- the majority of these fancy seminar speakers or writers haven't written a single line of production code for at least 5 years, let alone been involved in some real software project, so who cares about their opinion anyway (except that they can be nice to listen to if you're lucky, but I prefer Statler and Waldorf instead).
- one of these articles made the wise remark that business constraints like security, robustness, resilience, ease of maintenance, let alone transaction management or logging are never part of that fancy UML model which they want us to believe will be the only 'coding' effort required by these tools. So where will that 'supporting infrastructure' code be materialized then? Generated by your code generation tool? Provided automagically by your J2EE container? You bet. This industry survived already one huge 4GL wave, and we replaced all these tightly-integrated, monolithical, closed and unscalable installations with light-weight combos of n-tier Java applications. Let's not fall into the same trap again: there is no way to easily factor out the deployment context of your application, pretending you don't care. Your manager will still kill you if you're not able to keep that deployed application up and running. And like it or not: you will be required to think about the deployment context upfront to offer him that guarantee.
"Your manager will still kill you if you're not able to keep that deployed application up and running."
True. True. Which is one reason why the industry seems to be divided into "green fields developers" and "maintainers".
Plus that the "green field developers" gets paid more than the "maintainers", even though maintenence is harder and more boring.
"This industry survived already one huge 4GL wave, and we replaced all these tightly-integrated, monolithical, closed and unscalable installations with light-weight combos of n-tier Java applications."
Is there any specific reason why our industry couldn't evolve towards a 5GL environment, where our tools and platforms would provide us with loosely coupled, component based, open and scalable code? What CERTAINLY did not work was to throw away all the lessons we learned in 4GL and return to a (slightly altered, because Java/internet enabled) 3GL environment...
As a side note: don't think that 4GL tools aren't used anymore. You'd be surprised how much production code is generated every single day ...