Minutes HUBMIB (Ethernet Interfaces and Hub MIB) Working Group Recorded by Steve Moulton 50th IETF Minneapolis, Minnesota, USA Thursday, March 22, 2001, 3:30 PM Welcome by Dan Romascanu. Agenda Bashing. No changes from published agenda. Roughly 18 Present. In general information available from the slides will not be reproduced here. Agenda: . Introduction, agenda bashing . Status of the Working Group . Status of the 802.3 Working Group . 10Gbs Ethernet Management 802.3 ae . Power Ethernet Management: 802.3 af . RFC 2665 Editorial Issues . Discussions, working group actions, new charter (if relevant). Administrivia: . Working Group Chair - dromasca@avaya.com . AD and Advisor - bwijnen@lucent.com . New working group mailing list list - hubmib@ietf.org . To subscribe - send "subscribe " to hubmib-request@ietf.org. . Archive: http://www.ietf.org/mail-archive/working-groups/hubmib/current/maillist . IEEE 802.3 WG Web page: http://grouper.ieee.org/groups/802/3/index.html Liaison with the IEEE 802.3 WG The liaison function was set up by Scott Bradner. Dan Romascanu is the liaison person. The intent of this was to make 802.3 private documents available for those who are actually writing the MIB documents without having to attend IEEE meetings. HUBMIB Documents Status. RFC2108 Hub MIB - leave alone? RFC2668 MAU MIB - MAU MIB RFC2665 - Ethernet-like Interfaces MIB RFC2666 - ChipSets for Ethernet-like interfaces MIB (Informational). All documents save RFC2666 are at proposed standard status. In the mind of the working group chair, RFC2668 and RFC2665 are living documents. There is potential work in tracking changes in the Ethernet world. Should implementation experience for RFC2108 be gathered to forward it to draft status? While no consensus was established in the room, there was a frequently expressed opinion that since little new repeater implementation is taking place, there is little value in advancing this document. The question is referred to the area director for his opinion. A discussion took place on the relevant merits of having this document status changed to informational or historic. Since the MIB is in fairly widespread use, there was support for moving it to informational if IETF procedures would allow it. Referred to the area director. The room consensus is that implementation reports for RFC2665 and RFC2668 should be gathered with the view of advancing these documents to draft status. An open question is whether these documents should be modified to support 10gb interfaces and therefore spun at proposed, or if they should be advanced to draft as is. What is happening in the IEEE 802.3 Working Group Nowadays? There are three areas of activity: . 10 Gpbs Ethernet - 802.3ae . Power Ethernet - 802.3af . Ethernet First Mile - EFM The number of participants in IEEE 802.3 in the past 10 months has doubled with increased change and interest in Ethernet. The IEEE 802.3ae Charter is to define 802.3 MAC parameters and minimal augmentation of its operation physical layer characteristics, and management parameters for Ethernet 10 Gb/s using full duplex operation. Note that this takes out repeaters from considerations, since they run at half duplex. In addition to the traditional LAN space, we should consider implications of data rates compatible with OC-192c and SDH VC-4-64c payload rate. This has implications at layer 5, and has implications wrt the SONET MIB working group. We probably need to understand how we can work together and reuse the models that have been defined there. The IEEE 802.3af Charter is to define ways to provide power to Ethernet devices via the network connection [see slides for details]. Compatibility with existing equipment is an issue. Both working groups belong to the 802.3 committee. Defining management from their point of view is providing a set of managed objects, all hardware related, and providing a set of registers and considering other hardware implications necessary to implement these objects. The 802.3 committee explicitly leaves the MIB work to the IETF. The statement was made that there are probably 10 or 20 people in the IEEE that fully understand SNMP-based management. EFM (Ethernet First Mile) Proposed Charter. This working group was changed from Ethernet Last Mile, due to "last mile == end" connotation. The intent is to expand Ethernet applications to subscriber access networks, over copper (Ethernet over VDSL), fiber, and PON. This group applied for charter last week in the IEEE. This work is not yet approved - we need to be aware of it, but not to approach the work yet. On this round of work, we would like to be fully in sync with IEEE, so that when the IEEE standard is being released, we can produce the MIBs without a year gap. Presentation: Ethernet MIB Updates for 10 Gb - John Flick. Please see the slide presentation for the bulk of the material (IETF50EtherMib.ppt). 