conanxin

Compiled from: Alan Kay's 1972 paper — A Personal Computer for Children of All Ages

Abstract#

This paper speculates on the emergence of personal, portable information manipulators and their impact on children and adults. While this should be understood as science fiction, the current trends in miniaturization and cost reduction almost guarantee that many of the concepts discussed will actually occur in the near future.

“To understand this world, you must build it.” — Pavese

For years, the tradition of trying to cure the ills of our society through technology has persisted: “Got slums? Let’s build low-cost housing!” “Can’t afford that TV? We’ll make a cheaper one that you can buy on time, even if it breaks before you finish paying for it!” “Your child isn’t learning, education is too expensive? We’ll build you a teaching machine that guarantees your child passes the tests!”

Unfortunately, most of these “cures” are merely painting over rust; the root causes of the original problems remain. Educational goals are further obscured by the various existing “end product” models: society needs more members (cultural genetics), parents may want success, conformity, prestige, or may not care; children are not asked what they need (they might just want to plant beans and watch them grow). What about teachers? Of course, there are enlightened ones among them (who have good role models and try to communicate what they are trying to convey and what the child’s current situation is), there are well-meaning individuals who want to teach (but lack talent), and there are those who treat teaching merely as a job, or worse, because “ed” is the easiest way to go to college, and now they feel dissatisfied with their fate due to their youth.

Technologists point out that through teaching machines, at least the lowest tier of teachers will be eliminated. What they rarely understand is that such teaching machines are best suited for the middle category: well-meaning but lacking talent! Can technology provide a machine with first-class teacher attributes? Perhaps. But first, it must be decided that doing so is a necessary and desirable goal.

In this brief statement, what we want to do is discuss some aspects of the learning process that we believe can be enhanced through technological media. Most of the roots of the ideas have many theories about children. We feel that a child is a “verb” rather than a “noun,” an actor rather than an object; he is not an enlarged pigeon or mouse; he is trying to obtain a model of the surrounding environment to cope with; his theory is about the “practical” concept of how to get from idea A to idea B, rather than the “consistent” branches of formal logic, and so on. We want to understand how he currently thinks in order to influence him, rather than merely trying to replace his model with our own.

We do not believe that technology is a necessary component of this process, just like books are not. However, it may provide us with a better “book,” an active (like a child) rather than a passive book. It may have the attention-grabbing functions of television, but it is controlled by the child rather than the network. It can be like a piano: (yes, a product of technology), but it can be a tool, a toy, a medium of expression, a source of endless joy and delight... and, like most gadgets in ignorant hands, it is a terrible burden!!

This new medium will not “save the world” from disaster. Like books, it brings a series of new vistas and a series of new problems. However, books have encapsulated and transmitted centuries of human knowledge to everyone; perhaps this active medium can also convey something thought-provoking and creative!

Two children sit on the grass playing with Dynabooks

Quick! Accompanied by beautiful flashes and appropriate noise, Jimmy's spaceship disintegrated; Beth has won the space war again. The nine-year-olds lie on the grass in a park near their home, their Dynabooks connected, each able to see Beth's spaceship floating alone in the space world.

“Do you want to play again?” Jimmy asked.

“No,” Beth said, “it’s too easy.”

“Well, in real space, you’ll be orbiting the sun. Of course, you won’t be able to win then!”

“Oh, really?” Beth was sparked into action. “How do we simulate the sun?”

“Um, let me see. When the spaceship is in space without the sun, it just keeps moving forward because there’s nothing to stop it. Every time we press the thrust button, your program increases speed in the direction the spaceship is pointing.”

“Yeah. That’s why you have to turn the spaceship around and push back to get it to go up.” She illustrated by manipulating some exercise buttons on her Dynabook. “But the sun makes things fall into it... that’s not quite the same.”

“But look, Beth,” Jimmy aimed at her spaceship, “when you hold down the thrust button, the spaceship starts going faster and faster, just like Mr. Jacobson said rocks and things move under gravity.”

