A couple years ago we made the decision to the elements certainty, precision (and respons?) to model.glossLike so that they would be available on a few so-called empty elements such as gap and space. (See ticket http://purl.org/TEI/FR/2862151\).
I now think this was too blunt an instrument: while it is probably true that whenever glossLike elements are wanted, certainty etc. would also be useful, the reverse is not necessarily the case.
I suggest therefore that we create a new model (perhaps "respLike" or "certLike"?), for the three elements that point to specific subparts of a target element, and make this model a member of model.glossLike. This way everything that has glossLike gets certLike, but certLike can also be included on its own when <desc>, <gloss>, etc. are not wanted.