The ultimate goal of lattice QCD is to simulate dynamical (u,d,s,c,b) quarks at their physical masses, with sufficiently small lattice spacing and large enough volume such that the discretization and finite-volume errors are both well under control. In this talk, I discuss the theoretical aspects how this goal can be realized. Moreover, I perform the first exploratory study of lattice QCD with dynamical (u, d, s, c, b) optimal domain-wall quarks, where s, c, and b quarks are physical, while u and d quarks are heavier than their physical masses. Even with unphysically heavy u/d quarks in the sea, the mass spectra of hadrons containing valence s, c, and b quarks are in good agreement with the high energy experimental data. This implies that it is feasible to simulate (u,d,s,c,b) optimal domain-wall quarks at their physical masses, with sufficiently large volumes.