S., "A Computer Laboratory for Elementary Schools", MIT/AI Memo No. 246, LOGO Memo No. 1, October 1971. [Page 13] 43. Papert, S., "Teaching Children Thinking", MIT/AI Memo No. 247, LOGO Memo No. 2, October 1971. 44. Papert, S., "Teaching Children to be Mathematicians vs. Teaching About Mathematics", MIT/AI Memo No. 249, LOGO Memo No. 4, July 1971. 45. Papert, S., and Solomon, C., "NIM: A Game-Playing Program", MIT/AI Memo No. 254, LOGO Memo No. 5, January 1970. 46. INFORMATION PROCESSING MODELS AND COMPUTER AIDS FOR HUMAN PERFORMANCE Final Report, Section 3: Feurzeig, W., and Lukas, G., "Program- ming Languages as a Tool for Cognitive Research", BBN Report No. 2187, 30 June 1971. 47. PROGRAMMING-LANGUAGES AS A CONCEPTUAL FRAMEWORK FOR TEACHING MATHEMATICS, BBN Report No. 2165, 30 June 1971. Volume 1, Part 1: Feurzeig, W., Lukas, G., "An Introductory LOGO Teaching Sequence". Part 2: Lukas, J. D., and Lukas, G., "LOGO Teaching Sequence on Logic". Part 3: Feurzeig, W., Lukas, G., and Grant, R., "LOGO Reference Manual". Volume 2, Part 1: Grant, P., Falflick, P., and Feurzeig, W., "LOGO Teaching Sequences on Numbers". Part 2: Feurzeig, W., Lukas, G., and Grant, R., "LOGO Functions and Equations". Volume 3, Part 1: Lukas, G., Falflick, P., and Feurzeig, W., "LOGO Strategy in Problem-Solving". Part 2: Lukas, G., and Feurzeig, W., "LOGO Story Problems in Algebra". Volume 4: Weiner, W. B., Morgan, C. R., and Feurzeig, W., "The LOGO Processor, A Guide for System Programmers". [Page 14] ------------------------------------------------------------------------ Network Working Group Tom O'Sullivan Request for Comments: 313 Raytheon NIC: 9343 March 6, 1972 (CBI: 1) COMPUTER BASED INSTRUCTION Current development of Computer Based Instruction (CBI) systems seem to be directed toward two types of system: 1.) Small to medium scale, dedicated, stand alone systems (such as the IBM 1130, 1500, 1800 complexes) or medium scale systems with dedicated network implications (such as TICKET), and 2.) Large, centralized, dedicated systems with dedicated network implications (such as PLATO). Some attention has been given to the application of the resources of a General Purpose Computer Network to CBI (e.g., the EDUCOM efforts), however the full implication of the use of such resources do not generally seem to be understood by either the CBI development centers (at academic institutions or at the Armed Forces training or development centers), where most of the current activity takes place, or at ARPA Network Nodes, where most of the resources reside. This Request For Comment has two purposes To: 1.) Solicit comments from the Network Working Group, and others, on how selected classes of (and what specific) resources of a General Purpose Network might be applied to the field of Computer Based Instruction and 2.) Initiate a dialog between interested parties on the problems of Computer Based Instruction, not limited to, but including, the uses of General Purpose Computer Network resources. The attached paper discusses some of the applications of the resources of a large General Purpose Network to computer Based Instruction systems. Response and discussion are encouraged through the NIC system. O'Sullivan [Page 1] RFC 313 Computer Based Instruction March 1972 GENERAL PURPOSE COMPUTER NETWORKS AND COMPUTER BASED INSTRUCTION INTRODUCTION A high level of Computer Based Instruction (CBI) activity exists both in the academic and armed service communities, with the promise of a substantial amount of early development of instruction courses and instructional management facilities. The major functional areas of interest can be described as follows: 1.) Design and Development 2.) Field Tests 3.) Distribution and Operational Use 4.) Evaluation and Modification Specific computer support requirements are function of the philosophy and reflected strategy of implementation for each of the functional areas of interest. Design and development activities may focus on overall curriculum development or on specific training or educational goals involving a specific course. The focus of attention will have an effect on the support requirements, e.g., the type and size of data base, specialized processing capabilities, etc. Support requirements for Field Tests will be a function of whether they are to be performed at a central location, or are geographically distributed, particularly with respect to data collection procedures, computer support and terminal clusters, and communications. Solutions to the problems of the distribution and operational use of CBI systems and programs will be a function of the extent to which the training activity is proliferated (i.e., geographically or organizationally distributed). Both the level of activity, and the solution to problems, in the area of Evaluation and Modification will depend on the goals of the instructional process, the extent of dynamic change in the technology or specific application involved, and the degree to which the course(s) developed meet the needs for which they were intended. The above discussion has a heavy emphasis on Computer Aided Instruction (CAI) component of CBI, where the computer is directly used in the instructional process for lesson presentation, test, drill and practice, etc. Another component of CBI, Computer Managed Instruction (CMI), uses the computer as a management tool to guide the instructional process. CMI may be used in conjunction with CAI, or as an aid in guiding instructional processes of a more traditional nature. CMI, in addition to providing assistance in student selection, scheduling, and followup on past course performance, may provide guidance to instructors in the form of diagnostics of student O'Sullivan [Page 2] RFC 313 Computer Based Instruction March 1972 weaknesses, prescriptions for strengthening student understanding, and guidance in the redirection of students. In addition, CMI can provide management with evaluations of course and instructor effectiveness. CMI has corollaries to the discussion of CAI resource requirements and their relation to the philosophy and related strategy employed. Bearing in mind the effects on resource requirements of the complex considerations involved in CBI, there seem to be several areas in which the resources of a large General Purpose Computer Network, such as the ARPA Network, could be of high utility if properly applied. These include: 1.) The Network itself 