Introducing the RSTHD layout

[xsznix]

Hi guys, I've been learning a new keyboard layout that I generated using the simulated annealing technique demonstrated in carpalx. What do you guys think? I've done a little non-scientific write-up here:

https://xsznix.wordpress.com/2016/05/16/introducing-the-rsthd-layout/

[PieterGen]

What do I think ? I think it is FANTASTIC !  You wrote a new keyboard generating algorithm, right?   Plus, you wrote it in Rust ?   

You made my day already! BTW, loved how you wrote in the blog "I’m excited to be joining a small group of geeks who have a hobby that nobody really understands or appreciates."  )  When I tell my colleagues "Hey, a guy on Geekhack wrote a new keyboard generator and he did it in Rust too!" they don't understand what are Geekhack, keyboard generator and Rust. Let alone why this would be special  )

Looks like a Dvorak/Maltron cross over. How are the arrow keys in real life? Having them on one hand can be handy too, right? 

Iĺl dive into your layout later. In more detail. Great work !
 

[algernon]

I like how the arrow keys are split between the two halves, will try it, too!

The rest of the layout looks interesting too, albeit I like to place a bit more work on my thumbs: I have all the modifiers on the thumb cluster, and recently moved : there too. But I'm nowhere near 100WPM, let alone 120, so what do I know?

A very interesting post, nevertheless, something I will be coming back to for inspiration and ideas, thank you!

[duq]

Well done, and welcome to the community . I think the whole 'hack' part has lost some of it's meaning this past few years on this board, but this post absolutely does.

[xsznix]

Looks like a Dvorak/Maltron cross over. How are the arrow keys in real life? Having them on one hand can be handy too, right? 

The arrow keys are based on the placement of HJKL on Dvorak: JK are on the bottom left, and HL are on the right. I think using two hands to navigate in two dimensions makes more sense than making one hand do everything, simply because you have more fingers to do the same work. It would be nice to have arrow keys on the same hand sometimes, for example when using the mouse with the right hand, but I've fond that those cases are a lot more rare than when I'm navigating around long documents in vim or another text editor.

It's definitely helpful having them within reach of the longest and strongest fingers though instead of off in the bottom right corner of the keyboard as in most other layouts, because I can switch between typing and navigating without having to move my entire hand. I highly recommend the arrow setup; I'm surprised that more people haven't done it already.

I like to place a bit more work on my thumbs: I have all the modifiers on the thumb cluster, and recently moved : there too.

I find that having E, backspace, enter, space, and the symbol layer toggle on the thumb gives them enough work already… however, if I didn't put E under the thumb, that key would probably be a shift key instead, freeing up the normal shift positions for other modifiers—possibly the symbol layer toggle.

The 1u keys on the thumb cluster are too hard to hit for Ctrl/Alt/Gui, in my opinion; it's quite the reach getting to them. The center column is easier to reach, so for me at least, it makes more sense to put those modifiers there. The thumb cluster keys are used for my four auxiliary layers. You can see those keymaps in the comments of my keyboard firmware: https://github.com/xsznix/qmk_firmware/blob/master/keyboard/ergodox_ez/keymaps/keymap_rsthd.c

[algernon]

I like to place a bit more work on my thumbs: I have all the modifiers on the thumb cluster, and recently moved : there too.

I find that having E, backspace, enter, space, and the symbol layer toggle on the thumb gives them enough work already… however, if I didn't put E under the thumb, that key would probably be a shift key instead, freeing up the normal shift positions for other modifiers—possibly the symbol layer toggle.

