[Special thanks to Alexandru Petrescu for taking the meeting notes] o IP over IEEE 802.16 Networks Working Group (16ng) o 68th IETF - Prague, Czech Republic o Chairs: Soohong Daniel Park (soohong.park@samsung.com) Gabriel Montenegro (gabriel_montenegro_2000@yahoo.com) =================================================================== [Monday Session] o #1 Timeslot 2 hours o 1520-1720 Monday March 19 o Afternoon Session II, Congress I 05 Administrator Chairs - Blue sheets - Jabber scribe - Agenda bashing 50 WiMAX Networking implication for 16ng Max Riegel - Abstract: The Networking WG of the WiMAX Forum has defined a mobile radio access network architecture for providing IP connectivity over IEEE802.16/802.16e. The presentation shows the adoption of the work of the 16ng WG by the WiMAX network architecture putting emphasis on the implications coming out of the WiMAX network. Dave Thaler: Clarification, if ASN supports one of them is there only one per CSN or mixed ? Max Riegel: IPCS not visible to CSN Dave Thaler: Would ETHCS and IPCS both ever be used by same ASN that CSN is serving Dave Thaler: not taking one type of MS using IPCS and the other ETHCS and try to put these together in same CSN, not doing that Max Riegel:Not specified yet how ETHCS is served in CSN Unknown: much possible that same terminals, IPCS and ETHCS committed to same CSN Max Riegel: not on top of CS Max Riegel: GRE tunneling is used just to provide management of service flow, if you try to implement switch network you will run in trouble because you have to differentiate service flows. CID is not enough for that. GRE tunneling endpoint is the base station and IP interface of gateway Max Riegel: using VLANs won't work in mobile environments, not even in stationary due to limited of address space Unknown: how do you treat scalable multicast environments?, hundreds of subscribers, who's doing the actual replication to hundreds? Max Riegel: in Release 1.0 no multicast, so you have to do it in AR and bridge. It seems not solution sufficient and agreed 20 IPv6 transmission over IPv6CS Basavaraj Patil - ID: draft-ietf-16ng-ipv6-over-ipv6cs-08 - Status: IESG Evaluation - GOAL: Clean up issues raised by reviewers Basavaraj Patil: draft is in last call, march 21st is last day of LC Jari Arkko: you're saying draft specifies by default you use regular, but by configuraiton you use something else, is that? Basavaraj Patil: there's a 2ndary mgmt connection that can be established, which is not the regular, can be used. Jari Arkko: how do you get there?, what code do I need to have in my host? Or is it a configuration issue? Basavaraj Patil: you'll need support from MAC layer as well. Daniel Park: clarification on MTU recommending 1500. Any different MTU issues among 16, 16d, 16e? Basavaraj Patil: default value 1500 should be good for all 16x, minimum default in rfc2460 applies. Air interface link has capability to support, but backhaul links can be 1500bytes or less. Appendix has value lowerthan 1500. MTU in WiMAX is 1400 Max Riegel: 16g specification for GPCS, GPCS hasn't classification spec'ed, just a pipe. 802.16-2004 is specifying classification rules Max Riegel: for IPv4 then only IPv4 classifiers, there's no classification defined for GPCS. Suresh Krishnan: inter-RA, the host must take it. if the MTU in RA is smaller, then host must take it - makes not sense, because Host does MTU. Basavaraj Patil: needs to be a must Suresh Krishnan: not talk about PMTU, because that's a PATH Basavaraj Patil: right Suresh Krishnan: default value is 1500. you can also use RA to put MTU greater than 1500. Basavaraj Patil: default value you can always use greater than that Dava Thaler: is there any assumptions between transport connection and what IP sees? Can you have an IPv4 transport connection and an IPv6 connection and they both go to same transport interface? Basavaraj Patil: no, classifiers are the ones. Dave Thaler: it is not legal to a host to use both CS, with different transport connections and treat them with same interface is that correct?. Compare to PDP, negotiate with IPCP... PDP has multi-link... is that legal or illegal in this context? Basavaraj Patil: yes... what would have happened is that there are two transport connections. Dave Thaler: if one station chosses IPCS and the other station chooses the other CS - can they talk to each other?. Without that kind of explanation this text makes no sense. better not go there. Max Riegel: a BS is not a link, it's a concentration of hundreds of links Marco Liebsch: why idel mode support justifies such a huge interval? Basavaraj Patil: in order to deliver RA you need to wake up the mobile, the mobile has to then come up establish bearer. Marco Liebsch: setting huge value so that wait, why is there a requirement to deliver RA to idle mode? Basavaraj Patil: violating the IP stack Daniel Park: mailing list discussion for details, LC ends March 31th. [Friday Session] o #2 Timeslot 2.5 hours o 0900-1130 Friday March 23 o Morning Session I, Congress I 10 Administrative Chairs - Blue sheets - Jabber scribe - Agenda bashing - WG report 20 16ng problem statement Daniel Park (No Presentation) (on behalf of Junghoon Jee) - ID: draft-ietf-16ng-ps-goals-01 - Status: Active WG Document - GOAL: Ready for IESG Daniel Park: 01 is now available and wait for reviewers acknowledgements, Raise your comments on that quickly prior to moving it forward. It will be ready for IESG soon 30 IP over EthernetCS