Ridiculous, perhaps, but it's been that way for ever. Keyboards are not well adapted to human input; humans adapt to keyboards. Automotive controls and even writing systems themselves are shaped largely by implementation cost. Adapting a human is much cheaper than adapting a machine.

Of course, technology has advanced since the first Palms a long way, and it's to be expected we can come up with something with less adaptation on the user's part. But to me Graffiti seems much better than the HWR on the tablet, because it does one thing well. It uses one simplified (for greater recognizability) alphabet. And yes, you do have to learn that, just like learning touch-typing.

But the NITs' HWR tries to match many slightly different alphabets, each with less differentiation of characters, and fails. So you wind up with the need to re-train yourself anyway, if only for the worst few letters. If you can't get genuine adaptive HWR right, you shouldn't do it.

I'm not persuaded that the Newton's way of letting you write anywhere is such a great one; I'm not persuaded against it either, never having had a Newton. (I do know it, even confined to the dedicated area of a hildon-input-method, would spank the Nokia's HWR.) But one thing I can say against it: It inhibits porting of programs from "normal" OSes. When programs expect mouse and keyboard input, stylus and some dedicated text input method is one of the easiest ways to make that work. Maybe it should only be a fallback system, but for a device I use the way I use my N800, it must be there, and it must work right; hence my use of the finger keyboard at present. If graffiti or similar was available, I'd almost certainly switch.