The 1u keys on the thumb cluster are too hard to hit for Ctrl/Alt/Gui, in my opinion; it's quite the reach getting to them. The center column is easier to reach, so for me at least, it makes more sense to put those modifiers there. The thumb cluster keys are used for my four auxiliary layers. You can see those keymaps in the comments of my keyboard firmware: https://github.com/xsznix/qmk_firmware/blob/master/keyboard/ergodox_ez/keymaps/keymap_rsthd.c

Mhm. I had similar issues, but I turned my modifiers into toggles instead, so I don't have to hold them, and that made it much more comfortable to use them. (I'd rather reach the 1u keys on the thumb cluster, than fiddle with the bottom row, I just can't get used to using that :/). Having the E on the thumb cluster sounds interesting, I might try it, if I can fit it into my layout somehow

[naz]

How did you generate that awesome metric report table??? i would love to have one of those with spanish data

[squizzler]

I am running this layout on a Minidox. I selected it on the back of its designer's appraisal of Malt: my keyboard journey also took me through Malt, and his comments accorded with my experience. I have written a little on my creation of the firmware in a blog post. If the OP is still on the site and monitoring this thread, thanks for the awesome layout!

[squizzler]

I would also add that I think layouts with the 'E' on the thumb cluster will become more popular off the back of the success of the Ergodox and other split keyboards. This also includes the Malt, and such an arrangement allows greater efficiency than ones where the E must be with all the other letters.

Personally, I see no point in a split keyboard user converting to a staggered-row compatible layouts such as Dvorak and Coleman with the aim of adopting the new keymap on both the split and, for instance, their laptop. Better to learn something optimal to the split arrangement, and stick with Qwerty on the laptop and any machines one might be guest on.

[batfink]

I see people come up with layouts that have E on the thumb from time to time. Personally, I'm against this approach.

I can see the attraction if you start using a keyboard that has thumb keys. But I say this: thumb keys are excellent - but they are best used for layer selection.  Alphabetic keys should be in the main section, then use your thumb keys for cool extra layers like:

- having common symbols (brackets, etc) in great locations on the home row.

- have a navigation or extend layer for easy editing.

- having mappings for foreign characters / custom macros etc or anything else you have a personal requirement for.

[squizzler]

The two schools of thought are not mutually exclusive. I have layer-tap under the 'E' and space key.

Most traditional key maps adapted to the ergodox choose to assign the space and backspace on the home keys of the thumbs. Malt and RSTHD simply replace the backspace withe the 'E', which does involve a little shuffle of things on the thumb pads. Incidentally, as an aspiring writer on a minidox, I elected to put the quote mark on the thumb pad too, here is my keymap.

[Snarfangel]

The two schools of thought are not mutually exclusive. I have layer-tap under the 'E' and space key.

Most traditional key maps adapted to the ergodox choose to assign the space and backspace on the home keys of the thumbs. Malt and RSTHD simply replace the backspace withe the 'E', which does involve a little shuffle of things on the thumb pads. Incidentally, as an aspiring writer on a minidox, I elected to put the quote mark on the thumb pad too, here is my keymap.

You would really hate the layout I've used on my Kinesis Advantage2 for the last couple of years. 


Here is the comparison using the keyboard layout analyzer here: https://kla.keyboard-design.com/#/main with the books.short.txt from the RSTHD keyboard layout project:

(I know these analyzers are only as good as the background assumptions, but the layout works well for me. The only thing I would add is a dedicated backspace key conveniently near the thumb or index finger -- I use [shift]-space, which is normally fine unless I want several backspaces in a row. But I do like the rest of it, and can easily touchtype with it.)

If anyone wants to plop it into a keyboard layout analyzer, I will see if I can attach the file to the post.  YPHINAFU_Keyboard.txt (15.58 kB - downloaded 175 times.)

[squizzler]

You would really hate the layout I've used on my Kinesis Advantage2 for the last couple of years. 

Your layout is very interesting, but as a supporter of Malt and its descendants (of which I regard RSTHD) it certainly does not offend me! When referring to the Malt family I mean layouts that place the 'E' under one of the thumbs. But does your addition of further vowels onto the thumb clusters make it something totally new?

Always remember that Malt layout was designed with the Maltron split keyboard as an integrated system, and all columnar split keyboards are Maltron descendants. Whilst Qwerty and other layouts from staggered boards work fine on columnar, these layouts do not take full advantage of such ergonomics. With so many split columnars such as Ergodox currently available, now is an exciting time for layouts of the Malt family!

