Let $a, b \in \mathbf{R}$. It is an elementary fact that addition is a commutative binary operation on the reals, that is, $a + b \in \mathbf{R}$ and $a + b = b + a$. With the exception of commuativity, the same goes for subtraction. It is indeed clear that $a -b \ne b - a$ for some $a, b$. However, consider the intuitive notion of subtraction where one simply performs addition with the additive inverse of $b$:$$a + (-b) = (-b) + a$$
If one uses the intuition of "add the negative of the number" as a means of subtraction, you are simply using addition, a commutative operation. Of course, it is still absurd to ask the question "Is subtraction commutative?", seeing as we no longer consider subtraction as a binary operation, but addition on the additive inverse.
While the notion of subtraction as a binary operation is the one that is most widely used in elementary schooling (that is, a third grader will likely be clueless about binary operations, but he won't argue that $3 -4 = 4 - 3$), the second allows for the construction of algebraic structures such as the abelian group $\langle \mathbf{Z}, + \rangle$. I note that in the few introductory courses I have participated in, and in particular that of Abstract Algebra, this second notion of subtraction is used more widely.
Is subtraction (for simplicity, on the reals) the binary operation $-$, or the additive operation of $a + (-b)$. In particular, does subtraction as a binary operation serve a(n important) purpose in the research of algebraic structures?
I may be in the wrong here, but I think of substraction (well, from an abstract algebra point of view...) as :
first you enrich the monoid $(\mathbb{N},+)$ with symbolic elements $-n$ to get a group
now you have a group $(\mathbb{Z},+)$, and $a-b = a + (-b)$ by construction/definition. Of course this carries over to $\mathbb{R}$.
If you want to think about substraction of elements of $\mathbb{N}$ "while only knowing about $\mathbb{N}$", then I think it's reasonnable to think of it as the solution of $$a=b+x$$