Expose parameter to not write library-directory element in application.xml

I am confused. library-directory is only written if you set the defaultLibDirectory. If I understand you well, you're saying you must use it (to use the magic APP-INF/lib thing) but we should not write the value in that case?

I believe you're asking here to workaround a bug in Weblogic, right. Has this been reported by any chance?

You've got it exactly. Definitely a bug in WebLogic. I am working on putting together a bug report with Oracle, but from past experience, if they don't first tell me this is intended behavior, they won't release a fix to others since it's not a security flaw.

In that case, I prefer a weblogic configuration option, like we've done for JBoss. Your patch is a too wide option that sole purpose is to workaround a bug.

Something like

<configuration>
  <weblogic>
    <write-library-directory>false</write-library-directory>
  </weblogic>
</configuration>

Well, fair enough, I'm not too happy about an option just for working around a bug.

But you did bring up the point that library-directory would only get written if defaultLibDirectory is set. Can I ask, is there a reason why JARs are written to the root by default? From http://java.sun.com/xml/ns/javaee/application_5.xsd:
"If this element [library-directory] isn't specified, the directory named "lib" is searched."

Should jars be written to lib/ by default instead of the root? It seems to be the expected location according to the spec and would seem to eliminate the need for defining defaultLibDirectory in most cases.

Well, I think you got it wrong or I am missing something. Your request IS about working around a bug!

The answer is backward compatibility. This option was present way before JavaEE5 came out (actually it was mostly used on old weblogic version and the magic APP-INF/lib thing). Changing the behavior of an existing option according the version of the spec you use is a bit tricky.

The other reason is that users (for mysterious reasons) keep on willing to put libs in the root. If you change the default, it's harder to justify the behavior breakage.

I was agreeing with you about it being a workaround for a bug; I'm fine with rewriting the patch to make it be listed as for WebLogic only. I think I should just wait for Oracle to get back to me about this before I do, in case they have a more appropriate fix.

k