Because large arcs are highly magnified images, cluster lenses can be used as natural giant telescopes for photometry and spectroscopy of individual distant galaxies and may allow us to detect the most distant objects ever observed (Nottale 1988). In addition, for a given potential, the shear pattern and the arclets distribution around cluster lenses have an angular scale which depends on the redshift distribution of the background sources (Kneib et al. 1994, Bonnet et al. 1994, Smail et al. 1994a). Consequently, cluster lenses can be also used as natural spectrographs with poor spectral resolution to statistically infer the distances of the faintest galaxies by going much deeper than the faintest redshift surveys. However, in spite of this considerable interest, very little work has been done in this field, due to the observational difficulties of getting redshifts of such faint objects or in getting the high image quality needed for the proper use of the geometry of arclets.