This free ISBN database is interesting. We've observed that in our user testing a number of students recognize the ISBN as the "right" way to search for a particular book -- it's their known item search key. (We're not really set up for known item search, but we'll probably tune RedLightGreen to do known item searching using ISBNs only.)
I was at a meeting on Friday that discussed something we couldn't quite label: ONYX for libraries might be one way to express it, bibliographic record enrichment might be another. The LC has some "publisher descriptions" of new books up. It would be lovely if these could be retrieved in XML form boy LCCN or ISBN. Both a project like this ISBN catalog and like RedLightGreen could benefit.
I continue to ponder, though, whether there's a way to leverage this information up to the work-like level. Maybe there's some cut-off: if there are less than N (where N is like 10) editions (or RLG clusters), assume tables of contents and such are useful across all editions. If there are more than N, point to those specific editions for detail.