I think that a week or two of reflection on test cases 5 and 10 would
OK, just so that we can move along and get the other things done that
we need to get done for our deadline this week, I'm going to suggest
that we postpone a final decision on test cases 5 and 10 for the
community review document.
Now, as far as the process goes, I don't think we can postpone a
decision. So here is my suggestion.
* In the community draft, give the stronger model, that prohibits 5 and 10,
as the proposed spec.
* Note in the community draft that the expert group is seriously considering
relaxing the specification to allow the behavior in 5 and 10, with
request for feedback on this issue. Describe the changes to the spec to
generate the relaxed model. Lay out some of the pros and cons.
The reason for structuring things this was is as follows. If, after
community review, we wanted to strengthen the model, someone could
complain that this could potentially break their JVM implementation
and cause lots of process-related problems for the JSR.
On the other hand, if after community review we decided to weaken the
model, then any JVM's compliant under the strong model would still be
compliant under the weak model, and I don't think anyone outside of
the expert group and JMM community would have any grounds for
objecting to the change.
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:50 EDT