Abstract: Techniques to efficiently filter fragmented datagrams and route fragments are described. For each fragmented datagram, a filtering node obtains filter parameters as fragments for the datagram are received. When all filter parameters are available, the node applies one or more filters on the filter parameters to obtain a filter result for the datagram and stores the filter result in an entry in a routing table. Prior to obtaining the filter result, the node stores all fragments received for the datagram in a memory. When the filter result becomes available, the node processes all fragments already received for the datagram in accordance with the filter result. As each remaining fragment for the datagram is received, the node immediately processes the fragment in accordance with the filter result. When the last fragment is received, the node clears the memory and the routing table entry for the datagram.