Electronics Kit PC Keyboard Yonju is now on sale

I am excited to announce the launch of my new product, the electronics kit PC keyboard Yonju.

It has tons of exotic features.

The electronics kit PC keyboard Yonju is now on sale!

How did Enter take over Return?

I feel that IBM in the 1980s was like a fable. What if all the big tech companies were just one company? IBM in 1985 was it. Eight years later, 1993, IBM fell behind Microsoft and Intel, and faced a management crisis.

Rome wasn't built in a day. The decline and fall too. This is the story of a sign of the decline and fall.

The story begins in 1971, IBM 3270 terminal. Photo:

What is the curving arrow key? Is this Enter key? No, this is "New Line" key. System/23 Datanaster manual:

When did IBM start to call the curving arrow key "Enter" key? In 1981, at IBM PC manual:

It seems like a poor choice because the same symbol has two different names among the same IBM products. The customers will be confused for no reason. How did it occur? Look at the IBM PC keyboard:

The keycaps and layout (except color and printing) are the same with System/23 Datamaster keyboard:

The IBM PC keyboard was made by modifying the System/23 Datamaster keyboard.

Look at the "Enter" key of System/23 Datamaster. Yes, it is printed "Enter." This is a strange situation. If the designer of the IBM PC intended to place "Enter" key there, it was already there! Why went out of his way to rewrite it and cause confusion?

Here is a reasonable explanation: The desginer intended to place "New Line" key there. He probably gave an order "Replace Enter with New Line" or something like that. The order was misunderstood. The operator replaced the symbol, not key! The error was overlooked and the IBM PC was shipped.

What I've told so far isn't about the decline and fall yet.

PC is an extremely complex product. Errors are inevitable. If it's just a matter of revising the manual, it won't cost much. But IBM didn't. This is a sign.

"New Line" key was a good idea because "Carriage Return" and "Line Feed" keys were an awkward legacy of teletypes. In contrast, the current Enter key is a poor idea. It still makes confusion about the semantics of Shift+Enter. In messengers, does it mean "Send" or "New Line"? It depends on which messenger is used. We are still caught up in the collapse of the empire.

Theoretical maximum speed of typing

It is not safe to say, yes, it is polemic to say that "QWERTY is enough good."  You see, we are keyboard enthusiasts, hahaha...

The rest of us never think such a thing, but if they were to come across the QWERTY skepticism, they will view that with suspicion, I guess. No one believes that QWERTY is the best of all possible keyboard layouts, but "QWERTY is not enough good" is an extraordinary claim for them.

Why is it? What rules are applied to determine that?

I naively used to think that it comes from the cost-effectiveness of optimizing keyboard layout. We can get some effectiveness by swapping a pair of keys. For example, QWERTZ and AZERTY are made from QWERTY by this way. Swapping one pair, two pairs, three pairs... The law of diminishing returns applies. When the effectivenss becomes lower than the cost of the next key pair swapping, it is enough good: This is the cost-effectiveness hypothesis.

The hypothesis isn't entirely invalid, but now, I tend to think that the bottleneck lies in our brains rather than our hands. The bottleneck determines a "passing mark" of keyboard layouts, I think.

The new hypothesis came from my experience first. I designed a novel Japanese kana layout two years ago. Theoretically it makes possible to type as fast as speak without hassles. But in reality, by typing such speed with the kana layout, I found that my brain is in a hassle instead of my hands! Now I'm leaning toward that typing faster than speech (without hassles, i.e. sustainable for a few minutes) is a special skill like tongue twisters. It is a good exercise but not a practical skill except for some specialists like voice actors.

Today I found a paper about the information throughput of a human, "The unbearable slowness of being: Why do we live at 10 bits/s?" (2025). I think that the paper suggests the new hypothesis. Typing speeds exceeding 10 bits/s (120 WPM) is a special skill like tongue twisters, and 14 bits/s (170 WPM) is a big hassle even for specialists. The difficulty is the same regardless of the keyboard layout used.

10 bits/s, 120 WPM without hassles, is a "passing mark" of keyboard layouts, I think. It is safe to say, under this rule, QWERTY is enough good.

About Hisao Yamada: "A Historical Study of Typewriters and Typing Methods" (1980)

I am not very keen to criticise a paper written in 1980. The paper is Hisao Yamada: "A Historical Study of Typewriters and Typing Methods: from the Position of Planning Japanese Parallels", Journal of Information Processing, 2(4) (February 1980), pp. 175–202. You can download full PDF freely, without any account.

The paper has a fair amount of influence. For example, Google Scholar finds 192 citation. It seems that in certain fields, it has become customary to cite this paper. So I feel that I should do what I should do.

My overall thoughts: The paper describes about 1980 well, but not very well about 1873-1979. In 1980, Dvorak layout (Dvorak Simplified Keyboard, DSK) seemed like a good option, far better than it really was in 1980. The paper describes why DSK seemed so good in 1980, but did not describe how good it actually was.

Actually I read the Japanese translation featured in a Japanese magazine. It contains a must-see information which is not included in the original English paper. About the part of "Dvorak, being naive on business affairs" in the paper:

After reading the original English paper, Mrs. Dvorak sent a letter, “I have some doubts about this part. The facts were actually the exact opposite.” (Bit, vol. 13, no. 10, pp. 1248)

I think there are probably far more people who agree with Mrs. Dvorak now.

