The common proof goes like this:
Suppose $X$ is countable, then it must be Lindelöf. A regular Lindelöf space is normal (akin to the proof that a regular and second-countable space is normal). Using Urysohn's Lemma, it's easy to establish a contradiction.
Out of curiosity, does this result have a direct proof (without assuming $X$ is countable or making similar assumptions at the beginning)? It seems it's crucial here, otherwise, we couldn't deduce $X$ is normal, hence the proof couldn't follow. I've done some research on the internet:
Show that a connected regular space having more than one point is uncountable
Is every connected regular space having more than one point uncountable?
All are of the same proof. Can we avoid proof by contradiction? Thanks in advance.
The conclusion of the theorem is that $X$ is uncountable. This is defined, usually, as "$X$ is not countable". And logically speaking (in first order logic) the way to prove a statement of the form $\lnot \phi$ is to assume $\phi$ and derive a contradiction from it (even Brouwer, in his intuitionistic way of doing maths, did it that way; he did not accept, though, that proving $\lnot (\lnot \phi)$ proved $\phi$, as in classical logic). So the going via contradiction seems quite unavoidable.
Maybe you could define "uncountable" in a positive way (e.g. there is an injection from $\aleph_1$ into $X$), but then the proof of equivalence with the "negative" notion would (I think, correct me if I'm wrong) require a form of the Axiom of Choice (another source of possible controversy). And I don't see a proof route there, personnally, so that wouldn't solve it either.
So I say, embrace it, or state the whole result "positively", in a way that does not require a proof by contradiction:
Proof: $X$ is Lindelöf, so $X$ is $T_4$. Pick $x \neq y$ in $X$ and let $f: X \to [0,1]$ be continuous with $f(x)=0, f(y)=1$ (Urysohn lemma applied to $\{x\}$ and $\{y\}$). Then $f[X]$ is countable and $[0,1]$ is not, so pick $t \in [0,1]\setminus f[X]$ and then $f^{-1}[[0,t)]$ and $f^{-1}[(t,1]]$ form a decomposition of $X$. So $X$ is disconnected.