hey steve - do you have any plans to put out an xkb version for linux?
As a first step in this direction, I created some xmodmap files (as they are pretty easy):
Using Colemak-DH with Seniply.
This mod is interesting and makes me want to give it a try. I worry the M key is going to have the same problems the H key did, with ME and EM, but maybe it's not frequent enough to be problematic. Otherwise it looks really good.
I am currently typing a layout that is a combination of Workman and Colemak.
Q W R D P J L U Y ;
A S H T G F N E I O '
Z X V C B K M , . /
I prefer to move my index fingers down rather than stretch up, which is my only real complaint with this layout. That and typing "cd" on the Linux terminal.
What do you guys think, would Mod-DH be an upgrade?
I threw together a basic keymap for xkb.
After playing around with it for just a few minutes, I can already tell the EM combos are going to be problematic for me. However, if I swap M and K, that problem largely goes away, and has the added benefit making the J/K vim keys next to each other. Hmm, maybe I'll give that a try.
Hi xedrac, glad you like the look of Mod-DH! Of the changes Mod-DH makes, I agree the new location of M is the most difficult - All the other changes feel like improvements, whereas the M key is worse. The thing to bear in mind though, is at least the ME/EM only accounts for 1.15% of pairs, whereas HE/EH accounts for 3.49%, so you only have to do that motion about one-third of the time compared to before.
Your Workman-Colemak combination layout looks pretty good to me, I considered something similar to that myself previously. I think I would prefer to have XCV in the usual order at least, so I would make the bottom row Z B X C V (assuming angle mod). In that layout you still have the FE/EF combination though. If you prefer your index fingers to go down rather than up though, then Mod-DH should be a definite improvement!
I presume your most recent post means you now have K in the middle centre-column. Certainly KE is a less common than ME so if you want to minimize that motion it sounds a good idea, although it means an additional change relative to Colemak (but not relative to Qwerty). For standard keyboards I think that MK change is fairly reasonable, might be less so for matrix-like boards though.
P.S. thanks for the xkb contribution!
Using Colemak-DH with Seniply.
It turns out I didn't like the K/M swap as much as I thought I would. I think I'm going to just give it a try the way you've got it.
Heh, the more I look at this the more interesting it looks. Might give it a try soon. I don't think I want to change the H/M order (yet) because of the entropy of moving two more keys away from QWERTY so it'll be a left-hand only experiment for me if I do it.
However, I still feel that the F/U and even W/Y positions are nearly as good as the V/M ones. In particular, in your model I'd score more like this:
effort:
3.2 2.0 1.8 2.4 3.4 3.8 2.4 1.8 2.0 3.2
1.5 1.2 1.0 1.0 2.6 2.6 1.0 1.0 1.2 1.5
3.6 2.9 2.5 1.7 2.8 2.8 1.7 2.5 2.9 3.5
penalties:
2.8 | same-row bigram(?)
2.8 | one row bigram
3.5 | two row bigramThe weights are completely subjective, of course. :)
I first couldn't get your modeling tool to work as the Java RE plugin wouldn't install. Ugh. Well, better now.
My weights gave these results for your total finger effort measure (Keyboard Analysis Tool V1.01):
Colemak: 1.69 (you got 1.70)
DVBG only: 1.65 (you got 1.67)
DVBG+HM: 1.62 (you got 1.64)
I don't know what it means that my numbers are a bit lower than yours? Is my typing more carefree maybe? :) I suspect that the major impact is from the lower weights on the (W)F and U(Y) keys. The differences are at any rate minor.
Actually, here's a criticism: It seems your first penalty is for same-row and your second one a row jumping one. Well, instead it should be one for same-key and one or two key distance jumps (roughly calculated)! I don't see a big difference between typing a PG and a GD bigram, but a clear difference between PP and PG. To refine this, the PD bigram seems quite a bit worse than the GT one (a bit less so with DV vs TB). Don't know what to make of that...
*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***
Interesting! Your weights do seem more in line with others I have seen (e.g. Workman's). Perhaps I am just unusual in prefering the P/L keys over W/Y. As a fairly new touch-typist, maybe my ring fingers are not yet as strong as those who have touch-typed for years. I'm sure your final results come out lower because on average your weights look lower than mine, especially for those top-row middle and ring fingers.
