This is a great question, and one that we are just debating.
Can you flesh out your requirements a little bit more, and illustrate them with an example of a sequence of steps which shows how the reference data (e.g. the Legal Entity Name) would be queried and used?
The solution might be distributed (e.g. all necessary nodes have a copy of a Corda state that includes reference information, and nodes can query from their vault) or quasi-centralised (e.g. data is available from a network-wide service like the current Network Map service, which could be queried).
How many different types of reference data are you thinking of?
Is there a central existing authority for the reference data you are thinking of, or is it determined consensually by the participants?