At work in our team we're just hitting the merging-schemas problem.
We have an alerting system that uses rdf to describe the alert schedules for different types of event. It describes people and groups they belong to, in addition to their shift rotas etc..
On discovery of an event (e.g. system failure), the system consults the rdf store to decide which people to SMS and email about the alert.
The problem is that the data in the store was created before we standardised a convention for minting URIs. It contains data common to other systems (e.g. people details are also stored in the global company ldap directory, which also has an RDF export), but uses a different schema.
Now that my ifpstore handles equivalentProperty inferencing, we will be in a position to remove the people data from the alerter store, and have the alerter service get data from the global directory instead (which is more likely to be up to date).
The problem is that the alerter service will want results to the query in a vocabulary it understands - the question is, how do we do this whilst still managing the uri-triple-bloat problem discussed in my previous post?
I'm thinking about a system that allows the client to specify graphs of vocabularies that it would prefer to have the results in. When choosing URIs for logical resources, the system could use this to pick more appropriate ones for the client.