First, let me clarify the title: I think the touchscreen is in the future, I just don't believe that the future of computers is anything at all like the iPad. There's one primary stumbling block that lies in its way and I suspect it is an insurmountable one: the keyboard.

Can you type on an iPad? Sure. However typing on an iPad (I use the iPad as a specific example of this class of devices, but this applies to all tablets) is nowhere near as efficient as a traditional keyboard (never mind Dvorak et al).

When I consider the problems the virtual keyboard presents, I'm reminded of other fantastic technologies that got 80% of the way there, but couldn't get over the hump: OCR and voice recognition. Back in the 80's and 90's, everyone was certain that the last 20% was only a matter of more CPU power, incremental improvements to algorithms, etc, and that these would replace manual text entry in short order. Now they are considered... well they are rarely considered at all, except in highly-specialized applications.

So what would it take for the touchscreen and virtual keyboard (or some other touch-input method) take to get over this hump? When I consider the virtual keyboard, I have a single litmus test: could you write code on it? Yes, you could, but it would be akin to slowly dragging your knuckles across asphalt. You'd bleed and wonder why the hell you were putting yourself through it. So I look at the virtual keyboard and think: the day you can happily write code on one is the day it can replace a physical keyboard.

You might think to yourself, "well, you're a programmer, so of course you think this is the most important test", but this isn't really what I'm saying. What I'm saying is that typing is used to input much more than short messages and Google searches, and that program source code is a good bar by which to evaluate the capability of your input device. If you can easily input code, chances are you can easily input anything.

I can see a potential shift in programming itself to allow for developing code on touchscreens. Gesture-based programming languages seems a promising avenue, although I have difficulty imagining what one might look like or how it could be applied to problems that are inherently text-oriented. Context-aware text prediction might also help, although we already have something similar in IDE's that doesn't seem encouraging (another solution that can't achieve the last 20%). In any case, if some new way of writing code can be developed, the technique may find general application for touchscreen input.

In any case, when I consider the hurdles to achieving efficient text input on touch devices, I think we will be seeing hybrid devices for the duration. It's one of those technologies that seems so close that it cannot fail to reach the finish line, unfortunately the track it is running on appears logarithmic, so the goal is much further away than it might seem.