There are so many recent bestsellers in rapidshare and torrents. Hopefully you can convert them to text for your own non-profit use.

For libraries, they will not give you their contents even if they have, for copyright reasons.

Yes indeed, a coding keyboard would look much different from a writer's keyboard. So you can't really have a unified keyboard. I'm designinig a keyboard for general modern typing: communication, writing, short messages - that's why the word `modern'.

Where I would like to `improve' over Colemak is a number of areas:

- Primary finger/hand usage
Your strongest - most coordinated fingers are the index and middle fingers. For righthanded typists the right hand fingers also being much more coordinated. Thus, even on the home row these should be prefered.

- Homerow
My analysis basically confirms the traditionally expected: etaoin shrdlu, with an insignificantly higher frequency of `n' over `i' in Project Gutenberg. So the 10 homerow keys should be: etaoinshrd. The question is location of these keys with respect to the prior point, which I think can be improved.

- Extended characters / punctuation
It is apparent that some extended keys must be in a better position. Namely Return (Enter) has a frequency ranked right after `u', putting it in 13th. With `Shift' it gets even more interesting. Based on capital letter frequency, if there was one shift it is ranked higher than Return, actually higher than `u'. Therefore an interesting idea: put `Return' where `t' is in QWERTY and `Shift' where `y' is, and have them be accessed not in pure touchtyping mode, but both hands. Seems to make sense.

This also applies to backspace.

- Numbers
1 and 0 have a much higher frequency than other numbers, they should be in the middle not towards the edges.

Alas extended characters is where the modern typing sample issue arises. Project Gutenberg gives a 60% higher frequency of comma then period, which is apparenly determined by archaic sentence structure. This might be the case for other characters too. That's why I'd like a good modern sample.

I'd be very interested to hear thoughts / discuss at length my reasoning here.

I'm going to suggest that some of your points bear some research/thinking.  These are not intended as criticisms of your thinking.

This is an assertion which may be true, and certainly fits my intuition.  However, I've stopped short of claiming that I know this is true for a fact.  I have not been able to find any research to confirm/reject the hypothesis that the index and middle are stronger or more coordinated than the ring finger.  Pinky is pretty obviously not.  The only physiological fact I've been able to determine is that the middle and ring fingers share a tendon at the wrist.  While I'm not in disagreement with you, it would be good to have the claim backed up by something more than anecdotal evidence.

This ends up being possibly important because of another, provable effect, that of finger rolls.  It's easier for your fingers to activate from pinky to index than the other way around.  When it comes to the frequently used letters, there's the possibility that a ring finger placement might give more opportunities for rolls to occur in the proper direction.  This depends on whether the digraph frequency has the frequently used letter in front of the alternate letters more frequently than not.  If the ring finger is just as good as the middle finger or even just slightly worse when it comes to strength (coordination shouldn't matter as much since presumably we're talking about a home position), then it may make sense to place the frequently used key on the ring finger.

I doubt this is true.  Again, I've got no evidence.  However, I don't notice any perceptible difference in my left versus right-hand typing.  In any case, hand alternation is a more important metric than putting all the frequently used keys on the right hand.  Once you've grouped according to good hand alternation though, you can choose to put the group which happens to be more frequent on the right hand if you desired.

I think it's a mistake to focus on the home row without taking into account the natural flexion of the fingers.  You should take a look at the Workman layout's rationale if you haven't already.  The top row keys above the index and ring fingers are very close to the usability of the home row keys (in fact, I've seen suggestions that a better home "row" would have your index and ring rest on those keys rather than the traditional home).  If you've seen carPalx's model, which is strongly based on the home row, know that he thinks this may be a weakness of the model.

That's all I've got for you.  Just some food for thought, I'm sure you're on a good track for your needs.

@lilleyt
That was a great post, and I do agree that we can't take intuition and even so called expert research for granted. For example, many keyboard experts say that the bottom row should be avoided. That is, they have a metric that automatically groups all bottom row keys as the least desirable, so the least frequent letters would be put there. However, as I came up with my own effort model and developed a layout based on such a model, and even putting two frequent letters on the bottom row, I found that the layout is just as efficient as other popular layouts, and in some areas exceed them.

Here is the effort model that I came up with based on personal experience:

Lower values mean less effort. Actually the keys should be columnar, but I can only a find a staggered template image. So that's why the values are symmetric. Nevertheless, as you can see from this chart, that two keys on the bottom row actually use less effort than all the keys on the top row and even two keys on the home row. They are the index keys on the bottom row. This defies what many experts have claimed. To test this, I put two frequent letters, H and D on these keys. I quote from my results using an online layout test tool (my layout is called Balanced):

There are other similar findings from this chart that would seem unintuitive. Like that two side index keys on the home row actually have higher effort than some of the keys on the top and bottom rows. That's because the index finger is better at bending inward toward your body then stretching side to side.

Yeah, true. What I mean though, is break the touch-typing rules and have both the left and right index fingers hit the shift key, based on which side of the keyboard the shifted key is. There is a medium stretch when hitting with the left index finger, but seams feasible to me...

Well, exactly because of this I was interested in the extended character analysis. And it turns out that only 12 letters get used more often than the Return key and most other letters, much less so. So it seems to me Return deserves a comfortable position - not a pinky one. Currently I'm actually leaning towards 'V'.

Yes, my thoughts too, however the only key available then is the double-quote key, as the regular Return is way too far. And I'm just thinking that since Return is such a heavyweight, it should be operated by the strongest finger, not the weakest.

I'm currently experimenting, typing on it already, with a third version of such a layout already, once I settle on what seems a good solution, I'll follow up here, to have it subjected to criticism...

Different weight models bring different layouts. Even for one fixed model there would be several thousand layouts that fits the bill.

Although with all good rollings and key combinations, as a Colemak user I am only 98.7856235534% satisfied with Colemak, but somehow that's all right.