Interferometry is one of the most precise methods for optical testing. Nevertheless, interferometers are not ideal, they contain internal residual aberrations, and the methods for interferograms analysis are subjected to noise effect and depend on the interferogram parameters. The practice of the modem interferometry faces the following contradictions: on the one hand, up-to-date software allows processing an interferogram at an error of A1200-A1500, on the other hand, a really reached error of an interferometer does not exceed ?/50. Well-known interferometer designs applied in optical testing have been analyzed. A design of interferometer containing aligned branches was chosen. It is less sensitive to vibrations and more compact, allowing one to test surfaces with a larger numerical aperture. This is design of an interferometer, whose principal element represents an aplanatic lens, low sensitive to decentering. The interferometer is stable when defocusing the recording system. At the restoration of the wavefront from the object tested, samples of interferograms were used differing in the number and orientation of the fringes, which allows one to improve the test precision. This occurs due to the shift of the object signal in the area of spatial frequencies. In this case the estimate variance of the wavefront decreases approximately proportionally to the amount of sampling of interferograms. The present paper analyses the factors that influence the precision of testing the shape of the piece surface, the possibility of design and programmed algorithmic correction of the factors influencing the testing precision. An interferometer where such engineering means of correction would be implemented could be called 'an ideal interferometer'