follow up ...

[User Hayden <bob statland org>]

Back in the days of DOS there was a solution for folks who wanted
floating point calculations to act like calculations done with pencil and
paper algorithms.  It was called Binary Coded Decimal.  Instead of
representing the entire real with a base 2 float it represented each
digit with a binary integer.  It then did the usual pencil and paper
algorithms on those digits.  I am sure the code is still out there if
anybody needs it.  There are also free arbitrary precision calculators
(at least for Linux).  And the TI graphing calculators are largely
free of binary artifacts, though I suspect they use floating point
arithmetic and then trap anything that looks suspicious before sending
it to the screen.

More generally, much of the traffic on this list asks why Gnumeric
does not do some highly specialized task.  Gnumeric is a general
purpose program for non-specialists.  If you have specialist needs use
specialist programs.  
-- 

      _
     | |          Robert W. Hayden
     | |          5 Howard Street, Apartment 206
    /  |          Wilton, New Hampshire 03086  USA
   |   |          
   |   |          email: bob@ the site below
  /    |          website: http://statland.org
 |   x /          
 ''''''