RE: Activation groups for beginners

Carel Teijgeler wrote:

>There are two groups of programme's (PGM): OPM
>and ILE. OPM PGM's are written in RPG III or
>RPG IV. ILE PGM's are only written in RPG IV
>and uses elements of the ILE concept.

Since this is the RPG list, I'll say that ILE RPG programs are only written
in RPG IV. Other ILE languages include Cobol and CL (oddly enough.)


>RPG IV PGM's can be compiled with the following
>options: DFTACTGRP(*YES) (default) This AG is
>created at the start of the job and I assume the
>name of the AG is the jobname.

Do a DSPJOB option 18 and you'll be able to verify the name for yourself.
Here's my V5R1 system:
----Activation Group-----
Number         Name           In Use Indicator
0000000001     *DFTACTGRP        Yes
0000000002     *DFTACTGRP        Yes
0000001676     QLGLOCAL          No
0000001921     QTOSSAPI          No

-snip-
>In all cases: any AG has a name.

This is true; but when we speak of Named Activation Groups, we're speaking
specifically about other groups than *DAG.  The vocabulary does not make it
easy to talk about these concepts.  Non-*DAG is a mouthful.  At least NAG
can be pronounced, however horrible it may sound...

>You can run a PGM in three kinds of AG's:
>
>1) Default Activation Group (DAG). The DAG is
>the environment that runs a mixture of OPM
>PGM's and ILE PGM's.

Yes, but the behaviour of ILE programs compiled DFTACTGRP*YES) is different
than the behaviour of ILE programs compiled *CALLER being called from *DAG.

>2) A Named Activation Group (NAM). In the NAM ILE
>PGM's will run that have been compiled with the
>name of an AG provided or with the special value
>*CALLER.
>
>3) A *NEW Activation Group.

I've been calling a Named Activation Group a NAG.  The only difference
between a NAG and *NEW is who assigns the name.  With a NAG, the programmer
is responsible to see that the chosen name is appropriate (that it won't
clash with other applications).  With *NEW there can never be a name
collision; the system will assign a unique name every time.


>Considerations in defining an AG to an ILE PGM are
>manyfold; think about sharing resources, overrides,
>record locks, error handling, recursive calls
>(call stack), performance.

Exactly right; that's why this thread came about.  So many people have asked
"What's the best way to deal with Activation Groups?"  The very question
implies the belief that there is a single "right way" to think about the
subject.  I do not believe there is a simple answer, so we're "exploring"
different AG strategies in the hope of making some of the decision points
clearer.

Thanks Carel!
It's good to have another voice in the discussion!
  --buck