As I understand, with Jena adapter java interface, the client can connect to Oracle RDF database directly. Why does it still need SPARQL web service endpoint? Does the query request from Jena adapter at client side have to go through SPARQL webservice first then route to Oracle server? Or the query with SPARQL need to go through webservice and non SPARQL query directly go to Oracle server?
You are right that a client can connect to Oracle RDF directly through the Jena Adapter. The reason for providing a SPARQL web service endpoint is to give additional interoperability and flexibility. With the SPARQL web service endpoint, one does not even need to write any Java code to interact with the underlying database. It also makes SPARQL federation a lot easier.