Issue 222 - Proposal for vote

I formally propose the resolution below:  rescind our previous 
resolution, and modify it to indicate that in the case of a reply to a 
request-response invoke that violates a correlation constraint, the 
reply *is* delivered to the process, and correlationViolation is thrown.

Danny van der Rijn wrote:

> One of the dangers of moving as fast as we did this meeting- I think 
> we messed up part of 222.
> We said that an inbound message MUST NOT be delivered to a process if 
> the message does not match the correlations.  The case we didn't think 
> about is What happens if the an <invoke> is on a request/reply 
> operation, and the *reply* fails to match the correlation.
>
> My preference would be to rescind our resolution, and modify it to 
> indicate that in this case, the reply *is* delivered to the process, 
> and correlationViolation is thrown.
>
> Danny
>