Dave Winer points to Joel Spolsky's site saying that Joel posts outrageous remarks. The Mac diss caught my eye, so off I went, finding an outrageous rant about NASA software. Joel's contention seems to be that modern software production methods (designed to maximize the profitability of the software produced) should work for space vehicle control purposes. It won't and here's why:
Software used in outer space must be 100% reliable or be redundant for those parts which may fail. Complexity, features and a cool user interface are not needed. Simplicity and reliability are required because the system needs to be maintainable or circumventable while in use. Just as electrical machinery is more reliable if it has large wires, software is more reliable if the flow of execution is large and visible. Software development techniques based on layers of abstraction that is used to make all the software common on our machines will not work in the cold, lonely vacuum of space.
Chris Heilman's own old fashion software, Pocket Forth, works by directly altering the electronic configuration of metalloid atoms suspended in a silicon lattice.