The original English paper also contains a must-see information:

Malt also states that the long held belief of the superiority in speed of alternate hand keying over lateral keying (i.e., the stroking of adjacent keys by the same hand) has now been disproved, bringing back the belief of some early investigators such as Nelson (1920, 1921).

Malt is the Maltron's Malt. Maltron:

Needless to say, the belief that lateral keying is inferior stems from the mechanical limitations of manual typewriters. One of the design principles of DSK was this brief. In the age of good key rollover, lateral keying is superior. Malt studied and observed very well.

The author of the paper, Hisao Yamada, also pointed out a truth about chording keyboards. But this part is also not included in the original English paper:

Considering the history of chording typewriters, what is known today about stenotypes, and the fact that, in general, performing multiple tasks at once is much more difficult than performing a single task—it is difficult to argue that chording offer any advantages over sequential keystrokes except for very specific purposes. In fact, the opposite appears to be true. (Bit, vol. 13, no. 11, pp. 1554)

Although this remark was an implicit criticism of Fujitsu's Oyayubi Shihuto (親指シフト) in the context of that time, it remains a universal truth even today.

I hope you enjoyed reading this article, but I also hope you are not encouraged by this article. Truth is like the legal system in a normal society. If you make it your enemy, you’re in for trouble. If you want to make it your ally, you’re already in trouble.

keycap-designer Update (v0.1.7)

keycap-designer is now v0.1.7!

The biggest change: Now you can learn / write / execute it online!

keycap-designer is a Python-based open-source application to design keycap printing images. DecentKeyboards accepts the artifacts as complete printing-ready data.

Changes in v0.1.7:

• Now Pyodide ready. `safetensors` is replaced by `np.savez_compress()`. `pikepdf` is replaced by `pypdf`.
• Now Jupyter ready. `from keycap_designer.jupyter import *` and `show_inline([Manuscript])` etc.
• Now JupyterLite in Github Pages is available. See `jupyterlite/`.
• `trim` added to `Style`. Now you can print legends on margin area.

Abandoned idea: 3-axis joystick

Multi-touch touchpads are amazing. 4-axis absolute position input by two fingers! I think that gamepad thumbsticks are ridiculous. Why not use (single-touch) thumb touchpads? I saw a research paper about modern touchpads. As gaming M&K players prove, as it was during Doug Engelbart's research, mice are still the best option. Touchpads are not as good as mice (so there are quite a lot of mobile mice on the market), but clearly better than thumbsticks.

Bad news: Many of us already have touchpads for PC, i.e. laptops. Nothing to make a hardware.

Touchscreens are cheap and versatile. It's only natural that everyone uses it. So I'm interested in all the areas (except gaming) where joysticks are still being used. Look at this:

This is a CCTV controller, a.k.a. PTZ controller. The joystick operates Pan-Tilt-Zoom. You can twist the joystick to zoom. I am not well informed in this area, but there are many machines on the market. There must be a good reason why a joystick is needed.

The joystick module is available at Alibaba. 350 USD for 10 pcs (MOQ). 

The dimensions:

The minimum keyboard thickness is 30.7 mm. Bulky but manageable.

I looked for a good use for it, but couldn't find one.

Touchpads have two severe drawbacks: Susceptible to water and shock. Vehicles are a typical case. I don't want to ride an electric wheelchair with touchpad drive controls. In such cases, joysticks are the best. But my product isn't like that.

I think that CCTV monitor rooms are also free from water and shock. I couldn't find a good reason for using a joystick. It's a mystery. Static force feedback? Flat learning curve?

Abandoned idea: Motorized potentiometer

In old good days, the volume of TV / radio / amplifier was a potentiometer (rotary variable resistor). Remote controls already existed back then. How did a remote control move the volume? By combining a motor to the potentiometer:

Now motorized potentiometers can be found in exotic MIDI controllers.

I have seen rotary encoders on many enthusiast-grade keyboards. I am afraid that I feel they lack appeal. Without dedicated software running on PC, there are few good uses. Moreover, the rotary encoders are always incremental type. That is to say, the knob direction itself is irrelevant. I feel something essential is missing.

If dedicated software is unavoidable (such software is hard to maintain), why not motorize? It is not only an input device, but also an output device. I feel the interactivity is wonderful. Of course, it makes the knob direction relevant. So I investigated the area a bit.

The showstopper #1: Short life

The life of potentiometers is typically 1,000 to 30,000 cycles. Motorized products claim around 10,000 cycles typically. It is good performance as the volume of amplifiers. For TVs, it looks uncertain. For MIDI controllers, I feel much more uncertain. Many MIDI controllers feature motorised faders (linear variable resistors), and many people complain that motorised faders don’t last very long. The replacement repair of the component looks almost mandatory.

Rotary encoders have a much longer life because it has no sliding contact. Why not go motorized rotary encoder, absolute type? Because no one makes it. I hope it will be released someday.

The showstopper #2: Too bulky

This drawing shows the dimensions of a motorized potentiometer:

The dimensions are typical. It says that the keyboard should be thicker than 50.5 mm. I am too young to see such thick keyboards. Even if I place it on the enclosure's side, the height is 29.5 mm. Too bulky.

Motorized potentiometers and faders are fancy, but it is hard to bring the valuable product to life.