Actually, I didn't think about optimization. In case of jboss.xml, every bean there can be identified by the ejb-name, other metadata components that need filling in some details can also be identified by their names. So, parsing jboss.xml alone we can get EJBs and then merge them with the spec ones. In case of annotations, it's different in the sense that looking only at the jboss annotations we can't identify an EJB or any other metadata component. I.e. the JBoss5Creator would still need to look at the spec annotations. So, I thought why not combine the processing of both. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4149233#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...