> On 9 Mar, 2016, at 21:18, Wesley Eddy <w...@mti-systems.com> wrote: > > > Minor issues: Treatment of packets that don't fit into the hashing > > classification scheme: The default FQ-CoDel hashing > > mechanism uses the protocol/addresses/ports 5-tuple, but there will be > > packets that don't fit this scheme (especially ICMP).
I believe the current implementations treat “ICMP” as part of the protocol (viz. TCP/UDP), and set the port numbers to zero, then hash as normal. This stochastically provides a separate queue without explicit special handling. I agree that this is worth mentioning in the spec. ARP is another protocol worth noting here. The only addresses reliably available are at the MAC level. In fact, these are general problems for any flow-isolating qdisc, which is why Linux now provides a sophisticated (if slightly obtuse) flow-dissection mechanism which accounts for these factors and yields a straightforward hash value. Perhaps a document dedicated to discussing this problem would be valuable, if it doesn’t already exist (and if it does, it should be referenced). - Jonathan Morton _______________________________________________ aqm mailing list aqm@ietf.org https://www.ietf.org/mailman/listinfo/aqm