Given a model with a binary variable $b_s$ that describes whether taking an item $s$ from a set $S$ or not. Consider that some other constraint in the model depends upon whether all items of the set are taken (so to say the minimum of all $b_s$).
That other constraint can also be described with the help of a further binary variable, but given that $b_s$ and $c$ are "tightly connected" (for lack of a better term), I could also use a continuous variable $c$ in the domain $[0, 1]$ as given:
$$ c \leq b_s \quad \forall \, s \in S \\ c \geq \sum_{s \in S} b_s - |S| + 1 \\ b_s \in \{0, 1\} \quad \forall \, s \in S \\ c \in [0, 1] $$
Even though $c$ is continuous, in a feasible solution it can only take the values 0 or 1 due to the constraints.
What are the dis/advantages in this approach compared to describing $c$ as binary variable when solving such a model using existing solvers (CBC, Gurobi, CPLEX, etc.)?