The context of this discussion comes from CASSANDRA-293. Since it relies on keys, current hinted handoff scheme isn't going to work for when a range-remove operation needs to be hinted for a downed node. The idea I'm playing with now is to use a store-and-forward mechanism for hints where the entire message is stored and later sent to the destination when the host comes back up.
Cons to this approach are additional storage expense and inefficiency (since there is no way to tell if a particular message no longer matters). Pros are that it gives us a degree of latitude if/when other mutation operations (like 293) come along that aren't single-key based. Does anybody have objections or a better idea? I'm not stuck on this approach, but it was what came to mind first. Gary Dusbabek
