I think the Kinesis Advantage is a very bad keyboard. Particularly, although the key switches (Cherry Browns) are fine, the way they are arranged relative to each other has very bad ergonomics, and any of the many non-Kinesis boards in which Cherry Brown key switches are arranged in a single plane has much better ergonomics. The reason the Advantage is bad is that when every key has a different "altitude" (height relative to the desk or floor) and a different orientation (the direction in which the center-line or "axis of travel" of the key points) it becomes much harder for the person to memorize the location of the keys and in particular much harder to memorize the "vector direction" (i.e., distance and direction) between every pair of keys and the vector direction in which the person's hand would need to move to strike every key perpendicularly.

I used a Kinesis Advantage (well, a Kinesis Essential, which has the same Cherry Brown switches and the same arrangement of keys into two "bowls", one for the left and and one for the right hand) as my only keyboard for about 2 years. I bought it to try to get rid of my RSI, but my RSI continued until I was shown (by a piano teacher) the Taubman technique by which piano players avoid RSI, which is when I realized that to get rid of my RSI I would have to stop typing or I would have to memorize the locations of all of the keys on the keyboard. (Actually, certain keys, like backquote, right bracket, left bracket, backslash and backspace, I do not have memorized. I look at the keyboard when using those keys.)

The Kinesis Advantage is a consequence of thinking of typing as a sort of "sign language" to be perceived and understood by the machine. That is IMHO the wrong way to think about typing. The correct way IMHO is to think of typing as a long series of physical actions. Imagine you have a plunger in front of you. I.e., a 3-inch-diameter disk of plastic attached to a beefy value or switch. Imagine that you need to activate the plunger reliably -- to minimize the chances that your first "activation attempt" will fail. So now you have some "perceptual" / "non-cognitive" decisions to make, namely, will you have to get your shoulder or elbow in a different position in order for your hand to contact the plunger cleanly and perpendicular to its axis of travel? How fast should your hand be moving when it meets the disk? Do you want to take a "backswing" before your "swing" for the plunger? (I.e., move your hand first in the direction opposite to the direction it will need to move to activate the plunger.) Good typing is more like that than it is like making signs in the air the way that users of sign language do. (The DataHand keyboard takes the typing-as-sign-language metaphor even further than the Kinesis Advantage does, and the DataHand is an even worse keyboard than the Advantage is. Here is a pic of the DataHand: http://www.atpm.com/7.05/images/dh-handinunit.jpg.)

These comments on the Kinesis Advantage can be tied with the topic at hand. The reason the Kinesis Advantage is praised by so many programmers IMHO is that those programmers do not realize or understand the need for a keyboardist to memorized the locations, etc, of all the keys. That memorization is a perceptual task rather than a "high-level cognitive" task such as planning a complicated sequence of movements or operations or solving an algebra problem. Similarly, using a pointing device is a perceptual task. (More precisely, it is much more perceptual than the alternative ways of achieving the same result using the keyboard.) In both cases, we have large vocal populations of programmers (namely, Kinesis Advantage lovers and lovers of keyboard-centric tools like Vim, Emacs and Xmonad) who prefer to substitute cognitive tasks for perceptual tasks whenever possible. I suspect that that preference is a mistake and that those programmers would be better off in the long run if they tried to reduce the fraction of time they spend in high-level cognitive tasks. A possible exception is programmers whose frontal lobes are still developing, i.e., programmers under 22 or so, but certainly any programmer over the age of, say, 25 who prefers a high-level cognitive task when a perceptual task would work just as well to achieve the goal is IMHO misusing his or her brain more likely than not. (I am almost sure it would be misuing my brain for me to do so. I am much less confident about other people although I am confident enough to use this comment to share my suspicions.) Often there is no alternative to engaging in high-level cognitive work -- it is the only way to achieve some goal -- and in that case I am all for it if the goal is important enough.

