Various of my OCLC Research colleagues have long thought that a work-based approach to bibliographic data would better serve user needs. The ability to enter an information space at the work level, and then to drill down to the various editions, translations and so on that manifest that work has seemed preferable to the presentation of a list of those manifestations without any explicit note of how they are related. This is especially tedious if such a list is long. Think of a search for Hamlet in a large bibliographic database for example.
For this reason, we have been very interested in the (slow) development of the FRBR model.
Experiences with the data we have made available through the search engines seem to validate my colleagues' view. A work-based approach has at least two possible benefits. Firstly, it would reduce the scatter of relevant results through a large result set. Secondly, some clustering would potentially increase the 'weight' of a particular resource in ranking approaches. Our search engine partners certainly see the advantages of the FRBR work-based approach.
Note: I describe our interim approach to providing some work-level clustering in another post.