>> The only advice I've got is to do things in increments as small as >> possible. Don't do "big bang" integration. Make sure there is a >> runnable testable program after the first week of development. Maybe >> it doesn't implement any significant features, but you must have >> something runnable and testable at all times. Otherwise, you can get >> too far down the wrong road before you finally figure out that either >> a) what you specified isn't going to work, or b) they didn't >> understand the specification at all. > > ACK. And another important tip: split your big problem into smaller > and smaller generic sub-problems. Then you'll have great chance to > reuse an existing package or let some contractor develop/adapt one > without telling him about your actual project. > > Movie tip: "The Cube" ;-)
That's a good movie and a very appropriate recommendation. I think a lot of people are saying that when it comes time to execute a plan like that, it gets fouled up because the spec is hard to write and it's hard for coders to test what they've written. - Grant
