RE: system-wide JVM (threads) Was: Friendly co-existance

> From: Fred Kulack
> 
> That's very interesting.
> 
> Though I've never implemented nor had a need for a solution like this,
I
> wouldn't
> have wanted that as a solution (accessing internals of one job/JVM
from
> another) because
> although it gives me power, it violates some isolation principles that
I
> sort of prefer.

Couple of things: first, I'm not married to this issue.  Unlike some
people you see on the lists, I'm pretty comfortable skinnin' the cat
whichever way works.  I am a big user of data queues right now, the only
issue being those pesky QZRCSRVS jobs that get left hanging when a
server abends.  To be honest, that's really my only gripe with the data
queue method, and it's the same in any environment.

Also, from a security standpoint, note that I said "register the JVM".
That implies that the default action is to be not registered; i.e., not
accessible to other jobs, just as Gosling intended.

However, and here's the kicker, by having the CAPABILITY, you set
yourself apart from other vendors.  It would be a perceived advantage of
the iSeries and I think it might have some legs in marketing.
Especially if you allowed such a thing between partitions - could you
imagine?  A JVM shared by OS/400, Linux and AIX?  There are some
SERIOUSLY cool ramifications there.

But once again, I can do it through data queues or sockets or some other
less elegant method.  But the ability to use the JVM to unite the
various forces in your network - sort of an uber-RMI - that's gotta at
least tickle your fancy a little bit. <grin>

Joe