No, of course I didn't take it like that...
In fact, I 've been trying to do this (against my better judgement) and I 'm pro-"MyDMS code base" not because it's efficient for the project; it's definitely not. The reasons are practical: it's much easier to concentrate on addressing one issue at a time, possibly spend time rewriting subsystems, than to make "the mother of all specs" and start then implementing. The second approach would be most productive, but it requires experienced people with lots of time in their hands.
So, I 'm taking the other path because there just isn't time for the uber-design. You 've said so yourself. One step at a time, on the other hand, will be one step closer to the goal each time. And of course, don't forget the morale issue: people need to see progress, no matter how small! (and that includes us developers!)