Z:gnu-www-ja-stallman-kth--d8ac60-About three and a half year ag/en

About three and a half year ago it was clear to me that I should start developing a free software system. I could see two possible kinds of systems to develop: One: A LISP-machine-like system, essentially a system just like the MIT LISP machine system that had just been developed, except free, and running on general purpose hardware, not on special LISP machines. And the other possibility was a more conventional operating system, and it was clear to me that if I made a conventional operating system, I should make it compatible with Unix, because that would make it easy for people all around to switch to it. After a little while, I concluded I should do the latter and the reason was, that I saw that you can't have something really like the LISP machine system on general purpose hardware. The LISP machine system uses special hardware plus special writable microcode to gain both good execution speed and robust detection of errors at runtime, specially data-type errors. In order to make a LISP system run fast enough on ordinary hardware, you have to start making assumptions. Assuming that a certain argument is the right type, and then if it isn't the system just crashes.