Weekly Hangout Notes 23 March 2017

Notes taken and forward by Jason Eggleston

• gleb@ reports that julian has benchmarked listen sockets branch, no difference, which is expected for a single socket, promised to do more testing

• tuexen@ asks about PR-217637: what should happen on server side RST of a connection? jegg@ thinks we should transition to TIME_WAIT when the server decides to RST a connection, to prevent spurious packets from creating new sessions. tuexen@ and gleb@ suggest that we reject syncookie if we have not dropped syncache state recently.

• tuexen@ also wonders what should we do with half closed tcp sessions, specifically if we half close the read side, and we receive data, should we send a RST or should we continue silently throwing away the data like we do now.