Software applications follow this paradigm:

• Get data from somewhere
• Do something based on the data
• Put results somewhere

Java CAPS applications within the Java EE do the following:

• Get data from somewhere. This usually requires one or more adapters.
  
• (usually) Persist the data, using JMS topics and queues.
  
• Transform the data, using one or more collaborations.
  
• (usually) Communicate and orchestrate intermediate results amongst processes.
• Put results somewhere. This usually requires one or more adapters.

Within the Java EE container, Java CAPS applications can use web services, but they do not need to use web services. However, to communicate with the overall JBI container, web services are required. The services provided by the CAPS application are provided through the Java EE service engine; JBI uses WSDL to access them as web services.

JBI service units and service assemblies do the following:

• Get data from somewhere, using one or more binding components such as the HTTP BC.
  
• Persist and communicate intermediate results, using the Normalized Message Router.
  
• Transform the data, using one or more service engines such as the BPEL SE or Java EE SE.
  
• Put results somewhere, using one or more binding components.

From the point of view of the CAPS application, the bridge to the Java EE Service Engine looks and acts like an adapter. From the point of view of the JBI container, CAPS applications are described by WSDL files and served by the Java EE service engine.