A number of early small computers were non-binary. One that comes to mind is the IBM 1620 which was a decimal variable word lenght machine. The 1620 was in production about 40 years ago and was mainly marketed as a business machine. One of the interesting features of this machine was that it did all its arthmetic by table lookup.

The tables were stored in memory, so you could change how the operations worked!

A number of early small computers were non-binary. One that comes to mind is the IBM 1620 which was a decimal variable word lenght machine. The 1620 was in production about 40 years ago and was mainly marketed as a business machine. One of the interesting features of this machine was that it did all its arthmetic by table lookup. The tables were stored in memory, so you could change how the operations worked! A number of 1620s were used by universities into the late 1960s. Since they were variable word length, they were very nice for doing precise computations. Since early computers were based on analogue electronics it was much easier to do non-binary than it is now. Many early memory devices (except core) were really analogue devices with thresholds used to distinguish 0 and 1. You just needed to add a few more thresholds to get a larger range. -- Dr. Mark Green mark(a)cs.ualberta.ca Professor (780) 492-4584 Director, Research Institute for Multimedia Systems (RIMS) Department of Computing Science (780) 492-1071 (FAX) University of Alberta, Edmonton, Alberta, T6G 2H1, Canada

When all you have are ones (and nothing else to represent the absence of a one), what do you use to delimit the start and end of variable-length words?

The IBM 360 series was a "hexadecimal" machine. The result was that you got one or more fewer decimal digits of precision for floating point numbers as opposed to a binary machine. The CDC series of 3xxx/6xxx was a favorite of scientifically oriented users due to this fact, and many computer centers were CDC instead of IBM at the time. (There was a difference in word length, too, but many users didn't realize that you could get better precision on a PDP-11 than on a 360 in single precision.) The ultimate, of course, is the UNARY computer. If you look in the "Feynman Lectures on Computation", Richard Feynman, 1996, you'll see that he has you start designing a unary computer. Maybe not so easy when all you have to work with is "1"! (It has to have a variable word length, of course). Dave Mark Green wrote:

I think his idea was more along the lines of data transmission: instead of having digital bits of just 1 and 0, you could have "trigital" bits of 0, 1 or 2. So, for example the binary number 1001101 (7 bits) could be sent, in trigital, as 2212 (4 bits), theoretically saving bandwidth. (Whether it actually *would* or not, I'll leave to the theorists)