Hi, >Any reason we only use the kernel diff for big child node lists?
I'm not sure, but I guess the idea was that when the list of child nodes is small, the diff() method is slower. Maybe it's just that the threshold is still too high, for example by a factor of 10 or so (KernelNodeState.MAX_CHILD_NODE_NAMES = 100 currently; it used to be 1000). We might need to tune the value. Or possibly the threshold should be 0. It's great to have benchmarks! That way we can easily find out what the problem is. Regards, Thomas
