[Jtsummers]

Take an incremental approach.

Establish the core needs, what must be shipped as a basic system, and ship it. Then extend over a series of iterations where you incorporate more needs and wants. You won't totally avoid the deadline, but you will mitigate one of the deadline problems: All-or-nothing.

If you establish a deadline 1-3 years from now it is expected the entire system will be finished on that date. Well what if it's not? Establish quarterly releases (or more often, especially if you can work with the team deploying the system like if you're an internal dev shop) and aim for a first release in a few months that hits the essentials. Everything else is an extension, you may still take 1-3 years (or longer) but you'll be releasing and getting value from the system right away.

I've even done this with safety critical real-time systems in my work. We built what was necessary for flight test, but not desirable for operations (missed a few needs, a lot of wants). By the time the plane was rolling off the factory floor for release to airlines, we had all the needs covered and most of the wants. The rest of the wants were covered in the next year or so as well as some newly discovered needs. But if we'd waited to have everything done, the plane wouldn't have been delivered to customers for several more years (because flight testing would have been delayed by about 3 years).