If you are interested in the generalisations, there is a good case for starting from sheaves because the theory is slick. (Probably also true for R-R for vector bundles on curves, which Weil himself handled with bare hands c.1938, but makes more sense in Hirzebruch's book).