The fact remains that people want a device that they can wear on thier person that offers a reasonable way of accessing, entering, and modifiying text. There are only 3 technoologies that, even in the mid-term future, offer any chance of making this happen.
1.) Tiny keyboards
2.) Handwritting Recognition
3.) Voice Recognition
Number 1 sucks, has always sucked, and will for ever after, suck. They are fine for short text messages such as a crackberry.
Number 3, even if it is 100% perfect, is not suited to a myriad of situations and scenarios.
Number 2 is the only viable text input methodology and only 1 device, ...wait for it....
"The Apple Newton Message Pad"
....ever delivered on this feature.
For some strange reason, most people who say they want a tablet, suggest that they want one so they can take notes in meetings. I would suggest that it is never actually the case that these are effective note-taking devices. From everything I've seen, the most realistic usages are the following:
1. Mobile snapshots of your day planner
2. Capturing small one-line "To Do" tasks from meetings
3. Mobile snapshots of your address book
4. Capturing new address book information when away from your computer.
5. Viewing all manner of information, video, photo, text, document, etc that doesn't require a large screen and was probably put there by syncing with your computer.
Taking notes during class or at meetings just doesn't work on these devices. Everyone who tries eventually gives up and goes back to pen and paper. But if you were to make it ridiculously simple to create small one-line to-do tasks from a meeting that would end up in your calendar, and to capture contact information from someone that ends up in your address book, while giving you view-only access to a large variety of information, then you'd have something usable.
Most of the one-line stuff you'd like to capture would probably lend itself very well to voice input, but then you couldn't capture it _during_ a meeting (since someone suddenly talking into their UMPC would be a bit distracting). Typing is equal distracting (click, click) and so the most subtle form of input remains handwriting recognition. In general handwriting recognition is very, very poor (even 99% success means around 1 error every sentence or so - completely unacceptable from a general use standpoint). However, if the recognizer understood the _context_ of what you were writing, success could go up quite a lot.
If apple produces a UMPC, I hope it is focussed at a particular set of tasks, with the capacity to do more general stuff, but optimized for specific things.