information efficient text enty using dasher

[Chris Ball]

Hi,

I'm one of the Dasher developers, and am also interested in hacking on
OpenMoko.  So, getting Dasher going is fairly likely.

   > This pretty much means that you have to stare at the display all
   > the time when inputting text.

Yes, this is the main difference between Dasher and T9.  However, the
comments about needing a lot of screen resolution or CPU aren't so true
-- we did Dasher on the iPaq seven years ago at full-speed and using
150x150 resolution, and it works great.  The reason we get away with
not so much resolution is that you're only really ever being asked to
choose between five or so probable letters at each turn, and it doesn't
take much screen space to show those, and you can predict whereabouts 
you're headed by knowing the alphabetic order of which character comes
next.

   > Sure - in theory, dasher may approach arithmetic coding in terms of
   > information input. 

(I'm not sure what you mean by "approach" -- Dasher *is* an arithmetic
coder, and matches the information-theoretic efficiency of one in 
terms of bits/input to characters/output.)

   > But unless you can do the coding in your head, you've got to stare
   > at the screen, making it less useful for environments where you've
   > got vibration, sunlight, walking down the street, or less likely
   > for a phone, if you're blind. 

Yes, but the Neo doesn't have a keyboard, and doesn't have keys for T9
that you can use without looking at the screen, so I don't think this 
is a useful criticism.  Dasher's very tolerant of vibration and mistakes,
unlike T9 on a touchscreen -- it's much like driving a car, in that if
you oversteer or understeer you just correct yourself later, because
it's all about navigation and where you end up.  We can type easily over
20wpm on the iPaq with a touchscreen and stylus.

Thanks!

- Chris.
-- 
Chris Ball   <cjb at laptop.org>
One Laptop per Child