The Merits of Unification

With unification, databases can become part of the same information system by virtue of their sharing a common underlying data model. Entities in different parts of such a data system are of the same type because they have the same semantics (i.e., meaning ). 

Consider a molecular structure, which is the most important unifying information model in chemistry. Molecular structures appear in knowledgebases that represent catalogs of commercially available chemicals, pharmacology of named drugs, natural sources of bioactive molecules, protein-ligand interactions, measured molecular bioactivities, metabolic pathways, abstracted research literature, databases of synthetic reactions, and so on. 

With federation, databases can become part of the same information system because they share common communication language(s). Each database in a federated data system is free to establish and maintain its own data model. The utility of such a system is critically dependent upon the ability of each participating database to process queries using common protocols. It is only required that databases in such a system understand the languages that are appropriate for their local data model. It is not required (nor particularly desirable) that all databases understand all query languages. 

Communication between federated databases is much like communication between humans. If you ask someone what they know about Casablanca , the answer depends on what kinds of things they know about it - even more critically than, for example. 

For certain kinds of data sets, only the molecular structure makes sense, for example, databases of synthetic reactions. For a few others, only the salt form makes sense, for example, when you buy Lipitor or dispense it to a patient, what does 100 mg of Lipitor mean For the remaining, the molecular and salt forms are synonymous. 

---