Peter,
>Well, I also bear responsibility for a legacy application. It is completely impossible to restructure that stuff. Both your and my solution are impossible to implement in that legacy app. Here's an anecdote: Just this morning betatesters were testing the new version of that application. Guess what, an error occurred exactly because a prg had sneakilly got included in an app where it should not have got included. The routine also resides in the exe and uses an included table. That table is not known to the sub-app, so it could not be found. Bang..

not a completly foreign scenario you describe <g>. Don't give up on restructuring too early, and get to know the app [if it is a brand new child for you]. Sometimes even "whacky" seeming concepts have their own rythm and at least some good points. Most of the time the real problem is that later maintainance was done by people without realizing the intentions of the "framework" or ideas put into the app at first.

First order of business: clean out maintainance errors as seen from point of view of the original framework concepts, remove redundancy, build smaller components and enhance encapsulation if possible: basic textbook stuff. But don't follow "your" style, do it in the style of the original developer, if he wasn't a total idiot. If that is done, you will have a much easier part of reachitecting according to your "best usage patterns". At least it works for me<g>.

Main points of contention: Make everybody realize it is a long term plan running alongside usual maintainance / upgrade. Results for money spent are very dificult to "see" as lowered cost for new development building on the upgraded foundation is hard to sell. Biggest risk: diferent levels of competence if more than one developer is working on the project: "dumb" mistakes stemming from missing understanding can be bad for *your* image, even if the mistake was clearly done somebody by else - or to put the burden more on your shoulders: downgrade your concept[s] to the lowest level of competence found in your team. Scary, but the only way not to run into many errors.

my 0.02 EUR

thomas