We investigate how galaxy-galaxy lensing measurements depend on the knowledge of redshifts for lens and source galaxies. Galaxy-galaxy lensing allows one to study dark matter halos of galaxies statistically using weak gravitational lensing. Redshift information is required to reliably distinguish foreground lens galaxies from background source galaxies and to convert the measured shear into constraints on the lens model. Without spectroscopy or multi-colour information, redshifts can be drawn from independently estimated probability distributions. The COMBO-17 survey provides redshifts for both lens and source galaxies. It thus offers the unique possibility to do this investigation with observational data. We find that it is of great importance to know the redshifts of individual lens galaxies in order to constrain the properties of their dark matter halos. Whether the redshifts are derived from $UBVRI$ or the larger number of filters available in COMBO-17 is not very important. In contrast, knowledge of individual source redshifts improves the measurements only very little over the use of statistical source redshift distributions as long as the source redshift distribution is known accurately. Comment: 5 pages, 1 figure, submitted to A&A, major changes and extensions