“…Demangeon and Honda actually define subtyping for input, which in their presentation is unified with branch, to be covariant in the message type, and conversely subtyping for output, which they unify with select, is contravariant. It seems that this is just a typographical error, as other papers in the Honda et al style, for example by Chen et al (2014), use the expected variance in the message type. In some definitions of subtyping, for example by Mostrous and Yoshida (2015), the variance depends on whether data or channels are being communicated; the technical details need further investigation to clarify this point.…”