“Oh, yes. It’s like there’s a jet pointing to the earth on the rock. Hey, how about that gives the spaceship more speed?”

“What do you mean?” Jimmy was confused.

“Look here.” Her fingers began to fly over the Dynabook keyboard, changing the program she had written a few weeks ago, which she had “accidentally” come across through Mr. Jacobson after being exposed to the space war game. “Just act as if the spaceship is pointing towards the sun and increasing speed!” As she spoke, her spaceship began to descend, but not towards the sun. “Oh no! The sun is moving!”

Jimmy saw something was wrong. “Wherever your spaceship is, you need to increase speed in the direction towards the sun.”

“But how do we do that? Oh my!!”

“Let’s go ask Mr. Jacobson!” They picked up their Dynabooks and ran across the grass to their teacher, who was helping other members of the group solve problems.

Mr. Jacobson's eyes sparkled as the students were eager to learn. The students still thirsted for knowledge like two-year-olds. He and others like him would do their utmost to maintain the curiosity and thirst for knowledge that is every person's birthright.

From the words that spilled out of Beth and Jimmy, Mr. Jacobson could see that the children had intuitively rediscovered an important idea and only needed a little prompting to add the sun to their private universe. He was enthusiastic but somewhat noncommittal:

“That’s great! I bet the library has what you need.” At that moment, Jimmy connected his Dynabook to the class LIBLINK and inherited the thoughts and knowledge of past eras, all of which could be read through his Dynabook screen. It was like embarking on an endless journey through boundless space. As usual, he had a hard time remembering what his original purpose was. Whenever he found something interesting, he would send a copy to his Dynabook so he could look at it later. Finally, Beth poked him in the ribs, and he began to search more seriously for what they needed. He wrote a simple filter for his Dynabook to help them search...

Just as Beth and Jimmy were struggling to discover the concept of coordinates, Beth's father was sitting on a plane preparing for an important meeting. He was carefully studying the relevant background facts that he had extracted from his business files into his Dynabook that morning, stopping every so often to input voice annotations. He knew that not inputting his comments was untimely (Miss Jones had to do that too), and he eagerly hoped to add the long-promised voice recognition capability to his Dynabook. After landing, his eyes were drawn to a sensational poster on the airport's Storyvend. He connected his Dynabook to Storyvend “just to see” if the heroine really had “creativity.” She did, and when he pressed the copy key on the Dynabook (Alice would never know), Storyvend reminded him that he forgot to pay the copying fee (COPY).

He entered the taxi with a more pragmatic mindset and decided to verify the opposition's estimates. When he scanned the information with his Dynabook, he thought this was something he wouldn’t have done five years ago; doing it by hand or passing it to someone else was too cumbersome. Moreover, he had just thought of a new way to view their data on the plane.

At this point, Beth had discovered that if the sun was placed at the “zero” position, her problem would become extremely simple; she just needed to subtract a little from the spaceship's “horizontal” and “vertical” speeds based on the position of the spaceship. All the drawings and animations that she and the other children had completed before were achieved by using relevant concepts that matched their then capabilities. She was now ready to hold several independent ideas in her mind. The children's intuitive grasp of linear and non-linear concepts would become a wealth for their later understanding of great science.

After her spaceship completed its mission, she found Jimmy engrossed in his Dynabook, then thoroughly defeated him until she felt bored. When Jimmy went to find a less terrifying enemy, Beth retrieved a poem she had been writing on her Dynabook and edited a few lines to improve...

Current technology has made it possible for everyone like Beth and her dad to have access to a “DynaBook” anytime and anywhere. Although it can be used to communicate with others through future “knowledge tools,” such as school “libraries” (or commercial information systems), we believe that most of its use will be reflective communication that DynaBook owners have with themselves through this personal medium, just like the paper and notebooks currently in use.

Tools are things that assist in manipulating media, and humans are called “tool-making animals.” Computers are also seen as tools by many. Clearly, books are not just tools, and humans are not just tool makers... they are inventors of the universe. From the moment they learn to observe and use language, every new universe is a medium. Usually with the help of tools, the structures of imagination can be embedded (constrained) in expression. What about computers? They are clearly not just tools, although much of their content, in typical McLuhan style, absorbs previous media, their own attributes are just beginning to be discovered.

