Multi Dimensional Scaling for Dimension reduction



While reading on use of MDS for dimensionality reduction I came to know that:
1.MDS uses as input a matrix of pairwise distances between the data-points.?Does this mean that the technique can be used for geographical locations only?
2.MDS is equivalent to PCA when the pair-wise distances are Euclidian.What does this mean??
Also if there is any practical application of it in R,kindly guide me towards it(Note:I have already seen the one on R-Bloggers,so I am looking for a more comprehensive one)