RE: More ILE 101

["Joe Pluta" <joepluta@xxxxxxxxxxxxxxxxx>]

> From: Joel Cochran
> 
> Instead of having fifty copys of
> the service programs and hence its global variables, then there is
just
> one copy that gets accessed from fifty different places.  The
appearance
> is that the global variable is shared, but in reality it is the
> activation group that is shared.

Hm.  This is counter to what I always understood.  In my understanding,
there were two different pieces to any program: the code and the data.
While each job got its own data portion, the code portion was shared,
the caveat being you had to be in the same memory pool.

So, if 20 jobs in QINTER ran the same order entry program, all 20
programs would share the same copy of the code portion of the program.
They would just have their own copy of the data portion in their PAG.

It was thus fairly easy to implement re-entrancy by simply allowing
multiple copies of the data portion.  Unfortunately, RPG had some
limitation on re-entrancy that I never quite understood, but CL programs
could call themselves and each call level would have its own data
portion in the PAG.

Activation groups can be use to get around the re-entrancy issue with
RPG.   With *NEW, each call to the RPG program sets up its own data
group, and thus there are no recursion issues.  But there aren't
multiple copies of the program code, just of the data portion.

Now, there ARE problems if you use bound modules.  If four programs in
your program stack use the same bound module, then the code for that
module is repeated four times.  That's a different issue, but it's
probably one of the most important reasons for moving from bound modules
to service programs. 

Joe