I wrote this paper many years ago. It's so old I had to convert it from Word Perfect 5.0. I would put it differently now, but I still believe the basic idea should be implemented.
Algorithms now exist that could have a profound effect on computers, especially computers used for scientific computing, making them both faster and more accurate. I believe residue rational computers can be built that are about a hundred times as powerful as today's best machines. Building hardware and writing software to take advantage of these new algorithms is a challenge, but it could be well worth the effort.