FULL PRODUCT VERSION :
Java SE 6

A DESCRIPTION OF THE PROBLEM :
The problem is that the amount of network traffic has increased up to 25 times (20KB vs 578KB in one operation) compared to java 5. This has been traced down to every RMI request sent from our client to our server containing the list of all jars in the classpath for the application.

We can provide tcp dumps of a single (small) TCP stream conversation as well as have a test app to reproduce and test the problem.

The dumps clearly show what extra data is in each request, and it's clearly the whole classpath as specified in the JNLP file.

You can grab the dump files from these two URL's, and this shows the size of each RMI request on the wire.


STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
Here are 2 test links, it is our app with an automated front end that runs the same test every time. The problem occurs after our app is loaded so our test procedure for us is...

   1. Load the app
   2. Start Logging
   3. Click "Start Test" button in app
   4. After "done" appears stop logging

The numbers refer to the version of java (jre) that the app will run under, not the webstart version.


As mentioned in the ticket this is a problem with javaws 6 so both jnlp's above should have the problem when launched with javaws 6 as opposed to javaws 5.

In our testing we looked at the files that are linked to java.exe when our app is running (after being webstarted). We can see that deploy.dll is from which ever version of webstart that is used. So ws6 + jre5 = deploy.dll 6 where as ws5 + jre5 = deploy.dll 5, I'm assuming, but haven't tested, that this is the same using jre6. Hope this info helps.

ACTUAL -

temp/java6.dump | wc -c
23545
java5.dump | wc -c
889


REPRODUCIBILITY :
This bug can be reproduced always.

Release Regression From : 5.0
The above release value was the last known release where this
bug was not reproducible. Since then there has been a regression.