I can totally understand why you would consider experimenting with the left hand change first. I wasn't even sure about making the right-hand change at first. I have been using it for a while now though, and have decided it's worthwhile just on the grounds of how common H is and the newly lovely HE bigram. I still think doing the left-hand change brings the biggest benefit though.
About the bigram penalties: not sure on whether your bigram examples are referring to Colemak or Mod-DH, but to hopefully clarify: currently same-key repetition is not penalized. Do you think it should be? Same row penalty would apply to (in mod-dh layout) PB, TG, DV but not PP. There is currently no distiction between the different types of one-row jumps, so these are all treated the same: PT, PG, TB, BG, TD, TV, etc. Perhaps that should be changed, with the "diagonal" moves being considered worse? Similarly for the two-row jump penalty, this would cover PD, PV, BD, BV. I am happy to consider making changes to the tool to make it more flexible, as long as they're not too complicated :P
Using Colemak-DH with Seniply.
Same-key repetition is an interesting question. It doesn't lead to any jumps, but is it layout agnostic? I'd say no: I'd much rather hammer my index finger at something repeatedly than doing the same with my ring finger! So far, I haven't seen any analysis take this into consideration. But my initial feeling that same-key bigrams as such should have a separate weighting may not be correct? On one hand, such a number won't affect your choice of layout! On the other hand, it'll contribute to the typing "effort" which is what you're trying to model right? Depends mainly on how much meaning you want to give to the effort parameter itself, then. Not too much, hehe.
Not sure about modeling the diagonals, because it's not a simple matter of sqrt(2)=1.41 and sqrt(5)=2.24 because of row stagger. I've outlined the jump distances for a standard 1/4–1/2 stagger keyboard below:
Horz Vert Dist Where (on Cmk w/ or wo/ AngleMod)
1.00 0.00 1.000 T-D / N-H
0.25 1.00 1.031 T-P / N-L
0.50 1.00 1.118 T-VB / N-K
0.75 1.00 1.250 T-G / H-L
1.25 1.00 1.601 D-P / N-J
0.25 2.00 2.016 VB-G / K-L
0.75 2.00 2.136 VB-P / M-L
1.75 2.00 2.657 B_-P / M-JThe ones that really make this worth considering are the 0.75 and 1.25 stagger jumps which give 25% and 60%(!) error, and the 1.75 stagger two-row jump which gives 33% error if using a simple row jump model; the others are fairly insignificant deviations from 1 and 2. In actuality, the TD jump is quite different from the PG jump for instance – but part of that is baked into the weight for the D vs other keys too! Hmmm...
Interestingly, the diagonal jumps involve both vertical and lateral motion which means you have to use more muscles than for any other jump (somewhat simplified)!
Should vertical (row) and lateral jumps be treated differently? That may be a question of individual preferences. I don't perceive them as very different even if they do involve different muscle groups. Maybe most of those muscles are of similar strength and agility.
I do notice that the HM switch contributes a lot to the effort parameter in your current model! Also, it's a sweet little two-key loop that's completely separate so it could be its own mini-step in Tarmak (or bundled with the LUI loop if so desired). The left-hand mod, on the other hand, changes the Tarmak game, requiring a reworking of at least the 'T' loop and the "internalization" of a partial Angle loop:
Colemak loops: Y>O>;>P - R>S>D - G>T>F - E>K>N>J L>U>I [Angle B>V>C>X>Z>_]
Colemak-DH loops: Y>O>;>P - R>S>D - C>X>Z>_>B>T>F - E>K>N>J L>U>I H>MHmmm. That doesn't look too bad actually: Instead of the usual GTF loop we can have a CXZ_BTF loop; it may be that long without much trouble since the CXZ keys don't change fingering but only the wrist angle. Same number of loops, if you count the Angle mod. Or the C>X>Z>_ loop may be separated off for that matter; probably a good idea after all. The R>S>D loop, while having the same name here, will be slightly different of course since the D will end up at the QWERTY C position.
But I'm thinking way too far ahead: First let's see whether I like it enough, what! :-D
How's the XCV shortcut trio working out with this mod? See, the Z as Undo is a bit separate from the cut-copy-paste block which is now split up. Any misfires so far when trying to paste?
