If a curve is described parametrically as <math>\vec{r}(t) = x(t)\vec{i} + y(t)\vec{j}+z(t)\vec{k}</math>, the vector <math>=\vec{v}(t) = \frac{d\vec{r}}{dt} = x'(t)\vec{i}+y'(t)\vec{j}+z'(t)\vec{k}</math> will always be tangent to that curve if we imagine the basepoint at attatched to <math>(x,y,z)</math> instead of the origin.
+
+
However, when people speak of the tangent vector to a curve, they almost always mean the '''normalization''' of <math>\vec{v}</math>, which means a vector that points in the same direction as <math>\vec{v}</math> but with constant length =1. We can compute this '''tangent vector''' with the formula
This is not the only vector which commonly arises in the analysis of curves: we often also wish to have a description of how the curve is turning at any specific point. Since this amounts to knowing how the tangent is changing, it should not surprise anybody that we examine the derivative of the tangent for this purpose. Frequently in applications, we find it doesn't matter how long this derivative is, only it's direction. Hence we define the '''normal vector''' using the same process as above:
Since the tangent vector <math>\vec{T}</math> ''always'' has unit length, it only change direction, not length. This means that the derivative vector <math>\vec{N}</math> must always be perpendicular to <math>\vec{T}</math>, since if it were not, there would be a component of <math>\vec{N}</math> in the direction of <math>\vec{T}</math>; that is to say, the length of <math>\vec{T}</math> would be changing.