So what is a personal computer? People hope it is a medium that contains and expresses arbitrary symbolic concepts, as well as a collection of useful tools for manipulating those structures, and a way to add new tools to the computer instruction system. Another rarely mentioned constraint is that it should be superior to books and printing in at least some aspects, while not having obvious disadvantages in others. (Previous comments seem to disallow consideration of known commercial display devices.) “Personal” also refers to being owned by its user (costing no more than a television) and portable (to me, this means the user can conveniently carry this device along with other things). Should we add that it can be used in the woods?

“Before you learn to think, you must learn to think well. In retrospect, this proves too difficult.” — A. France

Recently, researchers in artificial intelligence and (to some extent) education have begun to study how children acquire their world models. (It was once thought that intelligent behavior could be simulated through non-anthropomorphic means.) Led by Newell and Simon, Papert and Minsky, Moore and Andersen, many are now interested in how children and adults acquire and manipulate human knowledge. Particularly interesting are theories of early development and model building, which have been completed by experts like Piaget, Bruner, Hunt, Kagan, and others who study children at different stages of development.

Another closely related group is interested in discovering what children of different maturities are truly capable of doing. We must mention Montessori, who was among the first to discover that children are much better at learning at an early age (2-5 years) than is usually imagined. O.K. Moore demonstrated through a reactive environment that even very young children can learn to read, write, and abstract. Shinichi Suzuki successfully taught thousands of children aged 3 to 6 to play the violin. Research by Bruner and Kagan shows that children have the ability to visually discriminate and generalize even in their first year (or month) of life, far exceeding previous assumptions.

The work and ideas of O.K. Moore and Seymour Papert particularly influenced the emergence of the DynaBook concept. Both believed that children are active agents, creators, and explorers, and are intellectually much stronger than generally thought.

Some principles of Moore's “talking typewriter” are worth studying. He believed that rather than children lacking long attention spans, they find it difficult to remain rational in thought or activity. For one idea, playing the role of a “patient listener” quickly becomes boring and attention-deficient unless other roles can also be played, such as “catalyst,” “referee,” or “player,” etc. An environment that allows very different viewpoints to be accepted is very conducive to children's activities of differentiation, abstraction, and synthesis.

A “safe and secluded” environment is also an important part, where children can play almost any role without social or physical harm. Although skills and knowledge are occasionally rigorously tested in front of peers and adults, there must also be absolutely safe times to “improvise” without blame. In Moore's words, a “productive” environment is one where what is learned can be used as part of new ideas (for further learning). Finally, an environment that can immediately respond to a child's activities and allow him to acquire his own models is extremely important.

The “talking typewriter” is the crystallization of these ideas, turning into a device (initially simulated by a graduate student behind a wall) that provides many beautiful insights into the abilities and interests of small children.

“Should computers be programmed by children, or should children program computers?” — S. Papert

In the process of “teaching children to think,” Papert, by giving them an environment where they can write programs for their own purposes (animation, games, etc.), is astonishingly similar to Moore, although the philosophical background is artificial intelligence and Piaget.

The LOGO language is used through terminals (via time-sharing systems), allowing children's programs to control text, graphics, music, and the cumbersome “turtle.” Papert's LOGO work is “CAI” only when the acronym represents computer-assisted intuition (or inspiration) rather than indicating instructions. However, much of the current computer-related education is based on programmed learning, which largely stems from behaviorists' experiments with rats and pigeons. On the other hand, Papert's perspective is highly influenced by his contact with Piaget and his research (strangely enough), which primarily comes from studying real children and how they view the world.

