There was a problem in the RMI-IIOP implementation in this situation that caused very
large requests to be sent in this situation. This was caused by the very long classpath
in the app server being set as the codebase (from RMIClassLoader.getClassAnnotation,
I think). Another error in the encoding caused this codebase to be sent on EVERY
java object, rather than just the first. Effectively the IIOP compression was not working
in this case, resulting in messages often 10x larger than required.
If this is the problem, you should see a lot of repeated class path entries in the data
that is sent. This problem was fixed in GlassFish version 2 several months ago
(probably in build 30-something), but I can't find the exact build number right now. Can you try
this with the most recent GlassFish v2 build?
[Message sent by forum member 'kcavanaugh' (kcavanaugh)]
http://forums.java.net/jive/thread.jspa?messageID=218450