Here is the comparison using the keyboard layout analyzer here:

I am puzzled as to what the two variants of RSTHD might be. Would those of us on 'columnar' split keyboards be using 'Ergolinear 2' and the other one being that arrangement adapted onto a traditional staggered board? Or am I using something out of date?

(I know these analyzers are only as good as the background assumptions, but the layout works well for me. The only thing I would add is a dedicated backspace key conveniently near the thumb or index finger -- I use [shift]-space, which is normally fine unless I want several backspaces in a row. But I do like the rest of it, and can easily touchtype with it.)

I know what you mean; but try deciding what gets the half dozen thumb keys with minidox! I was going round in circles for ages trying to figure the optimum way to use them. I am used to fn+bksp = delete on the MacBook, and I have emulated this so that the bksp becomes del on a layer (actuated by holding space key).

[squizzler]

Over on the Keyboardio forum the Malt thread has become host to a lively discussion on adapting RSTHD to smaller boards, especially the forthcoming Atreus.

The Atreus could drastically lower barriers to entry for Malt and RSTHD layouts. It will be mass produced and promises to be more affordable than previous Maltron descendants (Kinesis, *dox, etc). The discreet size and more traditional keyboard construction and styling ought not frighten the horses in your mainstream office environment! It will also have the benefit of all key caps being standard 1x1 size which means they can be rearranged with the 'E' (or indeed anything else) on the thumb home keys.

[squizzler]

I am pleased to confirm that my own keymap for this layout has been inducted to the QMK repository. It is for the Minidox but should also port to the gBoards Gergoplex, whose use of combos was an inspiration for another neat feature. Paired keys on the inner two columns create a virtual extra column each side, again the keys each side of the divide create a virtual centre column, and the the quotation mark is a combo of 'H' and 'N' (air quotes!). Combos are shown in blue in this diagram:

[hermes1908]

I would also add that I think layouts with the 'E' on the thumb cluster will become more popular off the back of the success of the Ergodox and other split keyboards. This also includes the Malt, and such an arrangement allows greater efficiency than ones where the E must be with all the other letters.

Personally, I see no point in a split keyboard user converting to a staggered-row compatible layouts such as Dvorak and Coleman with the aim of adopting the new keymap on both the split and, for instance, their laptop. Better to learn something optimal to the split arrangement, and stick with Qwerty on the laptop and any machines one might be guest on.

I think the problem with this argument is that it is predicated on the assumption that the thumb is the optimal position for a letter key. Even if you assume that this is true not all thumb keys are made equal. The positioning of the e thumb key on the maltron is carefully designed to avoid unnecessary strain when used as a main key, this is generally untrue for other keyboard (some of which badly try to mimick the position and end up causing thumb problems). The other problem is that most computer users do not spend most of their time typing. Keyboard shortcuts are often overlooked and the thumb cluster (assuming your keyboard has one) is prime real estate. This is especially true for programmers (who tend to be the principal consumers of alternative keyboards). For instance, I hit C-f and C-t far more often than I type the word 'the'. Any analysis which fails to take modifiers into account would seem to be incomplete. However doing so properly is itself highly fraught since modifier usage varies by individual and program. I remain unconvinced that layouts like Maltron and RSTHD are objectively superior to Colemak DH/Dvorak on curved/split boards when modifier usage is optimized.

[cheater]

generated using the simulated annealing technique demonstrated in carpalx

what makes a monte carlo method the right way to search for the best keyboard layout?

[jacobolus]

I think the problem with this argument is that it is predicated on the assumption that the thumb is the optimal position for a letter key.

It doesn't have to be the best position, only an acceptably good position.

The Malt, like a standard keyboard or most "flat" keyboards, uses a sideways striking type of thumb movement (flexion of the trapeziometacarpal joint at the base of the thumb). Some other "ergonomic" keyboards don't get the orientation right, but parallel to the other keys is close enough.

The other problem is that most computer users do not spend most of their time typing. Keyboard shortcuts are often overlooked and the thumb cluster (assuming your keyboard has one) is prime real estate.