Our project aligns well with the latter perspective. While some people measure progress in terms of “answers-correct/test” or “tests-passed/year,” we are more interested in “Sistine-Chapel-Ceilings/Lifetime.” This is not to say that skill achievement has not been valued. Without the superb skills of dreaming and depicting those dreams, “Sistine-Chapel-Ceilings” would not be realized. As observer Finch commented, “Where the spirit does not work with the hand, there is no art.” Papert points out that people willingly spend thousands of hours perfecting the sports they participate in (like skiing). It is clear that school and learning are not very interesting for children, and there is no way to immediately derive enjoyment from practical knowledge skills.

With Dewey, Piaget, and Papert, we believe that children “learn from practice,” and much of the alienation in modern education stems from the vast philosophical distance between what children can “do” and many adult behaviors of the 20th century. Unlike African children playing with bows and arrows, which engage them in future adult activities, American children either indulge in trivial imitation (children in nurse costumes caring for dolls) or are forced to participate in activities that will yield no results for years, which alienates them (math: “Multiplication is good for you—look, you can solve problems in the book”; music: “Practice the violin, and three years later we might tell you something about music”; etc.).

If we want children to learn any specific field, it is clear that we should provide them with something real and enjoyable on their path to artistic and skill perfection. Painting may be frustrating, but practice is fun because a completed painting is a sub-goal that can be achieved without mastering the entire subject.

Unfortunately, playing an instrument and acquiring musical thinking is far from that. Most modern keyboard and orchestral instruments do not provide satisfying sub-goals for children or adults over several months, nor do they truly help them understand what music is or how to “make” music themselves. This is often more akin to “practicing and skills” for painting billboards by “pushing numbers,” without even using their own numbers or paints!

In general, the situation for studying arithmetic and mathematics is worse. What can a child “do” with multiplication? The usual answer is to complete the problems in the math book! A typical reaction to this is, “Some things can only be learned through practice.” (Fortunately, in this case, children do not have to learn their mother tongue.) Papert's children need to use multiplication to change the size of the animations drawn by their computers. They relate to this matter.

Epistemology#

Jean Piaget's life work is both broad and profound enough to disdain any rough summary. Because there are summaries and commentaries (such as Furth: Piaget and Knowledge: Theoretical Foundations), a more selective strategy needs to be adopted.

From the perspective of computer scientists, two fundamental concepts of Piaget are appealing.

First, knowledge, especially that of young children, is preserved as a series of operational models, each of which is temporary and does not need to be logically consistent with other models. (They are essentially algorithms and strategies, not logical axioms, judgments, and theorems.) Logic is used in development, even through non-logical strategies.

The second concept is that development occurs in a series of stages (which seem to be independent of cultural context), each stage building on the past but exhibiting enormous differences in the ability to understand, generalize, and predict incidental relationships. Although the age at which a stage is reached may vary among children, the apparent dependency of one stage on previous stages seems invariant. Another important point later is that language seems not to be the master of thought but rather a servant, as Piaget and others have ample evidence that thinking is non-verbal and imagistic.

a. Stages#

Piaget and Bruner both created names for developmental stages. Bruner's are more descriptive, so they are included here.

If stage dependency is real, then if attempts are made to force knowledge from the previous stage onto children before they are ready, it may be worse than useless. For example, the current popular practice is to teach children (in “new math”) topology of point sets in two-dimensional Cartesian coordinate systems at as early an age as possible. A series of experiments by Piaget showed that children in the operational stage do not grasp the concept of coordinates until later, which contradicts the above practice. However, they do have very complex concepts of topology, connectivity, attachment, and grouping—all relevant concepts. Papert and Goldstein use these facts to teach geometry and topology without reference to global coordinate systems—a more satisfying state.

If we believe in the accuracy of “operational” (semantic) models rather than the accuracy of “predictive” (logical, syntactic) models, then we must argue with the most popular syntactic concepts in current “new math.” For example, in natural numbers:

“3 + 5”
“4 + 4”
“16 - 8”
“4 * 2”
“8”

It is said to be the “numerical symbol” of the number 8.

This concept is not only misleading and absurd but also incorrect. (What is the number for “8 / 3”?)

Minsky points out: “The problem with new math is that you have to understand it every time you use it.”

