What Does IBM Need to Do with RPG?

[David Abramowitz]

** This thread discusses the article: What Does IBM Need to Do with RPG? **
Robert Cozzi hits the nail on the head with his latest column. Many of the "newest" features are nothing more than a reinvention of the wheel. RPG now provides at least a dozen different ways of accomplishing the same thing. Other languages can be written in different styles, but RPG actually has redundant verbage for the same functionality. I believe that much of the impetus for the current flavor of RPG is due to turnover in Rochester, and Toronto. Many of the compiler developers who were familiar with the historic customer base are no longer there. Those now in charge are JAVA, C++, and Visual Basic enthusiasts who have spent a great deal of time and effort making RPG look a heck of a lot more like JAVA, C++, and Visual Basic. When I converted three shops to RPG IV, from RPG III, I had to provide some justification. I was able to do this by pointing out the new date data type features, available for use with RPG IV. No other feature generated as much interest among the powers that be. Having a programming language do something new, is far different from having a programming language do the same thing in a different way. Robert Cozzi is correct when he points out that code written using new features at one shop may be unusable, or undecipherable at another. Give me something substantial that I can use in day to day business applications, and I will use it. Dave

[R.Cozzi]

** This thread discusses the article: What Does IBM Need to Do with RPG? **
Thank you all for your valuable comments! Keep the coming. Remember, my point is that RPG IV is being enhanced in a fragmented manner, not that the iSeries is evolving to quickly. Yes I want continued enhancements to RPG IV (or "RPG V") but I don't think they need to be rolled out in each release of OS/400. I would rather IBM take the time to ponder the enhancements and make sure they are consistent and valuable, rather than just release them and be done with it. My hope is for a better language, not a language with the most features or the most redundant features as some of your have noted.

[R.Daugherty]

** This thread discusses the article: What Does IBM Need to Do with RPG? **
1. Does an AS/400 compiled Java program without database I/O execute as fast as an RPG ILE program? My expectations are that it would. 2. Does an AS/400 compiled Java program with database I/O through Toolbox record level I/O calls execute as fast as RPG ILE program? My experience is that two years ago it benchmarked significantly slower, but my expectations are that it could run 80% as fast as ILE. 3. Is there any reason the syntax for freeform wouldn't be Java? 4. How could RPG/400, RPG IV, Code/400, and /Free not be considered a fragmentation of an industry so serious as to threaten the continued existence of software development firms developing commercial packages for the AS/400? rd

[R.Daugherty]

** This thread discusses the article: What Does IBM Need to Do with RPG? **
Chuck wrote: "...yet no one considers Visual Basic a fragmented language." This is extremely wrong, as I read in the trade press weekly. VB .NET is not being adopted in any significant numbers, it is widely acknowledged that only 40% of VB code will port with the rest needing to be rewritten, and there is even more unhappiness with .NET than with the VB 6 changes that introduced OO to VB. This is not a statement for or against the change, this is to simply refute the very wrong notion that VB is also not recognized as becoming fragmented. Of course, the M$ plan is for all to jump to the latest version, running on the latest OS. In other words, it is fragmented and will become more so. rd

[R.Daugherty]

** This thread discusses the article: What Does IBM Need to Do with RPG? **
Chris wrote: "Even if the Java program is compiled with OPTIMIZE(40) and running natively on the AS/400, no. The time to initialize the JVM takes longer than the time to initialize an RPG program." Granted, I have read of the additional overhead of starting up the ILE environment over an RPG/400 program, and some might recall the nasty overhead of the C environment in the early 90's, but from an OS perspective I find this statement difficult to fathom. Wasn't the JVM built native into OS/400, wouldn't the JVM code be in shared read only memory, and wouldn't the initialization described be limited to the data segments (to use PC terminology) required for each new Java program? Given that the strategic direction of the AS/400 is as a Java machine, I find the concept of a JVM as a slow, monolithic monster which must be laboriously started up for each program to be incomprehensible. rd

[R.Daugherty]

Chuck wrote: "I don't consider them fragmented. The fact that IBM keeps older versions around is a plus." Certainly backwards compatibility is critical and has been the hallmark of the Sys/3x line. However, unless development is overwhelmingly chosen in one of the RPG flavors, either software is fragmented or is not being developed at all. From my limited perspective on the industry, it appears to not be being developed at all. rd