Thumb keys have the big advantage that they can often be independently typed or held alongside finger keys on the same hand, but this isn't an essential limitation. For example finger modifiers can put the keyboard into an alternate mode (not needing to be held down).

Alternately, if both hands are assumed to stay on the keyboard (for one-handed operation constraints are different), theoretically a single thumb modifier on each hand is enough; finger keys on the same hand can be used alongside as additional layers or modifiers, affecting keystrokes pressed on the other hand. Standard layer-switching with dedicated separate layer keys is arguably a waste of space.

Colemak DH/Dvorak

These are both suboptimal. They are built on outdated and limited assumptions about both physical keyboard shape and the way typing motions work (e.g. Dvorak is designed for mid-20th-century typewriters which are substantially different from any currently available computer keyboard). However there are diminishing returns to improvements in letter layout, so any of them is a reasonably acceptable alternative to QWERTY.

[hermes1908]

The Malt, like a standard keyboard or most "flat" keyboards, uses a sideways striking type of thumb movement (flexion of the trapeziometacarpal joint at the base of the thumb). Some other "ergonomic" keyboards don't get the orientation right, but parallel to the other keys is close enough.

Maybe, but this hasn't really been studied and there aren't enough users to know whether or not such implementations will cause potential issues in the long term (anecdotally there are already some complaints).

Thumb keys have the big advantage that they can often be independently typed or held alongside finger keys on the same hand, but this isn't an essential limitation. For example finger modifiers can put the keyboard into an alternate mode (not needing to be held down).

This is factored into my analysis. Whether or not the thumb keys are used as so called 'one shot' modifiers is irrelevant. The problem is that once you include modifiers and real computer use the equation changes and complicates the analysis.

Alternately, if both hands are assumed to stay on the keyboard (for one-handed operation constraints are different), theoretically a single thumb modifier on each hand is enough; finger keys on the same hand can be used alongside as additional layers or modifiers, affecting keystrokes pressed on the other hand.  Standard layer-switching with dedicated separate layer keys is arguably a waste of space.

This sort of chording has ergonomic implications that would have to be factored into the model. Intuitively this seems like it would involve a lot more strain than having dedicated keys for alt/ctrl/other layers.

Quote Colemak DH/Dvorak These are both suboptimal. They are built on outdated and limited assumptions about both physical keyboard shape and the way typing motions work (e.g. Dvorak is designed for mid-20th-century typewriters which are substantially different from any currently available computer keyboard).  However there are diminishing returns to improvements in letter layout, so any of them is a reasonably acceptable alternative to QWERTY.

It is not self evident that layout optimization for typewriters dramatically differs from the optimization criteria for computer keyboards when typing English text. Even if you believe this it is not clear what the optimization criteria should be or whether or not any appreciable gains can be made over common layouts like Dvorak or Colemak.

[cheater]

The Malt, like a standard keyboard or most "flat" keyboards, uses a sideways striking type of thumb movement (flexion of the trapeziometacarpal joint at the base of the thumb). Some other "ergonomic" keyboards don't get the orientation right, but parallel to the other keys is close enough.

Maybe, but this hasn't really been studied and there aren't enough users to know whether or not such implementations will cause potential issues in the long term (anecdotally there are already some complaints).

Thumb keys have the big advantage that they can often be independently typed or held alongside finger keys on the same hand, but this isn't an essential limitation. For example finger modifiers can put the keyboard into an alternate mode (not needing to be held down).

This is factored into my analysis. Whether or not the thumb keys are used as so called 'one shot' modifiers is irrelevant. The problem is that once you include modifiers and real computer use the equation changes and complicates the analysis.

Alternately, if both hands are assumed to stay on the keyboard (for one-handed operation constraints are different), theoretically a single thumb modifier on each hand is enough; finger keys on the same hand can be used alongside as additional layers or modifiers, affecting keystrokes pressed on the other hand.  Standard layer-switching with dedicated separate layer keys is arguably a waste of space.