The work of Piaget and others on the foundations and forms of children's thinking provides a fairly compelling argument that computers are almost an ideal medium for expressing children's epistemology. What is an “operational model” if not an algorithm for achieving goals? Algorithms are quite informal and do not necessarily have to be logically consistent (anyone who has ever spent hours debugging a program knows this well). This aligns with children's perspectives, which are global and interested in structure rather than strictly “truth.” On the other hand, computers also help form skills related to “thinking”: strategies and tactics, planning, observing causal chains, debugging, and refining, etc. A child rarely has the opportunity to practice these skills in a patient, secluded, and interesting environment!

DynaBook#

“I wish these calculations were done in a pipeline.” — Charles Babbage (19 years old)
“The analytical machine arranges algebraic patterns just as a Jacquard loom weaves patterns in silk.” — Ada Augusta, Countess of Lovelace

We now have some reasons to hope for the existence of the DynaBook. Can it be made from currently invented technology? Is there enough quantity to sell (or lease) at a price to millions of potential users? A set of considerations related to the practicality of the device (size, cost, capability, etc.) is as important as the more profound philosophy that initially prompted us to think. The following pages discuss some relevant trade-offs and attempt to convince the reader that a target price of $500 is not entirely outrageous. Current cost trends and the sizes of various components indeed bring considerable hope for achieving this goal. It is also important to remember the analogy with color televisions priced below $500. Now, what should the DynaBook be?

The size should not exceed that of a notebook; weight should be less than 4 pounds. The visual display should be able to present at least 4000 printed quality characters, with contrast close to that of a book; reasonably good quality dynamic graphics should be possible; there should be at least one million characters (about 500 pages of a regular book) of movable local file storage, with the ability to exchange several hours of audio (voice/music) files.

The active interface should be a language that uses concepts similar to those of the device owner. The device owner should be able to maintain and edit his text and program files anytime and anywhere. He can use his DynaBook as a terminal while working (or as a connection to the library system while at school). When he has read through and discovered the information he wishes to extract and carry with him, it can quickly transfer the information to his local file storage. The tether connection can not only provide information but also give the device extra power, while the central connection can provide information for any motors the device may have, allowing for approximately 300K bits/second high bandwidth transfer to file storage, or transferring 1500 pages of a book in 1/2 minute. During this connection process, the battery will also automatically recharge.

“Books” can now be “instantiated” rather than purchased or checked out. One can imagine vending machines allowing reading information (from encyclopedias to the latest adventures of whimsical women), but preventing file extraction until payment is made. The ability to easily copy and “own” personal information may not undermine existing markets, just as simple electrostatic copying enhanced the development of publishing (rather than harming it as some predicted), and the advent of tapes did not damage the record industry but provided a way to organize personal music. Most people are not interested in acting as pirates; rather, they enjoy exchanging and playing with what they own.

This combination of a “portable” device and a global public information facility like the ARPA network or two-way cable television will bring libraries and schools (not to mention stores and billboards) or the whole world home. One can imagine that the first programs written by device owners will include filters to eliminate advertisements!

Input will be through a keyboard (which most people have now learned to type) or through a traditional secretary-like keyboard. Or via voice. The device's file system can easily allow audio files (with digital titles); however, they must be transcribed before any editing. While “interactive graphics” will be limited due to capacity, sketches can be retained and edited as fax documents.

Display#

Whether it is a flat display, such as a plasma panel, or an external CRT connection, it is determined by size requirements. Power specifications do not allow plasma panels (which require 5 amps of current when fully lit), and the demand for using it anywhere eliminates almost (but not completely) the ubiquitous CRT. So what remains? We clearly want a technology that only requires power to change state, not for viewing, meaning it can be read in ambient light. Phase Transition Liquid Crystal could be x-y coordinates, becoming opaque under the influence of a low-power electric field. Furthermore, the display will maintain itself with very little additional power, and the electrode width can be as small as 1 mil, with the state of an entire 512x512 panel changing in less than 1/2 watt. (Note: This is a current technology, although no one has yet made a 512x512 panel.)

