The probability assessment of the dependability of a system considers the failures and repairs of its components as random events. Their time occurrences are described by stochastic temporal variables. An important particular case of stochastic finite state automaton (SFSA) is the one where all the transition rates are constant in time; it is then equivalent to a continuous time and discrete state Markov process. The SFSA is an abstract model. Its main interest is to be used as a basis of a simulation model. In order to do this, we have to add a time support. The number of histories to be simulated depends on the accuracy requested on the parameter to be assessed, that is to say the probability for the result of being comprised in the accuracy interval. This has been implemented in the Scilab/Scicos environment, a free software developed by INRIA.