For the distant clusters, a number of specialized techniques have evolved [9, 4, 38] that optimize the accuracy to which the LF can be measured at faint magnitudes. A recurrent theme is the importance of characterizing the background/foreground contamination and its uncertainty, as this is the main source of the error in the faintest points of the LF. The contamination is severe: the background/foreground galaxy counts are comparable to or greater than the cluster counts at R > 19, even in the centers of rich clusters like Coma. For example, high photometric accuracy in the zero-point needs to be achieved because the background counts are a strongly-varying function of magnitude and we must be sure that we are subtracting exactly the right amount of background galaxies at each magnitude. Other systematic sources of error (e.g. Galactic extinction, possible extinction from dust in the clusters, stellar contamination, globular cluster contamination, giant galaxy halo substructure, distortion of the background counts by gravitational lensing by the cluster dark matter) are important too and all need to be considered. The reader is referred to the papers listed above for details.
Measurements of nearby clusters as outlined in Section 4 pose different problems. Background subtraction is no longer a major problem because the faintest dwarfs have scale-lengths that are significantly larger than the seeing so that cluster members can be identified by their morphologies. Therefore the LF of Virgo measured by Phillipps et al.  is very well constrained at the faint end. Measurements in diffuse groups like Ursa Major are more difficult because counting statistics there are poor. Also, the surface density of low-surface-brightness (LSB) field galaxies which look similar to cluster dwarfs needs to be well characterized (for Virgo this is not a serious problem because the cluster dwarfs outnumber background LSB galaxies by a large factor in this relatively dense cluster - see the paper by Jones et al. in these proceedings).