Many micro- and nanoelectronic, micro- and nanooptic devices include different films and layers deposited at the surfaces of dielectric substrates. Properties of these films and layers sufficiently depend on properties of the substrate surface. Therefore substrate surface preparation before layers and films deposition is very important. One of the main parameters of the surface is its roughness that influence on such parameters of thin layers as specific resistance, electric strength, chemical resistance etc. By mechanical treatment of the surface its roughness is defined by the size of polishing powder grain. That is why preparation of the surface with low roughness is laborious one. Besides, preliminary mechanical grinding with coarse-grained powder causes appearance of near-by-surface cracked layer, which can be removed by deep mechanical, chemical or flame polishing. Laser polishing is sort of last one. The problem of laser polishing was repeatedly discussed. 10.6 μm CO2-laser is the only acceptable one, as all glasses well absorb radiation at this wave length. The regime of continuous radiation is used in most cases, for which typical duration of influence by surface scanning is several milliseconds. Though the beneficial effect was obtained (in before mentioned and other papers), the technique is not wide spread. In our opinion the reason of this is that there are problems connected with phenomena specific for laser polishing: hydrodynamical waves in the softening layer and thermomechanical tensions in the substrate. Both of them are aggravated by the three-dimensional locality of laser action. This situation can be changed by using of short nanosecond laser pulses - in this case the heated layer of substrate decreases and the problem becomes not so critical.