We discuss strategies for color separation and recombination, regarding 2-color pump-and-probe experiments planned for the DESY XUV FEL Phase 2. We show that efficient handling of the pump and probe components, with good spectral and spatial separation, is possible with either filters or gratings; the former are cheap while the latter are continuously tunable over a range. Nice spatial overlap of the pump and probe on a spot 5 x 8 mu m(2) FWHM, limited by surface slope errors on the optics, can be achieved with a Kirkpatrick-Baez pair of cylindrical elliptic mirrors placed about 500 mm away from the sample position. Thermal loading is also discussed. It is found that the filters cannot survive more than a few dozen FEL pulses in a train, while a grating on a water-cooled silicon thin blade will present unacceptable surface slope errors after about 1800 FEL pulses on a train. (c) 2005 Elsevier B.V. All rights reserved.5453568577