2.) Centralized Data Storage 3.) Language processors 4.) Dialogue Support Systems As questions of philosophy and general strategy are resolved, or assumed, the hard questions of implementation come into play. Tradeoffs between competing approaches of the instructional strategy or model, techniques of measurement, languages, hardware, etc., must be made. It appears that both in resolving the tradeoffs, and in the implementation stage, network resources could prove to have high utility. THE NETWORK The network itself seems to have utility for CBI that goes beyond the function of providing a communications base for linking terminal(s) (individual or clustered) to processors dedicated to CBI. The latter function, however, is important. The communications network exists, and can be tied into efficiently from many parts of the country. If there were dedicated CBI systems on the network, it would facilitate: 1.) Evaluation of a single system (or its several components) for adequacy, or of competing systems for relative utility, by an interested user center, to assist in the selection of a system for a specific use; 2.) Early use by a geographically isolated user center, through use of clustered terminals, of the full power of a major CBI center, O'Sullivan [Page 3] RFC 313 Computer Based Instruction March 1972 a.) For a continuing period of low level use, or b.) Building over time until total usage by the isolated center justifies the installation of a full CBI center of its own. Existing network resources also hold promise of utility. Many manufacturers' systems, with associated varieties of operating system software, are available on the ARPA Network. Within most of these, a variety of application software is available, some supporting CBI and data base applications. Therefore, even without the presence of a dedicated CBI system available as a service center on the network, use could be made in support of CBI interests. 1.) Testing of existing language and data base systems for appropriateness to CBI problems. 2.) Development of CBI systems or components for demonstration and/or test. 3.) Testing of existing courses, lessons, or lesson segments to determine if they meet specified performance criteria, eliminating potential duplication of course development effort. 4.) Development of CBI systems to be operated under a specific set of hardware and software resources available in the network in anticipation of delivery of a dedicated operational system. 5.) Greater flexibility in selection of test sites for field test of courses developed, and performance of those tests prior to the delivery of operational hardware. 6.) Formalization of hardware support and associated software protocols to clustered terminals to provide continuing service to geographically remote training activities. Even the interests of dedicated CBI systems can be served, since the network has established communications and terminal support protocols that could tend to reduce the software efforts required to establish contact between a large dedicated service center, and clusters of user center terminals. In addition, terminal types not normally supported by the CBI service center machine might be accommodated by accessing the network through a compatible port, and getting into the service center through established network protocols. O'Sullivan [Page 4] RFC 313 Computer Based Instruction March 1972 Terminal access to the ARPA Network could be provided in the following ways: 1.) Single Terminals - over common carrier facilities to the nearest compatible HOST or TIP. 2.) Large Clusters - Simple single purpose TIP, or simple mini- computer supported in the network as far distant HOST. 3.) Small clusters - either of the above two depending on the length and intensity of expected use and the number or terminals in the cluster. In addition to the above general uses of a large General Purpose Computer Network, there are several specific classes of network resources that may be useful. CENTRALIZED DATA STORAGE The effect of economy of scale could reduce costs for smaller CBI systems if they make use, through the network, of mass storage on larger systems. If duplicate smaller systems are distributed in the field, then the centralized storage would have a multiplying effect on savings for lessons and lesson material, but special attention would have to be given to the file structure to permit efficient use of look ahead techniques for lessons, lesson segments, and individual student pages. For CMI data there are savings that go beyond the economy of scale. A single management system could be selected or built on a large service center machine to be used by CBI systems on the network, even though the operational CBI systems are supported by different manufacturers' hardware. This would not only reduce the cost for programming and maintaining CMI systems, but also facilitate cross system analysis and intersystem comparison, even though each using system would have its own set of files. The user of the network data reconfiguration service and data transfer protocol should make such operations feasible. This approach to CMI would assist in early development stages of course material by easing the problem of accessing data on past performance and norms. In the case of geographically distributed testing, the evaluation team would have faster access to performance data. Both the distribution and modification tasks seem cleaner since there is only one copy of the released version to be updated. If the trillion bit laser memory proposed for AMES becomes a reality, then the economy of scale argument can be expected to be dramatic. O'Sullivan [Page 5] RFC 313 Computer Based Instruction March 1972 LANGUAGE PROCESSORS A basic characteristic of a large General Purpose Computer Network is that it is capable of providing support from various manufacturers' machines. That is, such a network can be comprised of a number of special purpose processors that can be distributed geographically and organizationally to locations where the best support exists for each process. This characteristic makes it possible to select and join the best match of capabilities for a complex application . It is no longer necessary to settle for a hardware/software system that does a reasonable job in most areas of the applications need. CBI is a complex application. In addition to a good management system and associated