Re: OVRDBF in ILE (was: diff btw Procedures and Routines)

On Thu, 22 Aug 2002, Keith Carpenter wrote:
>
> Granted you and Scott have made a reasonable effort to approximate a single
> application AG with *NEW/*CALLER, but why beat around the bush ?  A named AG
> would be the direct approach.
>

As I already explained, *NEW/*CALLER is more robust because it
automatically cleans itself up.   It's also easier to maintain because
you don't have to worry about making sure you don't write two applications
with the same name.

You replied to this saying it "comes at a cost".  As I demonstrated, the
cost is an extra wait time of about 1/100th of a second.  I can live with
that cost.

>
> As you pointed out yourself, there are those "special case" programs that
> get called from the menu and also as a subprogram ("function key") from the
> another program.  *NEW/*CALLER just doesn't keep that program in the same AG
> as the application.   A named AG would solve this problem.
>

Programs called from other programs should be *CALLER.  They should NEVER
be named groups, because that can create difficult to debug issues.  (You
call a service program, don't realize it's in a different activation
group, and when you reclaim your group, it remains... or the other group
gets reclaimed and ends the service program without the original group
ending...)  There are special cases  where they should be *NEW, such as if
they need to be able to run more  than once within a job, at different
call levels, but that's rare.