*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***
When thinking about how much penalty give bigrams though, it's worth bearing in mind that this is in addition to the "normal" key effort penalty from typing those keys - it's much easier to type TT than QQ, but the question is, should there be an additional penalty over and above that incurred by the key-effort values? I would say probably not.
I think the diagonals probably should have a higher penalty than the vertical ones. Maybe I'll add that at some point. I don't think I would take into account all those precise distances though, I would rather keep things relatively simple given the whole exercise is full of guesses and estimates anyway!
Nice work on the Tarmak loops that would be needed! Despite using tarmak myself, it's not something I thought about much. I guess I was imagining any take-up would be from existing Colemak users.
The copy-cut-paste change was tricky at first. I was starting from a non-angle-mod situation, so for me V stayed the same but Z X C moved. Definitely I found C particularly hard to adjust to - not just for copy but in general typing as well - as using my index finger for C had become ingrained!
Probably for existing angle-mod users the transition would be easier, but it would be interesting to know!
Using Colemak-DH with Seniply.
new years resolution, give stevep99 mods and dreymars extend layers a go..
One caveat with moving the shortcut keys is what will happen when you mistype them! I remember the chagrin of hitting Ctrl+W instead of the neighboring Ctrl+V using Dvorak – thus closing the tab instead of pasting into the text field on it, thereby ruining the post I was making!
So what'll happen with this mod? I can foresee hitting Ctrl+D instead of Ctrl+V, which can have a number of odd consequences. The worst would be in Linux where you'll close a terminal window with Ctrl+D, similarly to the Dvorak flaw above! :-o
Luckily, Ctrl+C instead of Ctrl+X is no problem whatsoever. I've been quite happy with that serendipity using the Angle mod. Ctrl+X vs Ctrl+Z depends on how good Undo capabilities your programs have, but I haven't had any big troubles that I can remember with it.
*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***
Slight weight tweak attempt:
effort:
3.2 2.1 1.9 2.4 3.4 3.8 2.4 1.9 2.1 3.2
1.6 1.2 1.0 1.0 2.6 2.6 1.0 1.0 1.2 1.6
3.6 2.9 2.5 1.8 2.5 2.5 1.8 2.5 2.9 3.6
penalties:
2.8 | same-row bigram(?)
2.8 | one row bigram
3.5 | two row bigramThese weights gave the following results for total finger effort measure (Keyboard Analysis Tool V1.01):
Colemak: 1.71
Workman: 1.68
DVBG only: 1.68
DVBG+HM: 1.65
Again, compared to your weights:
Colemak: 1.70
Workman: 1.68
DVBG only: 1.67
DVBG+HM: 1.64
It's fairly obvious then, that a little nitpicking from my side won't affect your results much. ;-)
I included Workman because it seems to be the one that got this thing started. It deviates a lot from Colemak, changes more keys and in the end doesn't really accomplish all that much for the effort it seems? It has a higher same-finger ratio which diminishes its achievements. In contrast, moving a few keys around a little to your mod makes it still look very much like Colemak to me but easily gains these advantages to the adversaries of the middle columns.
I tried fiddling with keeping the V next to C but in this model that bears a heavyish penalty: Just doing a BGD rocade netted me a 1.70 score with my weights which is too little gain really (and DGB top-to-bottom gets an abysmal 1.73). Ah well. At least, that's a little better than Colemak which suggests that Colemak might've kept the G in place and possibly been better off for it. :-)
*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***
Yeah agree, as long as your model is still based on the principle that the top-row middle/ring finger and bottom-row index finger keys are better than the centre-column keys, then I reckon you are bound to see a significant improvement.
Workman is an interesting case, as you say, it makes some improvements at the cost of making some things worse. In my view it correctly identifies Colemak's flaws, but its proposed solution is rather bizarre - as the criticisms it makes could have been overcome by making fewer (and simpler) changes, giving a result which is closer to both Colemak and Qwerty... which is where Mod-DH steps in - it solves the same problem that Workman solves, but without Workman's drawbacks.