To display book-quality characters at normal viewing distances, we need a good eye model and to leverage our laboratory's latest findings in character generation art. To establish an internal research terminal with CRT display quality, an experimental “loadable character generator” was designed and built. Any font of 128 characters can be viewed in matrices of up to 32x32 bits, dynamically loaded into fast dual-phase memory to allow real-time scanning conversion of ASCII text. Also provided are embellishments such as size, intensity, and covering characters (underscores, etc.). The photo is of the actual screen (875 scan lines), unembellished.

The first interesting finding is that the display looks much better than it “should,” meaning these characters appear much rounder than those displayed at the digitization level; however, when they are enlarged, they quickly become ugly. The intuitive reason for this phenomenon relates to the inherent noise-filtering function of the optical system, which essentially averages the small corners into a blurred signal (using an average window of about 0.02 radians) and then differentiates over a larger area, re-adjusting the scene into a clear image. The role of this filter is to eliminate small isolated spikes; fortunately, when the matrix is small, it allows the characters defined by the matrix to look nice. This also partially explains why 875-line television appears subjectively to be more than twice as good as 525-line television at a 22-inch viewing distance. The scan lines and their spacing are too large to filter 525, as they are about 1/50 inch high.

... Characters are difficult because the defining matrix is limited, but what can be done is more obvious. Two very effective tricks are to change the aspect ratio of the characters (height: width ~2:1, thus turning 45 degrees into 30 degrees) and even use multiple stroke widths on very small characters to achieve a bold appearance, even on very tiny characters (this deceives the eye's filter, trying to enhance the character rather than remove it as noise).

In summary, the display surface should be liquid crystal, with at least 80-100 raster points/inch, with an aspect ratio of about 2 points for each point in the vertical direction, and a total raster number of about 1024 x1024.

Keyboard#

Of course, the keyboard should be as thin as possible. It may have no moving parts but should be pressure-sensitive, providing feedback through speakers when successfully pressed. Such keyboards have existed for several years. Once a person is accustomed to the idea of having no moving parts, he is ready for the idea of having no keyboard at all!

Assuming the display panel covers the entire range of the notebook surface. One might hope that any keyboard layout could be displayed anywhere on the surface.

Four strain gauges installed under the corners of the panel will record the position of any touch, within a close range of 3/16 inches. The bottom of the display panel can be textured in various ways to allow touch typing. This arrangement allows a font of an input to be displayed on the keys, special characters can be windowed, and user identifiers can be selected with a single touch.

File Storage#

The only existing technology that can handle the moderate but important demand for writable file storage is the use of magnetic oxides on plastic in the form of tape or floppy disks. Until recently, tape handling typically required a collection of pressure rollers, spools, coils, and motors.

Now, some companies have solved the problems of constant tape tension and differential drive, most succinctly with 3M's tape cartridge, which uses a “magical” drive mechanism that contacts the outer side of the tape reel, requiring only one motor to read, write, search, and rewind. A four-track tape with a bit density of 1600 BPI allows storage and retrieval of 6400 bits/inch. Therefore, our requirement for 8M bits would need 1250 inches (or 105 feet) of tape in the cartridge. Of course, there will be gaps, etc. To be safe, our fantasy will add 50% more tape or 150 feet of tape. The file directory will be placed in the middle of the tape, so accessing it will average only 1/4 of the tape traversal time. From there, the average distance to any file is also merely 1/4 of the tape length, resulting in an average random access time of 1/2 tape traversal time. Search speed depends almost entirely on the required battery consumption rate and motor capacity. The 3M tape cartridge can position at 180 inches/second; about 7 seconds can traverse 100 feet of tape, so the average file access delay is about 4 seconds. This is very respectable. However, when using batteries alone, these speeds require too many watts. When using batteries, a more reasonable search rate is 60 inches/second, with an access delay of about 10 seconds.

Floppy disks require two motors (one is a stepper motor for positioning the read/write head), which usually run continuously. The latter cannot be operated on batteries, and the device must start and stop. One significant advantage of floppy disks is that they can be swapped on one track while still allowing proper access to files. (The concept and utility of swapping storage will be discussed in the processor section.)

