880: I don't think this is important as regions already do that (for a set of nodes though). What's the compelling use case for per-node eviction ? So a -1 from me 841: I think we should solve this problem fundamentally in a different way. Manik and I thought about functors, simple functions which (following the Visitor pattern) iterate over the nodes, e.g. in in-order, depth first order. So think of it as a closure that's applied to all nodes of a given subtree. This way we could implement sorted lists, map-reduce, *query* and in general, have a generic mechanism to transform a tree. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=3990374#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...