Software Reliability is considered the key factor for software system quality. Several models have been presented to estimate and predict this reliability, and Software Reliability Growth models (SRGMs) are the most commonly used for this goal. SRGMs assumptions, applicability and predictability are important issues and need to be addressed. This report discusses these issues, introduces time series models as a good solution to address them, and suggests a generic framework for reliability prediction at the different software development stages and at runtime.