Options and trade-offs for the POLLUX spectrograph optical design are considered. All the design problems are demonstrated on the exaple of the medium channel working in the range of 120-220 nm. The most of the current requirements can be fullfilled in a relatively simple optical scheme with an echelle grating and a single concave grating acting as a cross-disperser and camera. It's shown that the critical point for this design is the spectral length of the shortest order. With the gratings and the detector technologies available nowadays it's impossible to reach the target value of 6 nm in the shortest order.