Update from last meeting: hiren@ is looking into an issue where not going back to slow start causes a lot retransmits. Update: Jason thinks they have a fix?
Jason Eggleston reports that draft BBR code from rrs@ results in large end user metrics improvement vs LLNW's current proprietary TCP stack
Update from last meeting: Listen sockets. Update: glebius@ sees performance improvement in his tests, waiting for update from jch@
gallatin@ reported on his mbuf work. He is working on making mbufs point to an array of buffers. It makes a 20% performance improvement for his workload. But, mtod() loses meaning, which means it will take quite a while to smooth out the rough edges.
rrs@ reports a potential problem with parts of the inpcb not being zero'd.
glebius@ suggests we also make IPv4 and IPv6 information be anonymous union.
glebius@ also suggests that we remove spares from inpcb.