We propose a new approach to the analysis of limb-scanning measurements of the atmosphere that are continually recorded from an orbiting platform. The retrieval is based on the simultaneous analysis of observations taken along the whole orbit. This approach accounts for the horizontal variability of the atmosphere, hence avoiding the errors caused by the assumption of horizontal homogeneity along the line of sight of the observations. A computer program that implements the proposed approach has been designed; its performance is shown with a simulated retrieval analysis based on a satellite experiment planned to fly during 2001. This program has also been used for determining the size and the character of the errors that are associated with the assumption of horizontal homogeneity. A computational strategy that reduces the large number of computer resources apparently demanded by the proposed inversion algorithm is described.