It would be nice in theory to keep XCV together, but it's hard to reconcile that with having a nice position for D. Mind you, I don't think the new XCDV pattern is too bad though, the keys are still in the same order and the cut-copy-paste pattern seems familiar enough that it doesn't take *too* much effort to relearn. It's certainly no worse than having to learn Colemak's S position which took me ages to get used to!
Using Colemak-DH with Seniply.
Awrighty, here we go! Colemak-DH[eD] for XKB!
Use this download which is essentially the standard "Big Bag" download now.
(Plus, get a sneak preview of the new install script that can install to different folders and whatnot! To install to the X11 dir like before, use the -o option.)
What you have to do to use the mod is to uncomment these two lines in the symbols/colemak file:
include "colemak(cmk-hm_ed)"
include "colemak(cmk-dvbg_ed)"This'll enable two loops that remap to the DH mod using the "English (Colemak[eD])" aka colemak(cmk_ed) layout. Or, if you're using another layout you may include those two lines in that layout after any other definitions of the involved keys and hey presto. Lv3-4 mappings will be mine for the affected keys.
On a side note, I made a little trick for the lv4 M/N mappings with the mod! In normal Colemak there's an arrow cross that's fairly intelligently mapped to NEIM but with the DH mod that's a bit pointless. Instead I made an arrow line like the one Vim uses (←↓↑→)! I hope some will appreciate that. ;)
*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***
As an experiment, I had a bit of a play around with the idea of keeping XCV block together while keeping the nice D position. If you didn't mind making a lot of other changes, you could do this:
q w l p b j m u y ;
a s r t g f n e i o
x c v d z k h , . /
- Reverts S and G to qwerty position
- Scores 1.638 in my model, pretty much the same as Mod-DH. Same finger 1.54% similar to Colemak
- For me the left hand bottom row is a bit too weird though.
Doesn't really improve on things overall I'd say.
Using Colemak-DH with Seniply.
Awrighty, here we go! Colemak-DH[eD] for XKB!
Hey, nice work DreymaR! I'll have to fire up my spare linux box just to try this out!
Using Colemak-DH with Seniply.
ooh thats a nice xmas present dr. drey - thankyou very much
apologies in advance for my naivety, but how is one to access that pound sign on the '4' key?
I got it down and ran the install script, I ended up with a new directory in my home called drey-xmod. Even when run with sudo. Is that what should happen?
Update: OK, I get it now, what I needed was:
sudo sh install-dreymar-xmod.sh -oI did get this message but actually it seemed to install OK
install-dreymar-xmod.sh: 204: [: no: unexpected operatorIt worked! Well, it kind of worked. I edited the symbols/colemak and selected "colemak[eD]" from the GUI and sure enough it applied colemak with the DH mod! However I don't quite understand what to edit to get it to work for e.g. UK colemak and also to make it apply the wide- and angle-mods as well. Would be nice if the GUI recognised the available variants from the config and made them selectable, but I suppose that would too much to ask.
(I am using Ubuntu 14.04)
Using Colemak-DH with Seniply.
This is written using Steve's mod, both hands and all! Argh, it's very annoying right now as my mind want to rush ahead and use vanilla Colemak of course. But it feels interesting! I really like typing the 'd' and 'h'. For the 'dv' bigram I've already started thinking alternative fingering, heh.
BPH: To get the £ sign and other lv4 glyphs, you have to press lv3_switch (usually AltGr) with Shift.
Steve: Kudos, this is interesting! We'll really have to give it a better name. Maybe "Coilmak"† (it's not quite Coal, it's (C)oil – also, it lets you "separate the waters"...)? :-)
• The default operation of the DreymaR-install script is now a local install by copy-paste. You figured out how to make a system install instead I see (with '-o'). Nice!
• No idea what might cause that error. Could you please try a fresh download and if you still get an error, let me know its line number again? No such errors here (Ubuntu 14.04).
• To get it to work for the UK locale that's the simplest thing really: When you activate those lines in the colemak symbols file they'll work for both US and UK!
(Note that there's another UK Colemak that isn't mine – I will take no responibility for that one!)
• For other locales ('cmk_ed_us'), the only problem is if that locale remaps the 'h' key – then its name has to be changed from <AC06> to <AB07> in the locale symbols file.
