I am aware a similar question had been answered before on the site, see(A question about complement of a closed subspace of a Banach space). I find the answer given there unsatisfying since it relies on a categorical definition of direct sum where as my definition is strictly algebraic.
In my book the notion of a direct sum simply asserts a unique presentation of any element in $B$ as a sum of an element in $M$ and an element in $N$ (both of which are linear sub-spaces of $B$). In that case is it still true that $N$ has to be closed?