802.8ae adds three new objects associated with rate control. It is not clear that the Ethernet MIB needs to reflect these object, particularly since aStretchRatio is a read-write integer that can only have the value 104. Is the equipment going to be able to determine these things on its own, or it is going to be set at provisioning time? We need to send a question back to them to clarify these objects. Everyone on the mailing list is welcome to comment. Updates to existing definitions - dot3StatsAlignmentErrors and dot3StatsSymbolErorrs have editorial changes. Do we need 64-bit error counters? Some objects can now roll over in less than one hour. Of course, if these counters are rolling in less than one hour, you probably have a problem anyway. Do we need to reopen the interface MIB? Or do we define an extension for ifInErrors? ifTstTabletext and test OIDs have nothing to point to anymore. The table died and its replacement table died. Do we want to define our own test methods? I.e., do we want a way to do a loopback test on an interface via SNMP? This is an open issue. We need to fix ifInErrors definition since it currently double-counts symbol errors, since every symbol error causes a FCS error. We need to update ifMtu mapping to reflect implementations that support greater than 1500. 802.3 does not support MTU > 1500, but many vendors do. There is a section in rfc2665 that says Ethernet has no layers in ifStack table. There is an editor's note from six years ago that that might not always be the case. It is certainly not the case now. So this working needs to be fixed. We need to explain how to use the stack interface. There was such a long lag between IEEE implementing 100mbit and this WG dealing with it that many vendors registered their own types. The document should say that should you run into the old vendor specific ifType values, this is what you SHOULD do when implementing an agent. The comment was made from the floor that in general we should tighten up this text with respect to MUST/SHOULD. Action Item: Get procedure from IANA for fixing the ifTypes mess. The interface MIB does request that IANA deprecate these old ifType values. Presentation: MAU MIB Updates - John Flick. Slides are in IETF50MauMib.ppt. There should be MUST text indicating the relationship between MAU MIB and Ethernet MIB and that the implementations must be kept consistent. The rest of the changes are covered in the slides. Note that this new technology does not use repeaters, so the underlying tables need not be realized in this document. Presentation: WAN Interface Sublayer Management - John Flick. Slides are in IETF50WisMib.ppt. The edge objects defined by SONET do not match up with the EtherNet MIB definitions. The SONET MIB has 15 minute Ethernet interval counters, rather than the traditional SNMP free running counters. The story was told of the old telecom practice of taking pictures of counters on phone switches every fifteen minutes (literal snapshots), and how the current practice echoes this historical practice. The SONET MIB has a table that has values from the current interval, and a 15 minute table. You can derive the counters from the interval tables. If we do this work, we should do it in cooperation with the SONET guys. The co-chairs of the AToMMIB working group have asked for this issue to be crossposted. We have a volunteer (Tom Rutt) to contribute an analysis. [Note that the AToMMIB working group handles ATM, SONET, and Optical (DWDM) MIBs] The Power Ethernet MIB (draft-romascanu-hubmib-power-ethernet-mib-00.txt). This document is based on 802.3af. See the main meeting slides (ethernet_mib_wg.ppt) for details. We have several volunteers for document co-editors. It is nice that there is no legacy MIBs or other curse-of-the-installed-base issues. New Charter Proposal. Please see the slides for details. The changes are primarily with respect to the new work already discussed. Note that repeaters are removed from the charter proposal. We expect no new implementations of the current MIB documents, so it is not clear whether or not these documents should move forward. The area director (Bert Wijnen) indicated that this is a good thing to consider, but that the pressure to do so is not high at the moment. There was a discussion of the informational versus historic status. The reason that people are not implementing RFC2108 is that that companies aren't implementing hubs so much anymore, in favor of switches. Bert will research having RFC2108 going informational rather than historic. The chair's opinion is that having this document go to draft standard is not terribly important, but vendors may think differently. Please submit implementation reports. Goals and milestones. A final deadline of 2 Jan 2002 for the new documents is proposed. Volunteers are Tom Rutt Dan Romascanu K.C. Norseth John Flick Avi Berger - specific interest in 802.3af Mike Ayers - specific interest in 802.3ae End of meeting.