Max Riegel Hongseok Jeon and Sangjin Jeong - ID: draft-ietf-16ng-ip-over-ethernet-over-802.16-01 - Status: Active WG Document - GOAL: Ready for WGLC Daniel Park: whats going on on in terms in MBS in WiMax - what's going on there? Max Riegel: WiMax starting to address MBS. part of release 1.5, WiMax Foirum NWG has a team starting trying to define what kind of MBS, sublayer, just a couple of people wking together Daniel Park: probably couple meeting back, several 16ng worried about MBS, regarding mcast/bcast service over 16ng, you wanna take and this and include in this doc?, any relationship ETHCS and IPCS?. IPCS doc is not taking care of MBS and this sort in this doc, but ETHCS has MBS, but both doc obviously target proposed std in IETF. From WG perspective trying to clarify between ICPS and ETHCS, or do we have to ammend IPCS and ETHCS in connection to MBS? Max Riegel: timing. MBS just started and so IPCS is not able to address MBS in detail because no reference. For IPETHCS we're later, we may be able to address MBS service, but we have possiblility to provide provisioning, how things relate to each other Max Riegel: for MBS you must install a sort of multicast service flow management, maybe related to link model, but we have to see how design ed in spec. for current approach we not use MBS, even if designed, not sense, if other group is doing (market-relevant), but good idea, good hint, maybe even in annex Daniel Park: IPCS and ETHCS and consistency between those Max Riegel: we have to clarify in 16ng we are addressing it, IPCS is done, not reference Gorry F.: ARP must not be implemented, but must be turned off? what protocol rec? ND no relay, but ARP relay? why? Max Riegel: filtering out some, not a for bcast, that's why ARP relay to filter bcast. thats why Dave Thaler: mld snooping switch, and since nd is multicast. what not have igmp that works snooping for bcast. not option for igmp proxy there Bernard Aboba: other potential alternative: dynamic VLANs Max Riegel: Ive seen, but this RFC is a nice spec, same issues in wireless networks. mobile terminals arent handled in spec haven't seen support for enterprise lan scenario, and IPv6 support not completed, but thinks most difficult thing we have to cope in spec, 4562 is informational, no normative reference. little bit tricky to understand all details in there, missing explanatiopn of reasoning. Jari Arkko: not impossible, but most of time there's a reason why informational or experimental, not a particular grasp of this draft here, nit will make. Dave Thaler: 4562 is informational, is it independent or an IETF, not IETF consensus Dave Thaler: MLD snooping is informational Gorry F.: point to 4605 which is standard track Bernard Aboba: there are people deploying this, is this doc in sync with networks already operating? Max Riegel: believes there are people who are already deploying IP over Eth 802.16, these are in discussion, working, getting comments 30 IPv4 over IPv4CS Daniel Park (on behalf of Syam Madanapalli) - ID: draft-madanapalli-16ng-over-802-dot-16-ipcs-00 - Status: Candidate WG Document - GOAL: WG Adoption Dave Thaler: given this one is not Ethernet, why is default ethernet 2038 and not...the only req is 1500 or larger, but you choose lower - why? Bernard Aboba: because the draft looks like Ethernet, but this is not Ethernet, right? Daniel Park: not any MTU problem by using the IPv4 CS Jar Arkko: will be simple to just have one value. also in the other v6 doc, theres discussion backend archi architecture influencing how influences Dave Thaler: 1500 best value to avoid frag? or just picked up in the air? Jari Arkko: seems like wrong value because Eth frag Dave Thaler: maybe less than 1500, as long as above 1280 fine. 1484 or something Bernard Aboba: if Ethernet exposes an Ethernet interface then DNA triggered. then DHCP, so ARP is sent, then you figure out what to do Bernard Aboba: problem dropping ARPs - you wont get an address if you do that, because in DNA you don't dhcp. no connectivity if dropping the arp. In any operating system you'll have no address Dave Thaler: respond to any MAC address, sounds as if what you're proposing, manufcature ARP response... ARP goes on wire Bernard Aboba: not get DHCP but get (MAC) address Behcet Sarikaya: assigning an IP address, here the recommendation was to recommend a private address, with subnet mask unique for each mobile? because no prefix, havent read 3021, subnet mask is one thats used. subnet mask unique for each mobile node. 10 Using DHCPv6 and AAA Server for Mobile Station Prefix Delegation Behcet Sarikaya - draft-sarikaya-16ng-prefix-delegation-01.txt - Status: Individual WG document - GOAL: Initial discussion Dave Thaler: the sequence diagram, was the normal DHCP-PD stuff, right? that talks between AR and DHCP server with DHCP, but another slide AAA alternative AAA - DHCP? what relationship? Behcet Sarikaya: alternative, prefix delegation is still dhcp prefix delegation either way Dave Thaler: two different ways of doing same thing, which one is mandatory? Behcet Sarikaya: not recommending anything, DHCP is better Dave Thaler: but BA that disadvantage might not be there Jonne Soininen: is this the normal procedure, is there something special related to .16? Behcet Sarikaya: no new option, comes up in this context, AR can use it in this approach Jari Arkko: documenting the existing ways to do this, doesn tie to 16ng in any particular way, not in Charter 05 Closing Chairs