• X GUI is off the menu for now I'm afraid: It'd mean generating a bunch of new entries in the rules files for the mod (and maybe even both left and both hand variants?!).
• What I have been able to do though, is move the lines so all locales by default follow them: They're now in colemak(cmk_ed_letters). Some tweaks might still be needed...
†:Just kidding. It's a kinda cool name, but untenable since it'd sound so like Colemak that it'd be quite impossible to use. Heh, I knew that...
Update: 'Dholemak'? The Dhole is a type of wild dog-like animal, pretty much like this mod? ;-)
*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***
I'm in doubt about DVBG. Breaking up the ZXCV block is quite disruptive, and it messed with my Extend Paste binding somehow.
The middle column Colemak DH positions still are as awkward as ever, though. So what I'm gonna do is try a DBG loop for a while. When I sit and test presses, the Colemak(Angle) BK positions feel a lot better than the weights we gave them up there! Not quite as good as VM maybe, but still good. And a long shot better than DH for sure! The index fingers are the best at lateral movement after all, and the movement to BK isn't big either. (Almost makes me wonder if the K got too good a deal, heh.)
This requires an Angle mod to work, of course.
On a side note, maybe you want to make a forum signature like mine that links to your github mod pages? :-)
(I used http://goo.gl to shorten the links since there's a limit to the number of characters in a forum sig.)
*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***
Rightyho, again. Tarmak is figured out for the mod.
Tarmak2, modified for use with the DVBG loop (with the DBG loop, simply swap V and J)
Tarmak3, modified for use with the DVBG loop
To use the mod with Tarmak, I recommend doing the Angle mod some with or sometime before Tarmak2, and the HM loop whenever you feel like it really (probably late in the progression?).
With the DBG or DVBG loop set, Tarmak1 should now be unaffected and Tarmak3 onwards will include the mod. That way, it's completely up to you when to introduce the mod loops! It'll be less confusing overall to introduce them in Tarmak2 though – otherwise you'll remap the G to a D and then unmap it later which is a bit silly.
For Tarmak2 with D(V)BG some lines in the symbols/colemak file need to be activated/deactivated like this:
// Tarmak symbols for xkb on X.Org Server 7.x
// See the first Tarmak layout above for more info. This is a QWERTY mask for one Tarmak step.
// The Tarmak2(ET) step moves the G>T>F keys to their Colemak positions, leaving the J at QWERTY 'G'.
partial alphanumeric_keys
xkb_symbols "tarmak2_et" {
include "colemak(cmk_ed)"
include "colemak(tarmak_mask_prs_to_rsd)"
include "colemak(tarmak_mask_jyoc_to_yocp)"
// To include the standalone LUI loop in this step, comment out its mask below
include "colemak(tarmak_mask_lui_to_uil)"
// key <AC05> { [ j, J, U0283, U01A9 ] }; // Cmk D - stray J
//// NOTE: The below is for DBG/DVBG mods (VJ/JV on <AB04>/<AB05> respectively)
key <AB05> { [ v, V, division, Greek_gamma ] }; // QWE/Cmk V
key <AB04> { [ j, J, U0283, U01A9 ] }; // QWE/Cmk B - stray J
name[Group1]= "English/Intl (Tarmak2_ET - Colemak transitional)";
};So instead of making <AC05> (QWERTY G) the new temporary home of J, that key is left alone and the V and B keys are remapped instead. Note that the file will default to VJ (corresponding to the less drastic DBG loop) for those two keys so the AB04 and AB05 names may need swapping for you.
*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***
This is written using Steve's mod, both hands and all! Argh, it's very annoying right now as my mind want to rush ahead and use vanilla Colemak of course. But it feels interesting! I really like typing the 'd' and 'h'. For the 'dv' bigram I've already started thinking alternative fingering, heh.
Nice! Perhaps you should try the left-hand change only first? I found the new H quite easy to adapt to but the M bugged me for quite a while. I can imagine trying to cope with that simultaneously with the D change might be tricky...
Colemak(Angle) BK positions feel a lot better than the weights we gave them up there!
I think you're are probably right, they are not so bad thanks to the bottom row ½-key offeset. But, if I ever manage to get a matrix-style hand-separated keyboard, I might think differently. Those two keys would suddely be a fair bit worse I suspect.
