OWL and DBs have contrasting strengths and weaknesses. Combine the two. Integrity constraints cannot be modeled in ontologies. Users often mistake certain OWL features for integrity constraints. ABoxes and database instances are both relational structures. Databases perform query containment and optimization. They break queries into smaller queries. Queries in both ontologies and RDBMSs are first order open entailment problems. The only difference between the two is RDBMS uses closed world assumption and ontologies use open world assumption. Integrity constraints are used to express closed world assumptions. In databases, query is only considered on database instances. In ontologies, schema is also considered. Inferences come into play. OWL is an incomplete database. We need to extend ontologies with integrity cosntraints (ICs). The steps are: 1. Model an OWL ontology 2. Mark certain axioms in the ontology as ICs 3. Add ABox data 4. Check constraint satisfaction 5. Remove ICs 6. Answer ABox queries This actually improves performance because only valid assertions can survive the IC check. OBSERVATION: This seems to be very preliminary work. It was mostly stating the obvious. In fact, it is possible to model some ICs; the UNIQUE constraint in RDBMS is equivalent to the FunctionalProperty construct in OWL. Primary and Foreign Key constraints can be modeled using existentially quantified restrictions on object (and maybe, datatype) properties. Enforcement of constraints will be an issue because OWL uses OWA. How things such as DELETE CASCADES and so forth can be handled is something to be thought anout.