RE: CFINT: I understand it now...

["Joe Pluta" <joepluta@xxxxxxxxxxxxxxxxx>]

> -----Original Message-----
> From: Reeve Fritchman
>
> It’s simple: compared to 5250 applications, server applications
> are grossly inefficient.

In what way?  In fact, a well written server program is far MORE efficient
than a traditional monolithic green screen application.  By separating
business logic from presentation, your code (on both sides!) becomes much
cleaner.  Even before the CFINT issue came to the fore, the folks that ran
the major development shops were looking for ways to extract common code
from applications and put it into called modules - this is in essence the
concept of server programming.


> crunching-datastreams-in-RPG-so-I-can-write-a-simple-error-message
> -back-to-the-browser tax.

I must admit I don't understand this.  A server simply returns an error
code.  It's up to the UI to resolve this to something the user can
understand.


> What we’re spending for interactive features now will be redirected to the
> purchase of increased CPW systems.  And of course Big Blue can juggle the
> numbers so it will appear you’re paying less for the monsters
> system you’ll have to install to support the Brave New World of server
applications.

Again, I'm a bit confused by this opinion.  Where exactly does this
increased CPW requirement come from?  The only added code in a well written
client/server architecture is the overhead of the messaging.  With a decent
API, this overhead is trivial, both from a machine cycles standpoint and
from a programming standpoint.

If, on the other hand, you're talking about the overhead of the browser UI
as opposed to the 5250 UI, then yes, that's a finite, tangible amount.
There is indeed work involved in formatting HTML that we don't have to do
for green screens.  This work is done for us in the 5250 workstation support
routines (and in the microcode of the 5250 terminal, or in the emulation
package).  Since the HTML interface is far younger and less mature than the
5250 interface, there's more work involved.  If your concern is the cycles
involved in HTML formatting (which are actually less than you might imagine,
but that's a different issue), you can offload that portion quite easily to
a web serving box.

If, though, your argument is that you want to run an HTML interface with all
the processing on the iSeries and you want it to be as fast and as easy and
as efficient as the 5250 interface, without any additional work on your
part, then you're asking for quite a bit.

Interestingly enough, I think it's doable.  We're pretty close to doing just
that today with JavaServer Pages.


> The put override is a thing of beauty…

So is the JSP.

Joe Pluta
www.plutabrothers.com