What is the intuition behind the use of the so-called "kernel trick" with support vector machines?

A. It is sometimes possible to assume a Gaussian (multi-dimensional normal) distribution over the data, and use that instead.

B. It is often easier to use a dot product than it is to use a linear classifier.

C. For data where a linear classifier is not a good match, if we map the data into a higher-dimensional space, a linear classifier can still work well.

D. None of the above.