Nathaniel makes some very valid arguments for having a separate encoding
header and my only argument against it is silly-state elimination and the
elimination of the concept of an encoding for MESSAGE and MULTIPART. Either
way, it isn't of major consideration to me.