Yes, what Dan is saying here is tremendously important: the point here is to make the best engineering decision given our limited resources and the fact that we are trying to remain as much in sync as possible with ACS/Oracle. We can't redevelop a solution for every Java-in-Oracle situation that aD develops (and they *are* developing a number of such modules).
Dan, I'm not at all worried about launching a single, scheduled JVM every 2 minutes. I'm worried about the "more serious" problem, as you mention it. Sockets would be much slower, and would require starting two separate processes, so I would much prefer the JNI-in-module solution. I'm happy to help out and discuss this (on or offline).