No idea what might cause that error. Could you please try a fresh download and if you still get an error, let me know its line number again?
Sure. I downloaded and installed it again. Didn't revert the previous. This is the output:
$ sudo sh install-dreymar-xmod.sh -o
[sudo] password for steve:
â¢â¢â¢ DreymaR's Big Bag Of Tricks install script (by GadOE, 2014) â¢â¢â¢
⢠Found mod root dir './x-mod_v2-12-1_2014-11-12'
⢠Subdirectories to mod: 'xkb'
install-dreymar-xmod.sh: 158: [: 0: unexpected operator
⢠Backing up: None
⢠Replacing files in '/usr/share/X11/xkb' with mod
install-dreymar-xmod.sh: 204: [: no: unexpected operator
â¢â¢â¢ XKBmap activation skipped â¢â¢â¢
â¢â¢â¢ install-dreymar-xmod.sh finished! â¢â¢â¢After this, I edited symbols/colemak to uncomment the two lines and restarted X. Then, choosing different input types from the GUI gives me the following results:
English (Colemak[eD]) : Colemak with DH changes but without angle/wide mod.
English (UK, Colemak[eD], keep local symbols) : Standard Colemak with no mods.
Editing the file doesn't seem to affect national variants for me, unless I am doing something very wrong??
Using Colemak-DH with Seniply.
Lol, those nice bullet characters turned all wonky for you! Knew there was a good reason I shouldn't have used those. :-) Please find the latest version free of such typography.
What you did wrong was to call the scripts with 'sh' - which I now realize that I asked you to do in the instructions so it's my bad. The instructions have now been fixed. And my script flogged summarily. The easiest way is just to run it with './install-dreymar-xmod.sh' which will sort out everything. If you want to specify shell, use 'bash' not 'sh'. Also, I don't recommend that 'sudo' if you're installing locally as then you can't overwrite the local files without sudo later on. My script should sort out if sudo is needed otherwise.
Editing the file is potentially unsafe and may lead to strange results. Possibly even stranger than the ones I put in there... ;-)
The angle/wide mods have to be selected independently from the layout(s), as keyboard models. Using my setxkb script, that's the '-m' switch. For instance, «dreymar-xtools/setxkb.sh -m 'pc105awide-sl'» sets the ISO AngleWide-Slash "model".
Or using the shorthand option, «dreymar-xtools/setxkb.sh -s "5w gb ks"» will activate the pc105-AngleWide-Slash model with the UK(cmk_ed_ks) "Keep local symbols" layout. You can even call that from the install script using a '-s' switch, so you could use «./install-dreymar-xmod.sh -xos '5w gb ks'» for instance. The model shortcut names are two characters – 4/5 (ANSI/ISO) and n/a/w (normal/angle/-wide).
I found your problem I think: The UK layouts would work just fine from the GUI but with setxkb(map) you actually have to use the file name for UK layouts which is 'gb'! So the activation fails if you use 'uk' and you get nada (which in your case was standard Colemak). Is that it?
[Edit: No, that's not it. You stated that you did choose the layout from the settings GUI and not by setxkbmap. Then I'm stumped, as it works fine for me no matter which variant I choose. As an experiment, have you tried making the desired layout the only one available in the GUI (that is, unsetting all others)? Or it might just be an odd result of running the install as sh instead of bash?]
*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***
So, I'm at ~50 WPM with 97% accuracy now, using DBG+HM. Not too impressive but tenable. It feels interesting and somehow right? I love having the G back in place, which feels like an excellent compensation for moving the less common B (1.5% vs 2.0% for G) away from its old spot.
The HM switch is a step up in ambition for sure but it's nice to have this option as it's independent from the rest of the layout and it does seem much nicer to have all these 'HE' n-grams sorted out! That, and the substantial difference in frequency (H ~6.0%, M ~2.5%).
The D is dead comfortable to me in the old B-with-AngleMod spot on a normally staggered board! I don't worry about matrix boards as I don't find them interesting anyway: If I were to get nonstandard hardware it'd be something with a stagger on both sides unless it's split like the ErgoDox. For that keyboard I do see the merit of the DVBG switch over DBG.
*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***