>Would it be possible to propose two models, one "for programmers" and
>the other "for implementors"?
It is far too late to be considering such things. All JSRs that have
any hope of getting into Tiger should be converging *now*. Expert
groups working on these JSRs should now consider convergence to be of
paramount importance, and should--even more than ususal--be willing to
sacrifice secondary goals to get their specs out the door. It doesn't
matter how good a spec is if it misses the train.
This is not to say that we should sacrifice quality, but that we should
be well attuned to what is good enough.
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:47 EDT