Geoff Huston writes: > > No clean answers here, but for UDP I would suggest that higher (MTU) > is marginally less worse than lower. Concur. This matches the experience we got from the experiments we (Willem Toorop and I) did some time ago. jaap