RE: FW: IBM's RPG Strategy (was: Long Procedure Names) -- RPG400-L

> From: Scott Klement
> 
> > This says specifically that the JVM crashes ON THE 400 over and over
> > again, with "anything moderately complex".  As it turns out, though,
you
> > only know of one program that actually causes problems on the
iSeries
> > JVM.
> 
> No, that's not true. I know of one program that CONSISTENTLY crashes
the
> JVM on the iSeries. It crashes on other occasions, they're just
sporadic
> making it difficult to nail them down to a particular program.

Okay, semantics.  However, if you have a consistent crash, you ought to
report it.  The Java guys are VERY good about responding to that sort of
thing.  As to sporadic problems, well, there are dumps and you might
want to report them too.


> > How about toning down the rhetoric just a tad, eh?
> 
> On that score, you're absolutely right. When I'm frustrated, I tend to
> type something that's a little extreme.

Not me.  Never.  <grin>


> > For what?  For web-page serving it's really very good.  For database
> > access it's really very bad.  As far as I can see, the big problem
is
> > that you don't have a problem set for which Java is a good fit.
> 
> Or, perhaps there isn't a problem for which Java is a good fit...

No, there is.  I keep trying to make this clear: static solution set
problems, particularly translation, are absolutely perfect fits for OO
(and hence Java).  In situations where you add new features which are
variations of old features, but you don't change the base feature set,
Java kicks butt.  The entire class hierarchy things works wonderfully
when your base classes change rarely.  A good example?  HTML!
Generating HTML is an absolute dream job for a good OO.  That's why I
love Java for browser UI.  And by merging the HTML directly with calls
to procedures that generate HTML, JavaServer Pages are the best possible
technology.  Note that I do *NOT* include Struts or JSF in that
collection.


> Why do you think it's a "stunningly powerful language for user
interface"?
> Again, it seems to me that you're talking about software that's
already
> been written. Because they happened to write JSF or Struts in Java,
that
> makes Java a great language?

I hope I made part of my point clear in the previous comment.  Since
HTML generation is a static problem set, it is perfectly suited for
rigid hierarchical design, and thus for OO programming.  But I'm talking
about the base language: inheritance, polymorphism, that stuff.

Struts and JSF are the exact antithesis of that idea, attempting to
stuff application design into the same rigid hierarchical structure, and
because of that it fails either by providing too little capability or by
generating mountains of superfluous garbage to handle every possibility.
OO and business logic do not mix.


> If the same things had been written in C++ instead, would C++ be a
> "stunningly powerful language for user interface"?

Actually, C++ should be just as powerful.  I like C++, I just don't like
the pointers.  One thing I love about Java is the lack of pointers.


> That's my problem with the whole Java thing.  I can write the same
thing
> in another language, and it requires far less memory and runs far
faster.
> Why would I want to do it in Java?  Only one answer comes to mind:
> portability.

Scott, I have a thin-veneer UI client written for WebSphere that rivals
5250 for response time.  5250 will always win; the datastream is just
that much more compact, and the hardware and OS is optimized for it, but
my UI is as fast as you need it to be.  I wouldn't think of writing an
MRP generation in Java, but an EBCDIC<->HTML translation, absolutely.


> Having said that, though... I don't do much database programming in
Unix.
> Why would I use Unix for database when the iSeries does it so much
better?

Agreed.


> I haven't done any or seen any benchmarks. It's based on my personal
> experiences. Every program I run that's written in Java takes much,
much
> longer than C or RPG. Whether I wrote it or whether a vendor wrote
> it, it's always slower.

A lot of that is JVM startup.  Second it may just be poorly written
Java.  If your Java is not doing a LOT of code reuse, then chances are
you're not going to get the bang for the buck from the JIT compiler.


> But, RPG will perform well without me having to be an expert in it. I
> don't have to optimize it, or jump through hoops to make it perform
well.

I don't have to jump through hoops for the JIT compiler to work in Java,
either.  I just have to write intelligent Java code with a smart
hierarchy of objects.  Then again, I write a little differently than
some "Java experts" who believe in using someone else's code whenever
possible.  Me, I still think I can write most things better than anybody
else.  I only steal from the best <grin>.

(Not to blow smoke, but this would be an appropriate time to admit that
all of my limited sockets knowledge started life on your pages <smile>).


> I guess I haven't found anything lacking in RPG's support for CGI. It
> seems to me that it's well suited to both UI and business rules.

Ah, but CGI has a number of disadvantages, starting with lack of
portability.  Also, unless you are a masochist, you have to pick between
one of the template packages, or roll your own.  None of them will be
particularly amenable to newbies.  Me, I'd prefer to leave the UI to
newbies of college while I write the real guts of the application.
Then, I can pick which of the newbies might actually have what it takes
to make it to real programming <grin>.


> > Many languages fill similar though not identical niches, and
choosing
> > between them is as often a pragmatic matter of expediency or
personal
> > knowledge as it is a purely objective one.  But what I HAVE learned
is
> > that someone who consistently bashes a language just as a language
has
> > probably been using that language the wrong way.
> 
> Okay, well I guess I'm just an idiot, then.

Nope, just a bigot.

<smirk>

Joe