Have you ever implemented your own FSM?

Please give details!

My very own FSM was born ... let's say... yesterday(on paper) and its design included 2-bit ram for holding the "current address"/state, combinational logic for transitioning between different states depending on the previous state and the input, "Bios"/"rom" for the start-up state(address = 00) and one output variable. I was implementing a digital "lock" with a three-binary-digit code. I was wondering if the RAM could serve a more special purpose in these simple FSMs kind of like real computers do, except without the "temporarily hold CPU instructions", "temporary data" purposes and that stuff.

I have it written down on a flow diagram but I haven't figured out the actual electronic implementation of "wait until input comes in" before transitioning and other electronic implementation details... hmmmm...

Why not implement it fully in sw, using only 1 chip, one of the smaller (in SMD, really small) 6 to 8 pin PICs or AVRs?
Having human mechanical input usually means debouncing need somewhere which adds a bit more complexity. Is there some specific requirement that forces you to do this in discreet electronics? For synchronous state machines (require a clock) you have the "Mealy machine" and the "Moore machine" (very easy to find on the net).