Canon's compensating motion is rotational, not horizontal/vertical
as Minolta's system. Canon thus needs lesser motor movement and
much lesser force (if the element is balanced correctly) to acheive
the same amount of compensation.
When rereading this I see that it was very bad wording.
Canon sees camera shake as a rotational movement. They use a moving
lens element to "bend" (or counter-rotate, seen from the side) the
light coming through the lens. For this, they need pretty small
movements; even a minute adjustment will have a large effect on the
film/sensor. Minolta sees the shake as an up/down motion and uses a
counter-moving sensor instead, which works perfectly fine for a
small chip but is harder to implement with a larger chip. Both
approaches have their advantages and disadvantages.
Canon also uses (mostly in their camcorders, I believe) another
solution with two parallell glass plates with a high-refractive
liquid between. By varying the space between the plates at
different parts of the edge, correction can be applied by bending
the light as of above; this is an interesting system that Minolta
have researched (and patented!) extensively. In the Minolta
version, stacked shape memory alloy is used to vary the distance
between the plates in a fast and precise way. The advantage, as I
understand it, of such a system is that it can be easily
incorporated inside a camera -- especially one with EVF.
