to computing, the concept of compiling prior to execution will be a new one. Compiling tends to encourage the adoption of language conventions aimed at easing the compilation task at a cost of user convenience. For example, many languages use "counter" items to code user response paths because they can more easily be used in later decision points. Many aspects of PLANIT take on a definition only after the student responds (too late for efficient compilation). 2. Compute times are usually greater for interpreted programs than for compiled programs, ranging from a little greater for character shuffling to more than a hundred times greater for number crunching. However, much of CAI is character shuffling and compute times are characteristically small even for interpreters as was evidenced in the Purdue statistics. 3. Space will usually be of more concern in CAI applications than time. Since a compiled program grows in size according to the number of source statements while an interpreted program usually operates in a fixed size, there will be a point beyond which the compiled program will be the largest. Source code is normally more compact than compiled code. CAI programs are typically large (as programs go) and will pass that point very quickly. Therefore, interpretive CAI systems will ordinarily occupy less total space than compiling ones - this added to the fact that most compiling systems also retain the original source code for editing purposes. In general, space will probably be more costly to CAI than compute time and interpreting systems will normally require substantially less space. The PLATO terminal has the microfiche projection capability to alleviate this problem but raising the preparation costs somewhat. Let me now turn to the second part and offer some comments that attempt to put PLANIT in perspective with two well-known elegant CAI systems, PLATO and TICCIT. Today's CAI users are fortunate to have them as options. PLANIT is also an option. The three systems can be summarized in this way: 1. PLATO is for the person who can have everything and has the money to pay for it. 2. TICCIT is for the person who has nothing and wants a lot but has little money to pay for it. 3. PLANIT is for the person who has equipment and needs to get along on what he has because he has virtually no extra money for CAI. For some time PLANIT was considered to be an interim system, to be used until PLATO and TICCIT were ready. Support required for PLANIT has been a fraction of that for the other two. No hardware development efforts were involved since PLANIT is completely software, at least up until the time of installation. Thus, PLANIT was less expensive to develop and more quickly delivered, making it a likely candidate for interim needs. However, it is becoming increasingly apparent that PLANIT's portability is providing an option not yet available in another CAI system. PLANIT can be mounted on existing equipment with little or no extra hardware investment. PLANIT lessons on all such installations are fully compatible and can be exchanged freely. Experience at operating versions of PLANIT which have been installed on widely differing hardware show no discernable differences to the user. It is a striking experience to sit at the keyboard of a strange system where PLANIT is mounted and immediately be completely familiar with the entire operation. This has been particularly useful to the military with their diversity of hardware. The ability to produce a fully compatible system on existing hardware at nominal cost suggests something more than an interim role for PLANIT. It may well continue to be a viable option for some time to come. When discussing PLANIT's future, the question is inevitably asked, "Can PLANIT handle graphics?" The answer is "yes" but the implementation of that is probably 95 percent installation hardware and software and 5 percent PLANIT additions. "Holes" have been intentionally left in PLANIT's command structure to allow for these kinds of additions. In one experiment with graphics, a Rand tablet was used for a PLANIT terminal with the display projected onto the under side, giving the impression of "inking" a surface with an electronic pen. It has also been observed that the PLATO plasma terminals would make nice PLANIT terminals. These kinds of questions are decided at installation time. Probably the most significant recent development in CAI is that we are beginning to have some options - not just the name of the language but the kind of system. Formerly, we had only one option - invent our own unique system. Now there are several more, such as PLATO, TICCIT, and PLANIT. If this progress is to be sustained, then interested parties should feel obliged to see what is available before concluding that nothing currently exists. Articles on CAI appearing within only the past six months in respected national periodicals show that this has not yet occurred. One lists PLANIT and TUTOR (PLATO's author language) among others as "large and complicated and troublesome to learn" and then proceeds to describe a language of dubious improvement bound to specific hardware. Another describes yet a different language in which the technique of prompting the author for lesson inputs is ostensibly "discovered," not mentioning that PLANIT has been doing this for eight years in addition to several others that can also be named. It was especially interesting that after discussing the remarkable gains in efficiency which were attributed to prompting, the authors of the article then proceeded to describe the soon-to-be-released version II which will allow batch input. What we need are real, legitimate options in CAI, not contrived ones. PLANIT is one of these legitimate options.