Processor and Storage#

These two categories represent the cheapest and most expensive components in our fantasy machine, respectively. Because processors have a significant impact on the required main memory, they appear together.

The following attempts indicate that performance and packaging requirements are not necessarily incompatible with current technology (although sometimes they are). Just like the HP-35 pocket electronic “calculator,” our main savior is inexpensive LSI components. The HP-35 uses five LSI chips, equivalent to 30,000 transistors, with an average density of 6,000 transistors/chip. Now, better packaging densities are being achieved. The price of a packaged LSI chip seems to approach $12 within two years, then may suddenly drop to about $5.

A complete CPU can now be used on a single chip. The current challenge is more about determining what characteristics the processor should have rather than just using anything well-packaged.

LSI random access memory is now typically available in 10241 bit chips (700 ns cycle time), packaged at 1¢/bit. A 40961 chip has now been released, appearing to be packageable at 35¢/bit. Therefore, the cost of 8K*16 memory is about $460 (still expensive but encouraging).

With the advent of portable shavers, tape recorders, toothbrushes, televisions, etc., the technology level of rechargeable batteries has greatly improved. We may expect higher performance/cost in the future.

Since about 20 integrated circuits are currently estimated to be the number of chips required for the DynaBook, we can reasonably determine that the electronic part of the device will be very well packaged.

The processor is envisioned to implement 1 or at most 2 LSI chips. Such devices already exist, priced at less than $100, and are expected to be below $15. They typically contain thousands of transistors; registers for program counters, arithmetic operations, instruction return stacks, etc.; and may even use carry-lookahead arithmetic units. An independent “smart terminal” (including memory, keyboard, display, and two tapes) using one of the chips for the processor currently costs about $6,000 on the market.

Since the DynaBook is not just a terminal, the cost is much lower, so a lot of effort needs to be spent on processor memory design. Clearly, we want to maximize the use of expensive core-replacement RAM; this can be achieved by:

1. Effectively encoding operators for maximum instruction density/bit.
2. Encoding basic logical data elements (ordered sets) to minimize space requirements.
3. Removing any system routines from RAM (including the interpreter) so that users can use all the space.
4. Mapping the virtual address space onto file devices so that RAM acts as a cache for the most recently used memory portion. (Doubting Thomases may find this worthless on tape drives; they are advised to consult LINC literature for descriptions of similar schemes that have been successfully used by thousands of users for years.)
5. Eliminating the need for a resident “system” itself by merging the concepts of files with user variables, allowing users to interact directly with the interpreter, and allowing interrupts through the use of multiple control path evaluators.

“Medieval thought is not limited, but perhaps its vocabulary is limited.” — Williams

What kind of way should various potential users communicate with themselves through their machines? A language that provides “everything” to everyone is clearly impossible. In the usual sense, neither is “scalable language.” Given that these two factors are excluded (by definition, they are), what remains is to present the user with a very simple language (which reveals the true state of programming semantics), yet this language has a variety of expressions. So what impact does the computer have on other information systems? On one hand, messages can be indefinitely delayed (memory), information can be transformed into other messages (processing), and they can represent the transformation itself as a message (process).

The use of this language essentially consists of two activities: 1. naming objects and classes (memory association), and 2. retrieving them by providing the names of previously stored objects and classes. A process consists of these (activities), and when no names are under scrutiny anymore, the process terminates. While all such languages can easily be derived from these two concepts, to allow interesting things to be done immediately, some names will have a priori significance.

The following principles should be used in the design of the DynaBook language:

1. We need a unified concept to understand what objects are, how to reference them, and how they manipulate other objects.
2. If each object can have its own control path, then there must be a concise way to coordinate and “control” these paths when multiple objects are active.
3. The evaluation of control paths should follow simple rules that show how objects pass messages and return results.
4. Each object in the system should be redefined based on other objects.

