An interesting alternative to interval arithmetic is something called
"affine arithmetic". It models measurements as a center value +/- a
random value - which is basically the same thing as an interval. If
you create two new affine measurements

a = 10 +/- 1 // internally 10 + 0.5 * r1, r1 being [-1,1]

b = 10 +/- 1 // internally 10 + 0.5 * r2, r2 being [-1,1]

then a+b and a-b produce the same results as interval arithmetic.
However, if you subtract an object from itself, e.g. a-a or b-b, you
get 0 +/- 0 because the random parts cancel out.

The nice thing about affine arithmetic is that it allows you to treat
variables as measurements of physical quantities and in many cases
compute tighter bounds on their results.

There's a good description of affine arithmetic here:

http://www.sbmac.org.br/tema/seletas/docs/v4_3/101_01summary.pdf

Jorge Stolfi's project page describing it is here:

http://www.ic.unicamp.br/~stolfi/EXPORT/projects/affine-arith/

This might be an interesting addition to frink.

- Mike