When working with different coauthors, I have always had some troubles in using bibtex effectively. Everyone has its own set of bibtex entries, with keys in their own favorite format --- for instance, the same book could be indexed by different people as as golubvanloan, GVL96, matrix_computations, GolubVanLoan, GolVL96, mc-book...
I usually solve the problem by creating more or less a new .bib file for each paper, where each author copy-and-pastes entries from older files, and everyone keeps the keys he cites in their favorite format. Of course, this is not efficient. For every paper, a nontrivial amount of careful manual work is wasted adjusting the bibliography (e.g., pruning double entries).
What is your solution to this problem? Do you have shared .bib files, or common conventions across whole communities devoted to a single area of mathematics? Is there maybe any $\LaTeX$ command that lets you register double keys and refer to the same book with different keys? Is there a sensible method to avoid duplicate entries in the .bib files? Should Bibtex be dumped in favor of something better?
