Why should features often be scaled before k-means?

A. Scaling automatically chooses the best number of clusters B. Scaling supplies the missing class labels C. Large-scale features can dominate the distance and centroid calculations D. K-means can operate only on values between zero and one