For a subsorted signature Sigma the subsorted models are ordinary many-sorted models for Sigma# that satisfy the following properties (which can be formalized as a set of conditional axioms):
[AT] Is this really unambiguous? I am not sure what a "partial inverse" is, but with a bit of bad will it might be defined so that the composition of a partial function and its partial inverse in both directions is a subset of identity. Here of course the point is that one of these compositions is the identity, and the other is a subset of identity.
Discharged: Adjust wording.
[DTS,AT] Unclear. I do not like the verb "commute" here: with the current definition of overloading of functions, what we mean here need not be commutativity as such. Perhaps "are compatible" would be better?
Discharged: Adjust wording.