Domain decomposition of stochastic PDEs: Theoretical formulations