Never fun to have a mistake pointed out, especially when it is not fixable, but that's what happened with section 3 of this paper. Fortunately, when people have cited this paper, they've only used results from the first two sections. On the other hand, it *was* fun working out how to recreate the way the paper had been formatted by the publisher of the proceedings volume in which it appeared... This version looks like the published version, with notes in the margin. https://arxiv.org/abs/math/0002232