If $\mathbb{R}[X]$ denotes the field of real valued rational functions, then according to Bochnak, Coste and Roy's Real Algebraic Geometry, the unique ordering for which $X > 0$ and $X < a$ for all $a > 0$ is given by, for $a_k \neq 0$ and $n \geq k$, $$P(X) = a_n X^n + \dots + a_k X^k > 0 \text{ if and only if } a_k > 0$$
and we say $P(X)/Q(X) > 0$ if and only if $P(X)Q(X) > 0$.
This defines a total order of the field of real valued rational functions.
What I am trying to prove is that this is necessarily the only one for which we have $X > 0$ and $X < a$ for all positive numbers $a$.
It is enough to show, as $X^k > 0$, that we must have $a_n X^n + \dots + a_1 X + a_0 > 0$ if $a_0 > 0$. I was able to show this for the $n=1$ case, but for $n > 1$, I am unable to do so.
Let $\alpha>0$ and $n\ge 1$. We have
therefore $\alpha>\beta X^n$ $\forall \beta$.
In particular, if $a_0>0$, then $\dfrac{a_0}{n}>-a_k X^k~\forall k\ge 1$, so $a_0>-a_1X-a_2X^2-\dots-a_nX^n$ and $$a_nX^n+\dots+a_2X^2+a_1X+a_0>0$$ as we wanted.
The reverse implication is not true, as pointed out by a comment.