Comment by tehbeard

It just needs to download the pages (x KB chunks) to traverse from the header to the particular index and table schema, a fraction of the whole data.