The quality measurements in software generally give the number of errors (against the objectives of the product and normalized by size). The interesting number for the user is the mean time to failure with the possibility of using execution time (CPU or processor) or calendar time. In this paper time dependent reliability models are evaluated with respect to execution and calendar time, and the execution time domain is found to be highly superior (which is also intuitively obvious). It should be noted that execution times for a multitude of parallel users will converge towards calendar time (which is easier to measure).