Here's a very simplified view of programming:

1. Problems are either well-understood or not.

2. Problems are either small or big.

I think Squeak and its iterative methodology works best for problems that
are not well-understood. I think what Justin has been arguing for is changes
or a redesign to best deal with problems that are well-understood but large.
That is where CASE tools, modeling, and lots of up-front design has the
biggest payoff. When a problem is not well-understood then a great way to
understand the problem better is to jump in there and build some code
resulting in a better understanding. As a result of this better
understanding the code changes and one iterates.  My impression is that most
large software projects in insurance or banking are problems that are
well-understood and hence this iterative exploration isn't appropriate and
can be counter productive if the problem is large and hence the team is
large.

Best,

-ken kahn ( www.toontalk.com )