RE: Activation groups - named vs. *NEW

> From: David Gibbs
>
> Just for the purposes of discussion ... wouldn't it be a reasonable
> approach to use *NEW for the main entry point into a system (main menu,
> command processing program, published API, etc), and then *CALLER for
> everything else (I'm not exactly sure what the best thing to do with VCP's
> and Prompt override programs)?

This makes sense, David.  Personally, I also like *NEW for top level
programs, with *CALLER for called programs, because it ensures that the
subprograms get initialized on each call to the top level program.  When I
have control over the environment, that's the way I do things.

It's a little harder to figure out when you should create a *NEW AG for
server programs.  Do you do it once when the server starts, or once for each
request?  Empirical evidence suggests that IBM creates a new activation
group for each call to a host server, and so sharing data between server
invocatiosn requires either creating true persistent objects (such as user
spaces) or using a named activation group to store persistent data.

I'd be interested to know from anybody at IBM whether this is indeed the
case: do server programs such as the host servers create a *NEW AG for each
request?

Joe