“…The original proof of Lempert's approximation theorem [21], pp 338-339, relies on the general Néron desingularization, a deep and difficult result of commutative algebra for which the reader is referred to [1], [26], [27], [28], [29], [30]. Theorem 1.1 is expressed in terms of analytic geometry and has had numerous applications in the theory of several complex variables (see [5], [11], [12], [13], [21], [23], [25], [31]). It is natural to ask whether one can replace Néron desingularization by simpler geometric methods.…”