A minor point on this:
The assertion that AtomicOutcome is
tested in the interop scenarios but MixedOutcome is spurious. Since the WS-BA
scenarios concern only a single coordinator-participant, the only effect of the
AtomicOutcome/MixedOutcome flag is what the uri is. (which was of course
different when the interop was done in Jan 2005). Neither are tested in the
interop in the sense of checking the behaviour. There are no tests for
AtomicOutcome either - so new tests are needed for that too if it is to
remain.
Utility of BOTH AtomicOutcome and MixedOutcome needs to be argued for.
That's not the subject of this email.
Peter