RE: Trivia: Processor MHz

["Joe Pluta" <joepluta@xxxxxxxxxxxxxxxxx>]

> From: srichter
>
> >From: "Joe Pluta" <joepluta@PlutaBrothers.com>
> >>
> >> Better than:
> >>
> >> InTranL1.Chain( OhPo#, OhComp )
> >> If InTranL1.RcdFound( ) == TRUE
> >> ** Display IntranL1.Po#, InTranL1.RcvQty, ...
> >> End
> >
> >Um... because one exists and the other does not?  What language is that
> >second paragraph supposed to represent?  If it's a Java-like
> language, then
>
> It is pure C++.  Very clear and useable. The InTranL1 class, the
> Chain( ) and RcdFound( ) mbr functions could be easily generated
> from the external database file defn of INTRANL1. No pgmr coding
> needed. All that is needed is the CPW that other platforms have.

Excuse me?  C++?  No, if it were C++, it would be (naming and capitalization
conventions aside) written as follows:

        InTranL1.Chain( OhPo#, OhComp );
        if (InTranL1.RcdFound( ) == TRUE)
        {
        ** Display IntranL1.Po#, InTranL1.RcvQty, ...
        }

Although I would probably write it as:

        if (InTranL1.Chain( OhPo#, OhComp ))
        {
        ** Display IntranL1.Po#, InTranL1.RcvQty, ...
        }


I guess it was the lack of proper syntax that threw me.  In any event, you
still have to write the Chain method.  You say it could be easily generated,
but I don't know what "easily generated" means - who is going to write the
generator?  What does the generated code look like?  Is it ODBC?  How are
record locks handled?  Commitment control?  Do you have one class for each
record format in your system?  Who maintains them?  How does a change to the
database get proliferated through the application?  Most of these issues are
handled automatically by RPG, and are conveniently left out of your
simplified "C++" example.


> >> Any estimate on how much CPW is needed for each concurrent user ?
> >
> >Yeah - zero interactive CPW.  But I suspect it's more than the pure 5250
> >approach.
>
> Excuse me because I am trying to learn something here.  Where did
> you see "interactive" in my question ?   On a 1000 CPW 270, how
> many concurrent users can be supported ?

I meant to point out that my architecture requires no interactive CPW, which
is the largest part of the price of a machine.  That was a bit of a flippant
answer, I admit.  So I then went further to say that my performance is not
as lean as pure 5250.  Nathan, Brad Stone and I went over this a while back.
Performance measurements are always difficult to quantify, especially since
this is a development machine.  On my 370CPW model 270 (with 0CPW
interactive), I was able to handle 200 concurrent requests with subsecond
response time (about 700 milliseconds average, if memory serves).  I have
significantly enhanced the API, so it might be a good time to run some more
stress tests - thanks for bringing that point up.  In any event, it's far
more concurrent sessions than I could do using the 5250 interface, since a
0CPW interactive feature will seize up as soon as the second session starts
up.


> >> The as400 could meet these requirements much better if the CPU
> >> was as fast as it could be ( can even keep the price the same ).
> >> IBM would make more money, colleagues of mine would not be
> >> leaving the platform because of the lack of work and we could
> >> prove to the rest of the computer world that our platform is the
> >> best on the market.
> >
> >Thank you for your opinion.  Got nothing to do with your
> original statement:
> >
> >"This is not trivial.  The slow cpu of the iSeries prevents modern
> >programming languages from being used on our system."
>
> I try to be brief and count on the good intensions of the
> responder.  The original stmt is logically extended as: "The slow
> cpu also prevents our system from being a competitive file server
> and odbc/ado server"

There's no logic to this extension at all.  The choice of programming
language has nothing to do with file serving.  Or with ODBC - you can use
ODBC from languages as diverse as Java and Visual Basic.  In any event, the
AS/400 is not a file server, and was never meant to be, and ODBC is a bad
architecture.  I pointed those two things out the last time.  This
discussion was purely about the ability to mix OOP and procedural code
effectively.  I use Java and RPG together very effectively to build and
deploy web-based business applications, and have no problems with it.  If
you want a file server, buy a Linux box.

Joe Pluta
www.plutabrothers.com