Classification with Naive Bayes when certain features are missing.

Hi all,

I understand that Naive Bayes models can be trained with uncertain data (e.g. some of the cases have unlabeled features). However, in all of the cases I've seen, the Naive Bayes model only can classify a case where all the features are known. Does anybody know the defacto method used to classify if features are missing? Selecting the most probable state in light of the given features would be an approximation, but is there another way?