Now you're speaking my language (see my post analogizing sentence logic with algebra). Algebra would deal with variables(x, y z), the domain of which would be all real numbers. Arithmetic deals with Instantiations of these variables (represented as a,b,c in algebra and 1,2,3.5 in arithmetic I suppose). The domain of propositional variables in logic, rather than being the infinite set of real numbers, is the set of just two possible values (t,f).I agree that we do include propositional variables as a primitive notion, which take on the values T and F. Perhaps we can consider the application of the truth functions without variables as the 'arithmetic' of logic and truth functions with variables as the 'algebra' of logic.
That's what I have been thinking about lately, reading a book by Susan Haack on metalogical issues, when this thread came up. Your 'perhaps we can consider the application of the truth functions as...' comment goes to the heart of logical semantics - i.e. 'interpretation' of the primitive terms. I'm still working this out in my own mind, so I'm not saying I'm correct in my analogies and interpretations and my previous posts may not have been clear.
I think Arising is asking something like - Isn't the interpretation of propositional variables as true or false, but not both (since the domain consist entirely of two and only two choices) somewhat odd or circular? If the semantics allows only two values to variables, then is the principle of non-contradiction (or tautology) semantical (implicit the interpretation) or syntactic (stated as an axiom)?