Naive Bayes and numeric (continuous) features

I have a dataset with data instances that contain 20 numeric features and a class label. For the purpose of classification using Naive Bayes (with which I'm already familiar), how does Weka's implementation of Naive Bayes (e.g. "NaiveBayes" or "NaiveBayesSimple") handle continuous features, particularly when computing the likelihood probability portion? I know that some common approaches to treating continuous features are: (1) discretizing the values, (2) using Gaussian PDF, or (3) calculating a sum of kernel probabilities. A 2004 paper by R. Broukaert suggests that the discretization method is usually the best. Which approach does Weka take for Naive Bayes? Is the same approach taken for all the other classifiers?

Thank you for any help.