clustering

Does k-means make assumptions? Its drawbacks explained

Why k-means, though 'just an optimization', behaves as if clusters were round, equally spread and equally sized, and what happens when they are not.