Analogous to the replacement of an old machine, such as a car, replacing an aged software may contain its escalating cost of maintenance. Prior research has assumed that increasing maintenance cost is due to the deterioration of the system maintainability. However, cost of maintenance depends also on the number of incoming maintenance requests. While software maintainability is determined by its complexity and development environment, number of maintenance requests is affected by the business environment. This distinction is significant in analyzing economic uadeoffs in software maintenance because replacement improves system maintainability but will not affect the number of inalntenance requests. Unlike replacement of hardware, rewriting software takes an extended length of :&ne. Thus, the old software muststillbemaintainedbeforethenewsoftwareisready. Wedevelopaneconomicmodelthatconsiders the number of maintenance requests and the rewriting period explicitly. The model is an extension of Gode, Barua, and Mukhopadhyay (1990), which assumes a constant number of maintenance requests and instantaneous replacement. Our model allows us to draw some additional p o k y implications about software maintenance arid replacement. For instance, we show that in certain situations delaying a system replacement can be more cost effective when the user environment changes more rapidly, conuary to our intuition. Moreover, it is shown that rewriting should begin earlier when the instantaneous replacement assumption is relaxed.