If I understand the question correctly, you want the arc generated by a center point $C[c_1;c_2;c_3]$, a start vector $u[x_1;y_1;z_1]$ and an end vector $v[x_2;y_2;z_2]$.
The starting point of your arc is therefore $C+u$, and the end point $C+v$ (thus forcing $u$ and $v$ to have same norm). Let $\alpha$ be the angle $(u;v)$
Your arc is defined by the set of points $C+\dfrac{\sin(\alpha-\theta)u+\sin(\theta)v}{\sin(\alpha)}$, for $\theta\in[0;\alpha]$. Using values of $\theta=\dfrac{k\alpha}{n}$ should give you the points you need to approximate the arc with $n$ segments.
Edit :
Now that I have written this I realize that this does not work for $\alpha=\pi$, witch is to be expected as in 3D, two opposite vectors and the center of the sphere do not define an arc (as they don't define a plane). It does work otherwise. Given the midpoint instead of the center, one can find the center and make this work for all cases where $u \neq -v$. I'll look into this tomorrow, as well as computing the above formula using the cartesian coordinates.
Edit2 : after giving it some more thought, I think the best way to define an arc is giving the center of the sphere, the start vector, and the mid vector/point. That way the two vectors you are using are not colinear (thus defining the arc uniquely), unless you are trying to draw a circle. With those inputs, the above parametrisation should work with minor fixes. But first, some much needed sleep.