Knowledge of the orbits of visual binary stars has always been one of the fundamentals of astronomy. Based historically on the visual measures, nowadays the orbits rely more (or exclusively) on the accurate speckle data. This prompts reconsideration of the methods of orbit calculation, undertaken here and illustrated by 20 examples, from accurate to drastically revised and tentative orbits. Good understanding and critical assessment of the input data is a key requirement, especially concerning visual measures. Combination of visual and speckle data is still needed for long-period binaries, but the relative weights must match their respective errors. When the orbit can be fully constrained only by accurate speckle data, the old measures should be ignored. Orbits can be classified into three grades: (A) fully constrained, (B) semiconstrained, and (C) preliminary or tentative. Typical use cases of visual orbits are listed. Accurate parallaxes from Gaia, together with the orbits, will greatly expand the data on stellar masses. Continued speckle monitoring will be an essential complement to Gaia, but the vast amount of new pairs will restrict future work on orbits to the most interesting or relevant objects.