We study the hypothetical conditions for interstellar clouds dense enough to produce glaciations on the Earth. A simple differential formula (adequate to give lower limits to dust absorption) is used to relate mean temperatures and visual albedos today and during the glacial eras. For this, the geological and oceanographical records of the Pleistocene are used. The temperature decays are associated to an absorption of the solar light in visual magnitudes mu. As the effective albedo, integrated in all wavelengths are lower than the corresponding visual value, the adoption of a visual scale leads to an underestimation of the actual amount of dust. A minimum dust absorption Am0 = 0.02 mag, necessary to start a glacial era is then obtained. This should mean interstellar clouds with dust densities of 4100 mag. PC-' and sizes of 0.3 pc or more, taking into account the time span of the glacial eras and the mean velocity of the Sun with respect to the LSR. Such clouds were never observed and are uncompatible with what is known from the interstellar medium: the 'glaciation clouds' should be clouds with densities 50-100 times above the tolerable value for gravitational stability; on the other hand, the necessary number of clouds per cubic parsec should produce the collapse of the galactic disc as a whole. From a comparative analysis of the temperatures of the other planets it seems to be that the actual thermal temperatures in their surfaces depend less than one expects from the visual albedos. From this it is raised the suspicion that the cause of the ice ages was the Sun itself.