This sort of chording has ergonomic implications that would have to be factored into the model. Intuitively this seems like it would involve a lot more strain than having dedicated keys for alt/ctrl/other layers.

Quote Colemak DH/Dvorak These are both suboptimal. They are built on outdated and limited assumptions about both physical keyboard shape and the way typing motions work (e.g. Dvorak is designed for mid-20th-century typewriters which are substantially different from any currently available computer keyboard).  However there are diminishing returns to improvements in letter layout, so any of them is a reasonably acceptable alternative to QWERTY.

It is not self evident that layout optimization for typewriters dramatically differs from the optimization criteria for computer keyboards when typing English text. Even if you believe this it is not clear what the optimization criteria should be or whether or not any appreciable gains can be made over common layouts like Dvorak or Colemak.

some discussion of the thumb's role in preventing rsi: https://geekhack.org/index.php?topic=110354.0

[jacobolus]

It is not self evident that layout optimization for typewriters dramatically differs from the optimization criteria for computer keyboards when typing English text. Even if you believe this it is not clear what the optimization criteria should be or whether or not any appreciable gains can be made over common layouts like Dvorak or Colemak.

I don't know what you consider to be "self evident". Clearly it is not apparent to many people choosing (or even designing) keyboard layouts, as this is commonly overlooked. However it is pretty straight-forward to understand if you spend some time thinking carefully about hand anatomy and typing motions.

Manual typewriters, between their invention in the 19th century and well past when the Dvorak layout was created, are very steeply angled, with a tall step between keys. It is relatively easy to reach up to the upper row from the home row, but a much bigger challenge to reach down to the lower row, especially for the middle fingers.

Most electronic typewriters / most desktop keyboards, along the lines of the IBM/Honeywell designs from the 1960s, make the lower rows easier to reach than they were on manual typewriters, but make it a bit (relatively) tougher to reach the upper rows because the step is not as aggressive.

Flat laptop keyboards (and other keyboards with uniform keycaps) exaggerate this. Reaching the upper rows requires either moving the whole hand or stretching out the fingers into a position that partially compromises the strength of the finger flexors used to type the keys. However, moving the fingertips one row closer to the body without moving the hand is relatively easy.

When typing several keys on different rows in quick succession, these features are amplified. (Many manual typewriters have the additional constraint that keys need to be pressed more rhythmically to avoid jamming, so "arpeggio" kinds of sequences are less relevant.)

The Dvorak layout optimizes for the manual typewriter design by only putting rare keys on the bottom row. The implicit key comfort cost function used for optimizing the layout is not really relevant on a modern keyboard.

(A simple summary describing whole rows is oversimplified; how reachable each key is from home position varies from key to key, finger to finger, and typist to typist. But it is a start.)

[hermes1908]

Flat laptop keyboards (and other keyboards with uniform keycaps) exaggerate this. Reaching the upper rows requires either moving the whole hand or stretching out the fingers into a position that partially compromises the strength of the finger flexors used to type the keys. However, moving the fingertips one row closer to the body without moving the hand is relatively easy.

This seems highly subjective. Personally I feel the opposite is true (reaching the upper rows is far easier for me than reaching for the lower ones).

When typing several keys on different rows in quick succession, these features are amplified. (Many manual typewriters have the additional constraint that keys need to be pressed more rhythmically to avoid jamming, so "arpeggio" kinds of sequences are less relevant.)

Yes, but it doesn't follow that this isn't also good for your hands. It may or may not be. Anecdotally I find that the rhythm of hand alternation gives my hands time to rest whilst typing.

The Dvorak layout optimizes for the manual typewriter design by only putting rare keys on the bottom row. The implicit key comfort cost function used for optimizing the layout is not really relevant on a modern keyboard.

Again, you are making the mistake of assuming that this is an inherently bad thing on modern keyboards. If you are going to maintain that 'modern layouts' like colemak are superior because they weren't 'designed for a typewriter' then you are going to have to substantiate it with something. I would be interested in knowing what you regard as the ideal layout.

