From: Joaquin Miller (firstname.lastname@example.org)
Date: Tue 10 Sep 2002 - 00:31:08 BST
>... our definition of the semantic domain is cast as an OO model, rather >than in set theory. There are two reasons for this. (1) it leads us to >tools (such as those discussed in a previous posting) faster (2) it seems >easier to get OO models past the folks in the OMG than set theory. This is not only very clever (2) and practical (1), i expect it is sound. And it is certainly good for helping folks understand--so long as what is going on is made clear. The way this is said here does seem to confirm my understanding that what 2U is doing is formal semantics, just using its own domain, rather than the customary sets and structures of sets (such as the famous possible worlds). So i remain comfortable in saying that we also need to consider what a model means. This is where 3C and 2U could cover each others backs, since we do nothing formal, but a fair-to-middlin' job on meaning. You will recall that we say: "Because the UML 2 language is an order n language, there is guaranteed to be a set theoretic formal semantics. This guaranteed formal semantics will consist of a standard set theory and a semantic mapping from UML 2 to that set theory, providing an interpretation in set theory of any UML statement. In fact, we believe that UML is best served by a third order language with a Henkin semantics; but we leave formal analysis to others. We do not provide a formal semantics, but focus instead on providing a meaning: a clear, clean, concise and exact meaning." http://www.omg.org/cgi-bin/doc?ad/2002-09-15 at Part II Section 3 Meaning and Semantics, p. 26. ....... www.community-ML.org Footnote: By the way, i asked John Corcoran if he could point me to a reference for the the theorem about a guaranteed set-theoretic semantics for an order n langauge, and he said he would be surprised if anyone had bothered to prove that !