I agree that a lot more needs to be understood and imagined about how we may use the touch devices, but radical changes don't have to be radical on day one. Indeed, from experience radical changes fail precisely because users can only absorb so much change in one go.
As it turns out when you take the touch interface into account as our iOS 7 app does, you can do some sweet things. My personal take is that you experiment with the interface _while_ making sure you don't break the existing ways of inputting (= generating content). If you shove a lot of change down people's throats in one big spoonful, you alienate them. Evolving the UI means you can take user feedback and adjust the course on the way. Radical changes mean you assume you know exactly what they want and need (and you don't mind breaking their workflow to impose the new way). That's a big assumption and may appear a little presumptuous to users who've developed a workflow around your app.
We may not get it right completely in one go, but we can assure our users that the app is the product of their feedback and will continue to be.
Can't, or don't? Repeating keys isn't all that useful, and a different behavior could certainly be implemented in software.
It actually doesn't work well in TeX because it depends on having the correct input encoding set, which still breaks when you have multiple collaborators. Though \:e for ë is pretty easy to remember.
Language-specific autocorrect is common in desktop software that has autocorrect at all (e.g., wordprocessors) and, similarly, good access to international characters through methods faster than press-and-hold is pretty common in lots of desktop software.
Press-and-hold is a slow, clumsy approach that's only tolerable with a virtual keyboard because virtual keyboards are already so slow and clumsy that its not a big marginal cost.
I find myself using the Character Map (or Keyboard Viewer) all the time, because the keystrokes required aren't intuitive, and aren't labelled on my physical keyboard.
So for some users, press-and-hold on "e" to get é is far more efficient than having to open up a separate application entirely.
For that matter, I really like OS X's approach where you can use either input method.
Likewise, you can always switch to the specific language's keyboard on iOS, and press and hold is similarly obviated.
I don't know, I've always found the Word keystrokes intuitive (in fact that I was able to discover most of the ones I needed for French and Spanish by learning one and generalizing from there), and my main frustration in that regard is that they aren't standardized for text input in Windows.
As for labelled -- if I have to look at my keyboard to type, that's too much of an efficiency killer to start with.
> Likewise, you can always switch to the specific language's keyboard on iOS, and press and hold is similarly obviated.
Except that its generally not; even with the Spanish keyboard, you still need press-and-hold to get to the vowel accents.