Tony Hirst of the Open University says:
In the days when this blog was dominated by library related concerns, I used to spend a lot of time working out how to use ISBNs as pivot points for various book related searches; (librarians, of course, don't rate ISBNs - they'd rather focus on the city a book was printed in...). [OUseful Info: OU Course Codes - A Web 2.OU Crown Jewel]
Maybe a little harsh ;-) But it does highlight for me one of the major shifts that needs to take place in our thinking about bibliographic data. My sense is that a majority view in the library community is that bibliographic data is to support discovery and is for display to human users.
However, increasingly important is the use of bibliographic data to support automated processes. Think of resolvers. Think of the growing need to link data from discovery environments (google scholar, next generation catalogs, worldcat, ...) back to various library fulfilment environments. Think of the analysis and collection comparison being done to support digitization or off-site storage. Think of the processing required to support richer discovery experiences (faceting and frbrization for example). And think of future higher level services which build on the relationships in the data (see the related links here and here for example, or Worldcat Identities itself for that matter which does major batch processing of data).
This suggests that we need to be much more careful to facilitate this processing. The growing importance to us of identifiers is one - only one - example of this.