Posted by scwells72
on June 25, 2008 at 10:52 AM PDT
I'm playing around with converting a Java Web Start application into an applet for a more seamless Web-based experience using the next-generation Java plug-in. Generally, so far, so good. I love the deployment toolkit and the ability to use JNLP from an applet!
Because our application can be used from remote installations with significantly lower bandwidth between the client and the server, I've been using a throttling proxy server to test download, startup, and typical use case times. Startup and use cases are in general faster from the next-gen applet, but download times are significantly worse!
For example, in 1.5.0 using Java Web Start, our application's download time (18MB worth of jars) in a simulated remote data center environment is 4m17s. Using 1.6.0_10, the download time for a JNLP-based applet is 6m23s and the download time for the same app deployed using Java Web Start is 6m39s. For the last example, the ONLY difference between the two test cases is 1.5.0 vs. 1.6.0_10, so it does look like a significant regression to me.
I'm posting this in hopes that the folks from Sun monitor these forums and will see this as an issue for the next-generation Java plug-in to be resolved before its release. If anyone knows if there's a better place for me to share this information in order to raise its visibility, please let me know and I'll post it there as well.