Stopword lists are really only justifiable by specific research interests. The default should always be no stopwords, with the ability of the user to implement any stopword list they wish, as long as it is readily discoverable. I'm not sure that there is any point in trying to develop a reasonable consensus about stopword lists since research interests will vary so greatly and unpredictably. With any modern inverted index full-text database speed should not be a consideration across all of digitized classical literature.