Couldn't have said it better myself. I would note that many of the no votes
came from vendors and/or implementors (as did some of the yes votes). I am
sympathetic to Jarek's concern that we provide a useful model but am not at all
convinced that requiring barrier synchronization is a fatal flaw.
David makes the excellent point that if we are mandating portability then we
*must* have some sort of sync points in order to maintain coherence. A truly
1-sided API that did not require sync points would work just fine on SGI
machines, but what good is a standard if only one or two machines can support
We now have a real opportunity to simplify the get/put model. We should embrace
-- Eric Salo Silicon Graphics Inc. "Do you know what the (415)933-2998 2011 N. Shoreline Blvd, 8U-808 last Xon said, just firstname.lastname@example.org Mountain View, CA 94043-1389 before he died?"