[R.Daugherty]

Chuck wrote: "Today's iSeries is like the early '90s Mustang. It's got a great following but will IBM embrace it? Or has Rochester gone too far causing Armonk to lower the hammer even harder?" This is very interesting comments and analogy, Chuck. rd

[Guest.Visitor]

Ralph, Well, if the Java programs (servlets, JSP), are running on a server (in the context of HTTP and Websphere or Tomcat for example), then yes, it's resident and reentrant and shared by all requests. Under this scenario, IBM claims it's the fastest Java server in the world. But your 1st question was comparing the performance of an RPG program to a similar Java program without database I/O (perhaps a program that just counts?). AFAIK, calling a Java program from a green screen command line (or in QSH) will start the JVM every time and that's gonna cost you in performance. If I'm incorrect, I'm sure someone here will correct me (Don?). If the AS/400 could cache the JVM, why did IBM move just part of the JVM to the SLIC to boost performance? Regards, Chris

[rdean400@yahoo.com]

RPG is not evolving too fast. If anything, it's not evolving fast enough. Most of the college kids coming out nowadays are being trained from the start on object-oriented languages. It doesn't seem far-fetched to assert that the mental leap required to grasp procedural languages when your background is OO is nearly the same as the mental leap going the other way. It's a significant shift...it's hard enough to get RPG programmers now; imagine what it will be like when the new programmers don't have any experience with procedural programming. I'm not saying that OO should be on the table for RPG, but it would be good if syntactically, it were easier for these programmers to grasp the language. The recent changes, including free-form, and "Qualified" data structures, are a good step in that direction.

[D.Handy]

PTLMIS said: "I agree with Bob completely on IBM's rollout strategy. They are simply adding too many enhancements (changes) way too fast." Chuck said: "IMO, that's not even remotely possible ... IBM is doing what they should have started doing in 1990: Aggressive improvements to the iSeries." I agree with Chuck. And this includes aggressive improvements to the dead (?) RPG language. I realize that Bob's gripe is not that the language is improving, but that in his opinion the changes did not go through a long enough review process and produced inconsistent syntax, or inconsistent feature sets. I don't view it that way. Let's look at the specific examples cited. The CHAIN syntax was claimed to have four different syntax variations, and the implication is that they were not well thought out. Yet an alternative was not proposed. Granted, the fixed format vs free format have a different syntax: Fixed: operand op-code operand Free: op-code operand operand But that's no great surprise or learning curve; they all act that way. Calling the free format CHAIN three different syntaxes is missing the forrest for the trees. All of them are still "op-code operand operand", with the first operand being the search value. So what's the big deal between the operand being a single field name, a parenthetical list of fields, or a %bif() expression? They all specify the search key, and all follow the "op-code operand operand" syntax. Calling them different syntax and hard to learn is like saying it is hard to learn a command where you can specify a single value as a parameter, a parenthetical list, or a keyword with a value. In each case you are specifing one parameter, just like each of the "three" free-form chain syntaxes. I see absolutely nothing confusing about these "variations". In fact, I don't even consider them to be any different. Just like I don't consider a command syntax to be different because you coded a list instead of a single value. What's the real gripe here? If you waited 3 years to have more time to think about it, how else would you implement the ability to provide a list of fields to chain in lieu of a key list? The way it is implemented *does* seem consistent to me. I didn't see an alternative proposed. (And getting rid of free format is not an alternative, Bob ). Regarding the alleged inconsistency between UPDATE and WRITE, what is the advantage to waiting a few years and doing both at the same time? In my experience, it is comparatively rare to want to WRITE a new record and populate just a few fields, while it is very common to want to UPDATE only a few selected fields. I'd vote some of my $100 towards the UPDATE capability; I could care less about the WRITE capability, but it wouldn't get any vote dollars from me. I have no problem with getting the UPDATE capability sooner, as opposed to having to wait for WRITE to get the same treatment. I, for one, am estatic to see RPG getting rapid enhancements. RPG was a great language for business applications in spite of itself; now it is maturing into an even better tool. I also don't have a problem with trade rags publishing code using new language features -- even if I can't compile and run it as is. Articles *should* be educational, and show how to do things you don't already know how to do. I don't want all utitlies printed in RPG III... Doug