The prevailing opinion in most of the world is that drug susceptibility testing of Mycobacterium tuberculosis isolates may not be necessary, especially in countries where the diagnosis of tuberculosis is based predominantly on sputum smear examination without culture isolation, and even if it were functionally important it is too expensive to be practical. However, we believe that the spread of primary drug resistance calls for immediate action to prevent the real threat in a number of countries of an epidemic of incurable polyresistant tuberculosis. One step in this direction, among other necessary measures, is implementation of a system for timely detection of drug resistance in new patients. An overview of the currently available methods for drug susceptibility testing, and those under development, is presented in this article. Our aim is to stimulate discussion regarding where, when, and which methods can and should be implemented now, and which of them has the best potential for the future. This overview includes phenotypic approaches based on conventional cultivation and other techniques, as well as perspectives for the genotypic principle. Regardless of the methods chosen for now or to be implemented in the future, they are all laboratory methods, and we contend that dreams of a revolutionary method which can do away with the laboratory are unrealistic, at least for the time being. We argue that it is essential to create a system of direct, centralized laboratory services in order to make drug susceptibility testing reliable, practical and affordable. Such centralized laboratories would provide both more timely and trustworthy diagnosis of tuberculosis, and facilitate detection of primary drug resistance. Although such systems would require significant initial investment, we suggest that, over time, they would prove highly cost-effective for many countries.