Comment 8 for bug 379264

I am know leaning toward trying to hack file_io and pushing file_io writes to the tablespaces to memcached.... I think this will eliminate the secondary index issues. Additionally, I think we can then skip the "mru" memcached patch. Only an actual write will invalidate whats in memcached. I think we may need to push to memcached on read from disk ( maybe ) in order to get that to work.