Jeremy (to clarify),

Instead of having configuration data for an EJB living in only one namespace, 
why can't we create appropriate namespaces for the particular requirements for 
each EJB.  Some examples would be:

geronimo.<AppName>.<EjbName>.jndi
geronimo.<AppName>.<EjbName>.pool
geronimo.<AppName>.<EjbName>.transaction
geronimo.<AppName>.<EjbName>.security
geronimo.<AppName>.<EjbName>.methods.<MethodIndex>              (for Session & 
Entity)
geronimo.<AppName>.<EjbName>.queue                                      (for 
MDBs)
geronimo.<AppName>.<EjbName>.timing                                     (for 
Timed Beans)

(Note if speed is an issue, we can create the following to obtain all 
configuration data in one call - generally used at startup)
geronimo.<AppName>.<EjbName>.all

This hopefully will give us the freedom to maintain easy configuration 
interfaces, rather than trying to come up with a generic inheritance model that 
will suit all.  (Designed well, this should hopefully reduce code - due to 
reuse - not only in creating MBeans, but in creating behaviour on this model).

The population of this information can be performed in any manor (sequential 
for SAX parsing), in EJB blocks (from binary storage) or randomly (from 
properties file / deployment tool).  We are thus not locked into any 
configuration persistance model leaving flexibility to use any approach - the 
algorithm for loading configuration data, will also have an appropriate 
mechanism to persist the configuration data.

Regards,
Daniel Sagenschneider

Reply via email to