Oh boy, and no offense but I really think your aspirations to build a CPU are a bit more than you can chew. I base this on the questions you are asking.
3 state means the device can drive an output high, low, or go high impedance which means it becomes invisible to the rest of the circuit; in other words, it no longer drives a circuit. This is important on bidirectional lines.
Flipflops can store data because you can put a logic 1 or 0 on one data line, clock it with a strobe and the bit will maintain its state until another strobe, commonly called a write strobe. Later on you can read the stored bit on the flipflop. Reading the bit on the flipflop can be done in several ways. One is to have it connected to a 3 state device which can be turned visible again by a direction strobe and then a read strobe.
With the advent of Large scale integration, most of this logic goo has been incorperated into a single chip, and the designer need only worry about read/write strobe and maybe a few other signals. Without having a chalkboard and you in front of me, this is hard to explain.
How many years, and how many thousands of dollars, have you got available?.
A simple PIC, that costs almost nothing, is going to far out perform anything you're going to make in any reasonable time, and for any reasonable amount of money.
Yes, but you're talking about very very fast computers, that have all kinds of specialized circuitry. It wouldn't be anywhere near that hard to make a computer if it only had, say, 2 bytes of memory.
Take a look at this wikipedia 'home-made' computer made from 7400s:
An Implementation of a 4-bit two register computer, including six cpu assembly instructions: READ (read input), INCB (increment register B), MOVAB (move contents of register A to B), MOVBA (move contents of register B to A), RETI (return from interrupt), JMP (jump).