I think we can>> get durability out of HDFS-200 (and a few other patches) and stay on a>> fully released, tested, and (most importantly) trusted Hadoop release.>> Distributed filesystems are hard, and there is a reason there are>> very few actually in real production use out there.