RE: Activation groups - named vs. *NEW

> From: Mark Waterbury
>
> I don't believe that a *SRVPGM creates a new activation group,
> but perhaps if the *SRVPGM has ACTGRP(*CALLER) and
> the caller happens to have ACTGRP(*NEW), this might happen?

I think you're right about this, Mark, but that's not the issue.  I haven't
even begun to deal with all the issues surrounding AGs and service programs.
I was talking about the various OS/400 host servers, which I use a lot in my
applications.  For example, I believe that QZRCSRVS and QSRCSRVSD together
handle program calls.  I believe QZRCSRVSD is the daemon that listens on the
TCP/IP port and submits QZRCSRVS jobs as necessary.

The issue is down at the QZRCSRVS level.  If I create a connection to the
host (one way is to use the JTOpen toolbox) and do multiple calls to my own
API program, I get the same QZRCSRVS job, and so I can do some things like
changing library lists and so on, and they persist between calls.  However,
it seems that OS/400 creates a new activation group for each call.  Thus, if
I use *CALLER in my API, I will not be able to persist data using the return
with *INLR off technique between calls.  Instead, I have to used a named AG.

My assumptions, however, are based only on empirical evidence gathered
during my testing.  I'd like to know from any IBM folks whether I'm right on
this, and where there might be a little more information on how AGs are used
in servers.

Joe