RE: zero interactive CPW questions

> From: James W. Kilgore
>
> So it appears that even telnet would not work.  I looked at the
> Java based 5250
> emulator from Mochasoft that runs as an applet in a browser, but
> it appears to
> still be talking to the iSeries telnet server.  I'm guessing that
> IBM's telnet
> server, although a batch job itself, only talks to some
> interactive component,
> maybe the QT3REQIO component of QWSGET that I see in the program
> call stack.
>
> So does this mean that I'm pretty much stuck with using a browser
> or some thick
> client c/s approach?

James, I do all my development on a model 270 with zero interactive CPW.  As
long as I am using one session at a time, there's no problem.  There's also
no problem switching between multiple sessions, provided none are doing
anything CPU intensive - I regularly have three or for sessions up, and hop
between them.  However, if I, for example, start a mass compile
interactively on one screen, and then try editing on another, the machine
will lock up.

Of course, for a single user, the answer is simply to submit long-running
tasks to batch, and most of the time that's sufficient.  That's not to say
it's all sweetness and light.  There are a few things I've found that will
cause the machine to bog down, even on a single session.  Holding the page
down key to scroll through a long spool file is one way to bog the machine
down.  In fact, any repetitive large-scale screen I/O has a tendency to
occasionally slow the machine down to a crawl.  At that point, the only
solution is to wait a few seconds for the machine to "cool down", and then
go on with your business.

Far more annoying is the occasional habit the debugger has of seizing up.
This primarily happens if you try to single step through a piece of code
with a whole bunch of F5's queued up in your type-ahead (I think F5 is
single step - I forget).  It's especially bad when debugging a batch job,
for some reason.  It seems that the machine gets very confused and thinks
there are two interactive jobs and WHAM up goes the CFINT.

You're right in assuming it's the QWSGET that's doing it.  From what I
understand, any calls to the workstation functions causes the interactive
bit to get set in the job (this is the bit that the infamous FAST400
cleared).  Any technique that uses telnet or screen scraping will be seen as
an interactive session.  That's the side benefit of my approach; the jobs
run in batch, and don't use the workstation functions, so they in effect
stay "below the radar" of the CFINT governor.

So, your two basic approaches are to either rewrite your business logic
completely into true servers, or else update your I/O routines to use an API
along the lines of my server/client design.

Joe Pluta
www.plutabrothers.com