Hi Claudio,
I am no talking about a new language, rather a framework that takes care of the “assumed” running environment
I'm not able to visualize what you mean. What is it the developers will actually edit? What editors will they use for this and for which "parts" of the application? Do they have a free choice of editors?
The ProIV security stuff is not something I would try to put in a new RAD tool – I don't think it works particularly well and a better approach would be proper integration with security infrastructure(s) external to the tool.
I presume your reference to "Automatic design of screens and reports" refers to capabilities of Superlayer? Some visitors to this forum (including me) do not usually think of Superlayer as included by the general term "ProIV". For clarity it may be better to mention Superlayer explicitly when relevant.
Note it might be argued that, overall, Superlayer has been a failed experiment in RAD and looking at why that was could be very instructive for your proposed project!
Some (including probably me) would argue that screen modes (UI modality) are a BAD thing and a modern tool should strive to avoid this feature of ProIV.
Java could be arguably too verbose for "RAD logic" (declarations mandatory for all variables as an example). I also note again that you need to consider what subset would be legitimate/safe and how you'd enforce that.
For me, the "quality" issue as far as the tool is concerned is prevention of defects. The relevant question is what language/tool features help prevent application defects and how do they do so?
Starting by supporting only open software would be fine I guess - beware though that it's very hard to be "smart enough" to support the proprietary targets well later unless one knows enough about them right from the start..
I'd agree there's no point bothering with ISAM filesystems – just Object-Relational SQL databases (targeting an ORDBMS is probably important to properly exploiting Oracle etc.. later). I'd agree PostgreSQL is functionally rich but note that all the commentary suggests its performance is poor. I notice MySQL has teamed up with SAP to offer the SAP database rebadged as "MaxDB". I don't know if this has any O-O capabilities but performance is supposed to be pretty good. The cynic in me says the folks at MySQL finally decided that if you don't design transactions and performance in from the start then you can't retrofit them

"Integration" at the level of common database access is of course available with almost any programming tool. It's OK to aim for that but it would leave little reason for the ProIV community to prefer your theoretical new tool to any other RAD tool they could try.
My problem with the "bootstrap model" is that I can't use standard tools on my source code! For example none of the following can be exploited:
vi/ctags, sccs/rcs/cvs, grep/awk/sed, diff, make, cpp/m4, code-formattersNo modern app is developed using only one tool and one wants to be able to build the whole app in a "common tool framework" (the openness of Eclipse is advantageous in precisely this way). Scriptability of development tools is an important issue. Inadequate execution profiling, coverage testing and debugging support is also a problem in ProIV (although that is not necessarily the fault of the "bootstrap" approach, I think it hasn't helped).
There's plenty of open-source .NET stuff happening – just throw "open-source" ".NET" at a search engine.
And just to note the obvious finally, file processing PER FIELD is relevant only to the screen function 'model', not reports or updates