data base, it requires heavy text handling for lesson material, table lookup and branching logic for acting on the student selected answers to multiple choice questions, a student arithmetic problem solving language for drill and practice, simulation capability of both physical processes (for laboratory and circuit simulation), and of decision processes (for gaming experience), and a future need for natural language processors to permit evaluation of free form student responses. In addition, there may be need for heavy statistical and arithmetic processing for course, student, and instructor evaluation. Depending on the course, various mixes of languages to support the above activities will be needed. Some believe that the language required for presentation of course material and evaluation of student response (and associated appropriate action) may be heavily dependent on the type of course being given. As we develop a deeper understanding of the learning process, we are likely to require expansion of languages to provide new functions and perform processes not yet identified. To provide expandability of languages, Meta-compiler techniques can be applied. Meta-compilers are in an early stage of development; however, several are available on the network. In addition to facilitating language expansion with minimum effort while preserving the workability of code written in the previous versions of the changing language, the Meta-compiler can be made to produce either compiler or program object code that will operate on several different target machines. This feature can give both programs and, in some cases, compilers that are transportable across machines, eliminating the need to settle on a single manufacturer's hardware when it is expected that a CBI compiler or interpreter, or a course or set of courses is going to be used in a way that requires O'Sullivan [Page 6] RFC 313 Computer Based Instruction March 1972 substantial geographic distribution. Hardware decisions can be based on the most cost-effective hardware for the combinations to be run at one time. Use of Meta-compilers will permit the development and debugging of new course material in advance of the delivery of the system selected for operations, even though the selected machine is not yet represented in the large General Purpose Computer Network. Field test can also proceed before the selected hardware arrives. Experience to date in the use of Meta-compilers indicates that the use of their high order languages to implement compilers and interpreters result in dramatic savings in both turnaround time and the absolute cost of producing a finished language product. DIALOGUE SUPPORT SYSTEMS In a field developing as rapidly as CBI, and at a time when substantial implementation is about to take place, dialogue between theoreticians, developers, and users is an important issue. New tools for supporting dialogue among members of a distributed group are currently in experimental use in the ARPA network. These new techniques not only support dialogue more rapidly than the distribution of papers, notes, and memos, but in some cases tend to sharpen the thought process and yield a better result. The application of such facilities, when ready, will be helpful beyond the early planning stages or projects. After plans are set, during the development of a project, a broader group of experts will be able to be called on to work on problems and questions as they occur. Later, as the product is being field tested (especially if testing is distributed or separated from the evaluation group), these new tools can be used to allow the test implementors to interact with each other and with evaluators in a more timely manner than a post- mortum meeting, resolving problems and questions as they occur, and as a side benefit producing more complete documentation of test progress. After the tests, when the product is being used operationally, these same tools can provide an excellent vehicle for tapping the ideas, suggestions, and enrichments contributed by the more creative instructors, and facilitate acting on them more rapidly than is currently possible. Meanwhile, as these tools are being developed, present ARPA Network procedures for supporting the dialogue in a distributed group in more traditional ways may prove helpful. The Network Information Center (NIC), in addition to supporting the general ARPANET community, is O'Sullivan [Page 7] RFC 313 Computer Based Instruction March 1972 supporting special interest groups such as the Speech Understanding Research (SUR) group. The application of these procedures could establish a valuable link between the academic-nonprofit institutions working on CBI, the centers in the Armed Forces where development and operations are taking place, and members of the network community who have an intimate understanding of the network resources available. CONCLUSION This paper has argued that there are resources in a large General Purpose Computer Network that can be applied to CBI with high utility. The argument can be extended to suggest that large dedicated CBI systems can have greater utility to users (and in the other direction, greater use), if tied into a General Purpose Computer Network, with respect to current network capabilities, future network developments, and in some cases provide backup during periods of overload or system failure. There are certainly important CBI issues outstanding in areas of pedagogy, strategy, curriculum development, testing, etc. As CBI systems are developed there are important issues of control (of the development process, of the distribution of material, and of modification of those materials). However, these issues seem to be independent of the question of whether CBI takes advantage of the resources of a large General Purpose Network. There are important problems to be solved on the computer side dealing with better tools to handle and evaluate masses of data, language, and protocols for network utilization. However, there seems to be sufficient promise in what we know of present network capabilities to warrant serious consideration by the developers of CBI of how General Purpose Networks fit in, and by network people of how their resources apply to this important large application area. [This RFC was put into machine readable form for entry] [into the online RFC archives by Hélène Morin, Viagénie 10/99] O'Sullivan [Page 8]