[jacobolus]

I don't have a strong opinion about whether Colemak is "superior", but Dvorak is clearly suboptimal, with the form of 1930s typewriters strongly informing the design. I don't think 1930s typewriters are anything close to an ideal shape, but it is hard to deny that the shape is significantly different from a 2020 laptop.

Different people hold their hands in different ways. It would be fairer to say that a flat design forces the typist to make a personal trade-off between easier vs. harder to reach keys. The flat shape makes it impossible to reach all of the keys with anywhere close to equal facility. A better keyboard shape (split, tented, column-staggered, with height step set to leave the proximal finger joint in neutral position at the top of the stroke without moving the hand) makes off-home-row keys much less annoying.

All of the optimization methods I have seen for creating logical layouts are based largely on gut-feel heuristics and fanciful invented penalty scores, none of which seem right to me and few if any of which were driven by data or detailed anatomical reasoning (they still make a big improvement vs. QWERTY; very low bar there). You could get better results, but doing it in a principled way would take serious research. Especially if you want it to fit a wide range of typists.

[cheater]

I don't have a strong opinion about whether Colemak is "superior", but Dvorak is clearly suboptimal, with the form of 1930s typewriters strongly informing the design. I don't think 1930s typewriters are anything close to an ideal shape, but it is hard to deny that the shape is significantly different from a 2020 laptop.

Different people hold their hands in different ways. It would be fairer to say that a flat design forces the typist to make a personal trade-off between easier vs. harder to reach keys. The flat shape makes it impossible to reach all of the keys with anywhere close to equal facility. A better keyboard shape (split, tented, column-staggered, with height step set to leave the proximal finger joint in neutral position at the top of the stroke without moving the hand) makes off-home-row keys much less annoying.

All of the optimization methods I have seen for creating logical layouts are based largely on gut-feel heuristics and fanciful invented penalty scores, none of which seem right to me and few if any of which were driven by data or detailed anatomical reasoning (they still make a big improvement vs. QWERTY; very low bar there). You could get better results, but doing it in a principled way would take serious research. Especially if you want it to fit a wide range of typists.

there's no such thing as a "proximal finger joint". Which do you mean, Metacarpal-Proximal (MP) or Proximal-Intermediate (PIP)?

[jacobolus]

there's no such thing as a "proximal finger joint"

Dude.

Proximal in anatomy means toward the body. See https://en.wikipedia.org/wiki/Anatomical_terms_of_location

A joint is a connection between adjacent bones. A finger is one of the digits of the hand, an articulated bony appendage.

If we look at a human finger other than the thumbs, there are 3 hinge-like joints, and one of them is proximal relative to the others. This is the joint between the proximal phalanx bone and the metacarpal bone, which is typically called the metacarpophalangeal (MCP) joint.

(The "phalanx" bones are called that because bones in the same row of adjacent fingers fit together like a soldiers in a battle line. "Metacarpal" just means "beyond the wrist".)

This joint is the main one flexed for typing a key, mainly by the flexor digitorum profundus muscle (Latin for "deep finger bender") in the forearm. Other finger/hand motions are used for locating the key top, but actual keypresses are done via flexion at the MCP joint.

Flexion of the two distal finger joints is much weaker and slower, and would not be a comfortable or effective means of typing with a typical keyswitch.

The proximal (MCP) joint is at the strongest part of its range of motion when the proximal phalanx bone extends straight out in the plane of the palm of the hand. When it is partially flexed, it loses a significant part of its strength. (At least, this is my personal observation for typing motions. I am not a trained professional expert in human hand anatomy; it might well be that the story changes a bit when discussing strength for gripping a hammer or something, where coordinated sustained flexion of the other finger joints comes into play.) For this reason, reaching the finger outward on a flat keyboard leads to less comfortable keypresses than simply extending (uncurling) the middle finger joint (proximal interphalangeal joint, PIP) to reach a further key which is vertically stepped: the latter keeps the MCP joint in a neutral position where it has the strongest, fastest, and most comfortable flexion.

While we are at it, flexion of the wrist also compromises strength of MCP joint flexion used for typing motions (and often leads to RSI).