Note: timestamps in PDT. Connection details • Date: 22 September 2017 7-8am Pacific http://www.worldtimebuddy.com/?qm=1&lid=100,12,5392171,1850147&h=100&date=2017-09-22&sln=14-15 • Webex Recording: https://cisco.webex.com/ciscosales/lsr.php?RCID=fc612b41b955470f82876b586c243599 o Recording password: Si67iq7D • Material link: https://bitbucket.org/6tisch/meetings/raw/master/170922_webex/slides_170922_webex.ppt Present 1. Diego Dujovne 2. Fabrice Theoleyre 3. Jonathan Munoz 4. Keoma Brun-Laguna 5. Malisa Vucinic 6. Michael Richardson 7. Pascal Thubert 8. Qin Wang 9. Remy Liu 10. Robert Barton 11. Simon Duquennoy 12. Sumankumar Panchal 13. Tengfei Chang 14. Thomas Watteyne Taking notes (using Etherpad) 1. Michael 2. Pascal 3. Thomas Action Items • TODO o Pascal to trigger WGLC for 6P o Thomas to add text in 6P revision as indicated below Agenda • [07:05] Administrivia [10min] o Note-Well, Scribes, Agenda Bashing o Status of drafts (WGLC / forthcoming WGLC) o Last meeting todos • [07:15] Using the datapath for faster local repair (Pascal) [15min] • [07:30] Influence of Link Metric on reliability; ETX^n (Simon) [10min] • [07:40] Open Discussion [18min] • [08:00] AOB [ 2min] • TODO Minutes • [07:05] Administrivia [10min] o Note-Well, Scribes, Agenda Bashing o both approved o Status of drafts (WGLC / forthcoming WGLC) • new milestones published, • Last meeting todos o Pascal to publish the agreed milestones o Pascal to publish Michael's zerotouch 00 doc [confirmed!] o Thomas to review the draft 6P o Xavi to publish (may delay to include the relocation discussion) o Thomas to propose a new meeting next week (Friday same time?) to discuss further RPL lossless. • 6top changes o Thomas: there is now a relocate comment. Node A asks node B to move one or more slots somewhere else in the cell list. Neighbor answer by success, I moved thiose there. o Thomas gives examples from the new 6top protocol document (08) just published • 6top status: almost published for some time, 08 published today o Q on the ML waswhether there is a way to move not the first but only the second cell in the list? o Thomas: I think thios is over complicated o Qin: yes that is the ML discussion. At the beginning we wanted to use combination. In case only one can move out of 2, should we move one or delete 2? o Thomas: this is a case of collision, we do not want to delete the cell. We need to add text on that. o Tengfei: Same case as delete; if there is only one that exist it should return error code  action item: add IOW what happens if a relocate command is issue for a cell that doesn't exist  action item: add sentence which says that for cells that couldn't be relocated, do not delete o Remi: If teh 1st fails to relocate than the second will not be relocated either o Thomas yes, all cells are equivalent, and we move the first first. For non equivalent cells, use separate commands • [07:15] Using the datapath for faster local repair (Pascal) [15min] o Pascal presents slides.. o Q: do we take the risk of attaching to a parent with a higher rank (risks creating a loop) o is there a way to know that B is a better parent? (slide 10) o MCR: asks if detection packets are real packets, Pascal says they are probably ICMPs to the DODAG root (slide 11) o Simon: does something different, when we switch parents, if the parent has a higher rank, we do not send extra messages, just send real data, and look for RPI direction inconsistency.... we keep forwarding the packet to new parents. o Simon: What if the inconsistency occurs before you detect it? This is different from standard RPL, because the looped packet is not dropped. o Ensuing discussion about count to infinity and other things about slide 12, which were too fast to capture. • [07:30] Influence of Link Metric on reliability; ETX^n (Simon) [10min] o Simon: 5-9's of delivery. o TW: downward traffic... storing? non-storing? Simon: both. o TW: why is a 50% is worse than two 100%? Simon: because 50% link will eventually lose traffic. o Simon: ETX tells you that the energy cost of the two situations are the same. o Simon: explains that the as the n (ETX^n) goes up, the churn seems to increase o Pascal: suggests that hysterisis might solve that. o Simon: explains that "LR" is Loss Rate. o Simon: points out tht the upwards link success seems to go up as well. o Simon: quick take through non-storing state to fix things. • [07:40] Open Discussion [18 min] • [07:58] AOB [ 2min]