I think it is important for people expressing an opinion about a tool to state how familiar they are with the tool. Consequently, let me state that I used text-mode Emacs ("emacs -nw") every day for 18 years. When I finally switched to an editor that allowed me to position the cursor (and do other things) with the pointing device, I realized I should have made the switch much earlier. Parenthetically, the editor I switched to was graphical-mode Emacs, but I customized it to extend the range of tasks that could be done with the pointing device, and I temporarily disabled the arrow keys to train myself to use the pointing device to move the cursor. (In fact, I permanently bound up arrow and down arrow to commands that scroll about one-eighth of a screenful, which is similar to what what up arrow and down arrow do in Plan 9 and even more similar to what they do in Firefox.) Also, I would probably have switched to Textmate, Sublime Text 2 or Gedit if I had not already invested a lot of time and energy into Emacs, including the writing of 5000 lines of Emacs Lisp code to customize Emacs. I would probably use some keyboard shortcuts if where using Textmate, Sublime Text 2 or Gedit.

Why would it be easier to remember how to hit keys on a flat 2d plane? My fingers are all different lengths.

If what I have been saying in this comment section has any value, then most people have already committed to muscle memory how long each of their fingers is relative to any other finger on the same hand.

If you just hit the D key and want to hit the T key next, on the Kinesis Advantage, in addition to having to know how far the D key is from the T key in the two horizontal dimensions, you have to know the delta in altitude (i.e., how far it is in the vertical dimension).

In addition to the "delta in altitude" information just described, almost every key on the Kinesis (the axis along which the key will travel when pressed) "points" in a different direction, which is two more degrees of freedom. So, for almost every key on the Kinesis, I would have had to "memorize" (i.e., commit to muscle memory) 5 "scalar values" in contrast to the 2 scalar values for every key on a normal keyboard.

I am not saying that no one can "memorize" that information; I am saying only that my estimate of the time and effort required for me to memorize it caused me to ditch my Kinesis for the standard planar arrangement.

Note that I used a Kinesis for 2 years. It was only after I realized that I would need to "memorize" where the keys are (and what direction they travel when pressed) to recover from RSI that I gave up on Kinesis.

And yes, I have been free from RSI in the six years since making this realization and ditching my Kinesis.

And yes, I think most of the ergonomic keyboards currently sold by Microsoft are bad for the same reason the Kinesis Advantage is bad -- particularly those keyboards that could be described as "wave shaped" or "hill shaped". (At least the Kinesis Advantage has decent keyswitches!)

ADDED. There is more wrong with the Kinesis than just the amount of information it requires a person to commit to muscle memory. I just got done stating that the DataHand is an even worse keyboard than the Kinesis, and the DataHand requires a person to have only a few "muscle memories". If I understand correctly and if the DataHand device is a good fit for the person's hand, for every character, the person need know only which finger to twitch and which of four directions (up, down, left or right) in which to twitch the finger.

The reason the DataHand is such a bad keyboard is that it causes the person to repeat over and over again a kind of movement that is very different from the kinds of movements people are used to making and are good at making.

An example of a kind of movement that humans are good at is striking something with a hand or a finger. There is a lot of (conscious and unconscious) "data processing" going on when a human is getting ready to hit something with a hammer or with his fist, hand or finger. My belief is that the best way to avoid RSI and other problems is to get the brain to devote a lot of "cpu cycles" to planning out the physics of every keystroke.

The DataHand (and to a lesser extent the Kinesis) has bad ergonomics because it makes it hard for the brain to do detailed physical planning. This planning would entail perceiving or remembering the exact location, relative to the striking finger, of the key to be pressed. In contrast, when all the keys are on the same plane, it is more likely the the brain will plan out "ballistic" movements resembling hops, or plan out a "backswing" before the "swing", and, again, we want to get the brain deeply engaged in planning movements.

Designing the keyboard so that every key is right next to a finger tip is bad ergonomics because the brain is poorly adapted to carrying out the resulting key-pressing movements.

To summarize, the most important things for avoiding RSI are (1) choosing a keyboard that encourages the right kind of movements (namely, the kind the brain is good at planning) and (2) reminding oneself during typing that pressing a key should be more like hitting a ball with a tennis racket than making a sign with a flag, baton or the hand. Part (1) consists mainly of avoiding so-called "ergonomic" keyboards like the DataHand, Kinesis and most modern Microsoft "ergonomic" keyboards. The Microsoft Natural Elite keyboard by the way was a good keyboard (well, besides the pretty crappy rubber-dome keyswitches) because even though it was a "split" keyboard, all of the keys designed to be hit with the left and lay in the same plane, as did all of the keys designed to be hit with the right hand. But they might be hard to get now. And like I said, the keyswitches are pretty atrocious even compared to the (scissor-style) keyswitches on the good modern Logitech keyboard like the K750.