>>30
What I was getting at (and I had a huge rant before firefox crashed) is that the costs of "reinventing the wheel" aren't as dominant as you might think. We have hundreds of different processors with different designs and performance characteristics, but the fact that many processors share the same instruction set is so that they can sell them to the same customers and they don't have to port their software. It is a convenience, not a key design point.
That's why the general methods of hardware computation have been abstracted, so that we can instead focus more on the actual problem of what computations need to be performed rather than how to physically perform them.
Your conclusion is fine, but your reasoning to get there is flawed.