Section 12.1 of the spec states:

When determining which archives are bean archives, the container must consider:
• Library jars, EJB jars or application client jars
• The WEB-INF/classes directory of a war
• Directories in the JVM classpath

For CDI 1.0, I think the inclusion of "directories in the JVM classpath" would not be a big deal. However, since the introduction of implicit bean archives, I'm wondering if implementations are required to scan the JVM classpath for classes annotated with bean-defining annotations. This seems like it has the potential to be an expensive requirement, and it should not be required.