The idea of a Common Information Model to rule all within a business has been around for a long time. It's gained a lot more traction recently with SOA. Here is a post from Nick Malick at Microsoft and some of their efforts. His post is part of an ongoing conversation with Alex Maclinovsky at Sun.
They both bring up good points but I find myself leaning towards Alex's realism a bit more. I can bring up the point that the average business IT shop does not have a lot of luxuries when it comes to time. CIM's take time... a lot of time. They also take knowledgeable data architects, expert negotiators and the ability to bend the space time continuum. Simply out of reach to most IT shops.
Would a large model that rules all hold up to the daily demands of the constantly changing business? I know the academic arguments around versioning, governance etc but that within itself can be difficult even with a governance product. By the way Microsoft has a new product in this space, only for MS products though. :(
I've stated before that one size doesn't fit all meaning what works for my organization might not work for yours. I have found that smaller simplistic data structures combined with more granular services can achieve a lot of flexibility and reuse. These can be wired together to make more coarse grained services but the reuse tends to go down and the complexity goes up.
Ultimately data translation and manipulation has to happen anyway to satisfy the needs of the off the shelve systems as well as old legacy systems. Some common simplistic data structures can help with that without having to go for the big bang approach that a CIM can create.