The basic idea is to leverage the duality between functions and tables (or processes and memory). In English, there are nouns referring to “objects” and verbs referring to “actors” and “relational words.” This is Newtonian epistemology. Modern physics and philosophy tend to view “objects” and “actors” as merely different aspects of process concepts. A process has a state (a set of relationships associated with it), and over time (defined as interaction with other objects), the state changes. Using this perspective, “data” is a “slowly” changing process, while “function” is a more rapidly changing process. Each process has the logical properties of a complete “micro” computer: they can have inputs, feedback outputs, act as memory on file systems, perform computations, be interrupted, etc. Because a “computer” can simulate all other computers (modulo time and space), using one language to describe the concept of a process can yield useful ideas such as arrays, records, recursive processes, etc., which can be added to the computer instruction system at any time.

The technology to directly evaluate this language through hardware is well known and achievable by single-chip processors.

The concept of multiple control paths allows independent concepts such as “files,” “operating systems,” “displays,” etc., to be replaced by a single idea that the user is also a process (thus having a state composed of variables and bindings, etc.). When he leaves the machine, his process is suspended until the next time he rejoins his DynaBook. When he is not there, his state (now activated) constitutes a “file.” By executing user input directly without any additional mechanisms (the “direct” mode of JOSS, LISP, etc.), control over the evaluation of various programs can also be achieved. Since multiple control paths are allowed, many processes can be at various stages of evaluation and debugging.

Size and Cost#

The evaluators' experiences we discussed earlier indicate that about 8000 bits of control memory are needed to implement the hardware. This memory currently requires 1 ROM LSI chip and another processor. Assuming both can be combined in one package without affecting the current level of technology, the realization of this idea is not far-fetched. The price of LSI packaging tends to approach $12 - 14 per package, as most of the manufacturer's costs come from testing, die, filling, etc., all of which are relatively independent of device complexity (as long as the yield is reasonable).

Intelligent encoding of “data” and “code” can reduce the memory required to save equivalent structures in languages like BBN-LISP by more than three times. This means that 8K 16-bit RAM is roughly equivalent to 12K 36-bit words of BBN-LISP on a PDP-10.

The DynaBook computer can now be assumed to be a bus machine, including:

1 processor chip
16 (8K*1) RAM memory chips
4 IO controllers (also including the processor chip—why not?)
21 chips (electronic device cost 14~294 dollars).

Due to the influence of science fiction and handmade creations, this price has little credibility. However, some brave readers may find it laughably high, not just ridiculous!

Conclusion#

Committed to speculation and fantasy, most readers may agree that the content of the previous pages is merely conveyed (with some credibility...).

We do believe that this thing has undeniable advantages for teaching algorithmic thinking, ease of editing, etc. (all contained in an environment that can go anywhere and belong to anyone). Considerations for packaging, power, and weight come from current technology and electronics, which may also be real. Software knowledge, language design concepts, and user interface ideas have been at least 5 years in the making. Three unreliable speculations are about flat low-power displays (which currently do not exist but seem possible), guesses about how much work can be done “independently” on an 8K machine (not yet simulated), and pricing.

Assuming the DynaBook can be sold for $500 (ridiculously low compared to current mini-computers, ridiculously high compared to current television technology); where is the money for most children (and adults) to have such a machine? The average annual expenditure for all this education is only $850/child. Some people care about the high-quality character generation because about $90-95 per year of student funding is used to purchase and maintain school books. If the DynaBook can take on this function over its lifetime (at least 40 months), then about $300 can be used. Perhaps the device itself should be given away along with a loose-leaf notebook, with only the content (tapes, files, etc.) for sale. This is spiritually similar to the current distribution of packaged television or music.

We deliberately do not quarrel with those who believe that sharing resources is most beneficial for living. The analogy with books still holds: libraries are very useful, but people do not want to endure the library's schedule, location (or content) 100% of the time. As Larry Roberts suggested, how about terminals via radio? Well, suitable for inverse large matrices, but not for graphic animations or any other high-bandwidth output. No need to elaborate.

Let’s get started now!