TRILL Working Group meeting, 22 March 2010 15:20 to 17:20, California B Room Chairs: Erik Nordmark, Donald Eastlake 3rd Minutes by Radia Perlman, edited by the Chairs Donald Eastlake opened the meeting with the usual "Note Well" slide, distributed the blue sheets for people to sign, and then arm-twisted to find a note taker, at which point everyone in the room carefully examined their shoes until there was a reluctant volunteer. Arm-twisting to find a jabber scribe failed. The posted agenda, below, was reviewws and there were no comments or requests for changes. AGEDNA 5 min. Administrativia (scribes etc), Agenda Bashing, Chairs 15 min. Review of Existing Milestones, Document Status, and Non-IETF Code Point Allocations, Chairs 25 min. Definitions of Managed Objects for RBridges Anil Rijhsinghani, Kate Zebrose draft-ietf-trill-rbridge-mib-00.txt 15 min. RBridges: VLAN Mapping Radia Perlman, Dinesh Dutt, Donald Eastlake draft-ietf-trill-rbridge-vlan-mapping-02 15 min. Additional Documents, Donald Eastlake 3rd 15 min. Charter Update, Chairs REVIEW OF EXISTING MILESTONES Donald reviewed all the completed milestones, most prominently that the Problem and Applicability Statement was long ago published as RFC 5556, the Base Protocol Specification was approved on March 15 by the IESG for publication as a Proposed Standard, and the first draft of an RBridge MIB has been posted. Unusually for an IETF WG, TRILL has no overdue milestones! The only remaining unfinished milestone in its current Charter is to re-charter or shut down the WG by July 2010. REVIEW OF NON-IETF CODE POINT ALLOCATIONS Donald noted that TRILL needs a NLPID and an Ethertype and a block of multicast addresses. In addition L2-IS-IS needs an Ethertype. 0xC0 has been allocated as the NLPID for TRILL. This value is used in the IS-IS Protocols Supported TLV. The Ethertype and multicast address applications have been submitted to the IEEE Registration Authority (RA), requesting that any fees be waived as these codes points are for standards use. The IEEE RA is run by the RA Committee (RAC) which is currently chairless; however, it is believed that allocation decisions are made by a expert working under contract to the IEEE and should not be affected by the chairless state of the RAC. REVIEW OF DOCUMENT STATUS As above, the Problem and Applicability Statement has been published as RFC 5556 and the Base Protocol Specification, draft-ietf-trill-rbridge-protocol-16, was approved as a Proposed Standard on March 15, 2010. Three more WG documents are in process; VLAN mapping, MIB, and header options. Donald asserted that the VLAN mapping document was in quite good shape and asked if there was any objection to Working Group Last Calling it. There was no objection and this will be brought up on the mailing list for confirmation. Relevant non-TRILL WG documents are IS-IS extensions for layer 2 (draft-ietf-isis-layer2-03), which the TRILL base protocol spec is normatively dependent on, and a document in the ppp extensions (PPPEXT) working group for TRILL over PPP (draft-ietf-pppext-trill-protocol-00). Both of these are in fairly food shape and people are encouraged to review them and comment. RECESS By unanimous consent, the WG recessed for ten minutes to celebrate the approval of the Base Protocol Specification. There was the (perhaps new tradition) sharing of an excellent two-layer cake. [In discussion after the WG meeting, it was claimed that perhaps is was a Layer 2 Cake rather than a two-layer cake.] MIB: DEFINITIONS OF MANAGED OBJECTS FOR RBRIDGES Anil Rijhsinghani gave a detailed overview of the MIB document, with a few clarifying questions about alerts, whether certain objects should be read-only, where they should be placed in the tree. Anil thanked Kate Zebrose, Radia Perlman, Anoop Ghanwani, and Donald Eastlake for their work and help on the MIB. Since a station might be a co-located router and rbridge, both using IS-IS, there is a question of how to distinguish the layer-2 and layer-3 IS-IS MIBs. Anil suggested that the SNMPv3 Context facility was the way to go and that draft-kkoushik-snmp-context-map-mib-02.txt was a good exposition of this facility. Dinesh Dutt and Anoop clarified that the FDB table should be indexed by {MAC address, VLAN ID} and not just MAC address. Donald suggested that inhibition should be per port. A question arose as to wheher there is an IGMP snooping MIB. Reportedly Cisco has one but it may be proprietary. The WG was urged to read the document and take questions to the mailing list. VLAN MAPPING AND TRILL Radia Perlman gave an overview of the VLAN mapping document. There was a question about whether it would extend to more than 2 regions (which it does). Another question about whether VLAN mappings are reversible, and Radia said they would only be if the mappings were 1-1. FUTURE TRILL AND TRILL RELATED WORK Then Donald talked about potential future work items: We should undertake additional work that is important/urgent. But perhaps the most important factor is whether there is someone willing to work on a document. On data center bridging, there were some clarifying questions, such as from Tina Tsou, of why TRILL would need a data center bridging spec, or why TRILL would need to change to support IEEE bridge features of priority based flow control, enhanced transmission selection, and congestion notification. The answer was that probably the first two of these features could be implemented without needing to change TRILL but Congestion Notification needs TRILL work to handle cases like a Congestion Notification generated by a bridge between two RBridges in response to a TRILL data frame. In any case, each should be thought about in the context of TRILL to ensure there are no subtle issues. John Messenger pointed out that there are also two new recent work items in the 802.1 Data Center Bridging task force: 802.1Qbg (Edge Virtual Bridging) and 802.1Qbh (Port Expander). Donald mentioned the notion of provider RBridges, and several people said that this duplicated work being done in IEEE 802.1. Donald pointed out that provider RBridges is a possibility that he routinely mentions [first instance appears to be here from July 2007 http://www.ietf.org/proceedings/69/slides/trill-4/sld2.htm], that there have never been any IETF drafts in this area, and that it is not mentioned in the proposed new Charter to be discussed later in the meeting. Additional people spoke suggesting that before work was started in this area, it be coordinated with 802.1 OAM-BFD for TRILL was discussed. Dinesh Dutt asked why we would need BFD for TRILL. Donald said for faster link failure detection. Vishwas answered that some L2 support is required. Tina Tsou mentioned that a draft of BFD for IS-IS already exists. Himanshu Shah asked why not CFM (Continuity Fault Management, 802.11ag) instead of BFD? Donald answered is that BFD is normally used in conjunction with IS-IS and that TRILL's multipathing creates problems for using CFM. Dave Allen mentioned that if we were to define a traceroute type OAM facility, it is important that exactly the same code path be followed for traceroute packets as for data packets, including ECMP capabilities, in order for the traceroute information to be useful. Dinesh pointed out that IP traceroute does not meet that criterion but people still think it is useful. PROPOSED TRILL CHARTER UPDATE Erik Nordmark then presented a proposed new TRILL charter and set of milestones, and said that suggestions for new work and comments on the draft new charter should be brought up on the mailing list.