XO-2

[Chris Ball]

Hi,

   > Will be cool if we can take dasher and run with the concept. For
   > some languages and usage patterns it makes a lot more sense than
   > keyboard input.

Given that we employ someone who worked as the lead Dasher developer
full-time for two years, it's quite possible.  :)

   > (OTOH, it's extremely awkward for other stuff - like programming!
   > And I do suspect that dasher is a big battery hog too.)

Yes, repeatedly calculating probabilities for nodes while you zoom in
uses lots of CPU.  That said, we had it running totally smoothly on
100MHz iPaqs five years ago..

The main problems with Dasher will be:
  * the visual/cognitive load from the zooming metaphor
  * the time-critical nature of typing (not present at all with an
    on-screen keyboard)
  * the requirement to understand tree structures -- ie. "I want to
    write 'chair', so I have to find the r inside the i inside the a
    inside the h inside the c, and no other path containing the same
    letters will do."  This kind of hierarchy turns out to be a way
    of thinking that doesn't come easily to many people.

For those reasons, I wouldn't encourage Dasher as a panacea here.
As an optional input method, though, it can be much faster than hunt
and peck on an on-screen keyboard after some practice.  Here we
compared an eyetracker with Dasher to an eyetracker with an on-screen
keyboard, finding that Dasher was faster with fewer errors:

http://www.inference.phy.cam.ac.uk/dasher/presentations/Closing05/mgp00021.html

- Chris.
-- 
Chris Ball   <cjb at laptop.org>