--- In executableuml@yahoogroups.com, "Hagstrom, Erick G."
<ehagstrom@...> wrote:
> The suggested change in semantics is not required to make any arbitrary
> Executable UML domain model deterministic, nor does it seem desirable,
> nor would it make the Wikipedia description correct in the general
> case.
>
I actually put that wording into the Wikipedia entry. I'm not sure
what prompted me to use "deterministic system" vs "system", but the
Wikipedia description isn't completely wrong. The message 704 example
is non-deterministic by the rules of Executable UML, but it will
produce a deterministic executable, for a given architecture.

It would be a really bad candidate for reuse, and therefore a bad
Executable UML model. The question that arises is, "Can a bad model be
termed an Executable UML model?" For instance, you can do procedural
programming in C++, but if you showed the code to an experienced C++
programmer, they would likely say, "This isn't C++! It's C with a .cpp
extension!"

You wouldn't need an experienced Executable UML modeler to point out
the flaw; a static Executable UML model parser could point out that
the second action statement is non-deterministic.