Sometimes I like to think about the most fundamental assumptions I can about how things work and wonder "what if this were not true"? Here is an example of that sort of crazy thinking.
In most programming languages the statement "a=b" assigns the value of "b" to the variable named "a" - the two things "a" and "b" play completely different roles (although after the statement has executed they both have the same value). What if assignment instead were reflective, so that "a=b" meant the same thing as "b=a"?
In that case, the information about whether this statement says something about "a" or something about "b" must come from elsewhere. For example, if earlier we had said "a=1" then both "a=b" and "b=a" would mean "b=1". So each variable would have some information associated with it (let's call it the "direction") that says whether it accepts values from other variables or passes its value to other variables when assigned.
If the direction of variables can change over the course of a program's run, it won't generally be possible to determine statically that two output variables are never assigned to each other, so that would have to become a run-time error. Unless we have the ability to change the direction of two connected variables at the same time, we'd have to allow two input variables to be assigned to each other (and give them both some default value).
Perhaps surprisingly, you can end up with a Turing-complete language by going down this route - all you need are two other elements: one is a way of connecting three variables together (obviously no two of them could be outputs at the same time) and a NAND or NOR operator with two inputs and an output. Then you can build any function like a logic circuit.
What's the use of this? Well, it's handy for describing electronic circuits (where assignment means connection). Also, when I write my modular emulator it will have a language like this built in for describing how things are connected up. For that the "two outputs connected together" isn't just a theoretical problem any more - the Commodore PET computer had a Killer poke which connected two outputs together, damaging circuitry. An emulator could have some fun with how it handled this - I'm thinking an "explosion" graphical effect might be a nice easter egg.