This paper presents the basic principle and theoretical relationships of an original method allowing to retrieve the Wavefront Errors (WFE) of a ground or space-borne telescope when combining its main pupil with a second, decentered reference optical arm. The measurement accuracy of such a "telescope-interferometer" is then estimated by means of various numerical simulations, demonstrating a high performance excepted on limited areas near the telescope pupil rim. In particular, it allows direct phase evaluation (thus avoiding the use of first or second-order derivatives), which is of special interest for the co-phasing of segmented mirrors in future giant telescopes projects. We finally define the useful practical domain of the method, which seems to be better suited for periodical diagnostics of space or ground based telescopes, or to real-time scientific observations in some very specific cases (e.g. the central star in extrasolar planets searching instruments).