I agree with most of the commit, but the reason I left of some of the proto_tree_add_text/proto_item_add_subtree calls alone in the dissectors was that I thought they should be converted to use proto_tree_add_bitmask (or something similar). When I did the original "conversion", I just didn't want to take the time to create the necessary hf_ fields to make that possible. To me leaving in the proto_tree_add_text makes it easier to detect the need for proto_tree_add_bitmask (which is a function that should really be used more in dissectors), than a group of proto_tree_add_items.
Michael