May be you should look at :
glassfish/entity-persistence/src/java/oracle/toplink/essentials/ejb/cmp3/persistence/PersistenceUnitProcessor.java
method computePURootURL()
&
glassfish/entity-persistence/src/java/oracle/toplink/essentials/ejb/cmp3/persistence/ArchiveFactoryImpl.java
method createArchive() this creates the input stream to JAR containing the entities.
If you figure out what the persistence-unit-root URL is and follow up on how the ArchiveFactoryImpl treats that URL, it may point to why downloading is happening.
Hope this helps
-Prashant
[Message sent by forum member 'cprashantreddy' (cprashantreddy)]
http://forums.java.net/jive/thread.jspa?messageID=206800