From reading UC's Bibliographic Services Task Force report on "Rethinking How We Provide Bibliographic Services for the University of California," and participating in an initiative to write a similar white paper at MPOW, I've been thinking a great deal recently about what people mean when they talk about the future of the "catalog" or the "OPAC" in libraries.
Many people, when referring to the future of the catalog, mean the future of MARC. These arguments tend to center around how we can adapt the MARC record to handle new types of materials. Others mean the cataloging/metadata creation system present in the library's Integrated Library System. Many vendors (and OCLC) are talking about including metadata formats other than MARC in these systems. This sounds like a reasonable idea on the surface, but given the track records of ILS vendors, I'm not holding my breath for this one to work out very well. Another common usage is to mean "those things which the library owns," but this model has become problematic with the advent of licensed and free online resources, so this meaning is falling out of use.
I do think we need to figure out what systems locally-created metadata will go into. However, it's not realistic to expect we're moving towards an environment in which everything our patrons want access to is in a single database. As I'm fond of saying in this context, "That ship has sailed." Consider article-level access to the journal literature as the "elephant in the room" example of this phenomenon. Many vendors provide databases for this purpose that we happily subscribe to. It would be madness for libraries to try to replicate this information. We need to focus our attention instead on systems to make all the various information sources (including the catalog!) work together to provide seamless access to our users. Federated searching products on the market today are a step in this direction, but I've been decidedly underwhelmed by their functionality. We have a long way to go, one step at a time.
After all of this, I'm still not sure what my definition of "catalog" in ten years will be. I toyed briefly with the idea of "metadata records we created locally," but our current models with my library having a local copy of a shared record don't really fit with that definition. It could be something more like "records we manage locally" but that seems to administrative to be useful to anyone other than ourselves. Perhaps we should just bite the bullet and call the as-yet-still-imaginary single front end to all resources of possible interest to our users the catalog.
We're starting to tackle these issues in a big way, and I hope we can continue to make progress by agreeing on some semantics so we're not constantly talking past each other.