Proteomics has traditionally used the separating power of two-dimensional electrophoresis for the quantitative analysis of protein amounts in complex extracts. However, the limitations of this approach in terms of throughput and analyzable protein range have elicited the development of other proteomics approaches, based either on peptide separations instead of protein separations, or based on direct protein recognition and selection on dedicated arrays (protein chips). These recent methods seem very promising, and probably look more promising than they will ultimately be, just because their weaknesses are not fully characterized yet. The purpose of this paper is thus to highlight the strengths and weaknesses of all the proteomics approaches proposed to date and to try to deduce the respective niches in proteomics that these approaches will have in the future.