The "writing dirty paper" capacity result crucially dependents on the perfect channel knowledge at the transmitter as the presence of even a small uncertainty in the channel realization gravely hampers the ability of the transmitter to pre-code its transmission against the channel state. This is particularly disappointing as it implies that interference pre-coding in practical systems is effective only when the channel estimates at the users have very high precision, a condition which is generally unattainable in wireless environments. In this paper we show that substantial improvements are possible when the state sequence is drawn from a discrete distribution, such as a constrained input constellation, for which state decoding can be approximately optimal. We consider the "writing on dirty paper" channel in which the state sequence is multiplied by a fast fading process and derive conditions on the fading and state distributions for which state decoding closely approaches capacity. These conditions intuitively relate to the ability of the receiver to correctly identify both the input and the state realization despite of the uncertainty introduced by fading.