with JBoss 4.0.x we used the class org.jboss.deployment.scanner.PrefixDeploymentSorter to get a deployment order which was idependent of the file content. With JBoss 4.2.x the filename is the first part of the global JNDI-Name. This is a very goog idea, because we have the version number in the filename and with 4.0.x we had to fiddle around with the deployment descriptor to get a version number in the path. No we get it without doing anything. Especially for EJB3 this is very useful. But now the number prefix for the sorter is part of the global JNDI-Name, too. If we had to change the ordering, e.g. add an additonal layer, the JNDI-Name changes, too.
To avoid this I wrote a DirPrefixDeploymentSorter, which sorts according to the prefix in file and directory names. It sorts the whole hierarchy under the directory which was given to the scanner. Are there any interests in the code? Or is there a simplier solution for the problem? Another solution would be to strip the prefix in the deployer, but I think this has a bigger impact on the server code.