OPS-NM Open Area Meeting Minutes
March 20, 2001
Minutes by Andy Bierman and David Harrington
1) A-D status summary of OPS-NM area WGs (Bert Wijnen & Randy Bush)
2) IPv6 MIB Design Team Report (David Thaler)
3) Reusable RowPointer TC Proposal (Bob Moore)
4) Open Q&A
1) WG Status Summary
Randy Bush presented a quick summary of the OPS WGs:
AAA - proceeding along okay; recently held an interim meeting. Major areas that have been worked on are transport profiles, security, and data modeling. Transport has settled down. Security and data modeling are still being discussed.
BMWG - very productive. Focuses on devices, not systems. IPPM tests systems. Starting work on routing parameters.
DNSOP - working on experiment in distributed DNS using AnyCast. Will start working on v6 DNS transition requirements, including decisions affecting v6 (should different types of records be kept?).
GRIP - draft on security aspects of ISP behavior approved by IESG recently. The WG will look into logging requirements this summer.
MBONED - how to build up mcast infrastructure; currently documenting existing practice and improving diagnostic support.
NASREQ - requirements for AAA; WG currently dormant.
NGTRANS - overview and how to transition. Currently examining how to find tunnels on the WAN and how an Ipv6 devices gets out of the tunnel onto the WAN; Also working on integration of IPv6 with DNS.
MULTI6 - multi-homing causing increase in small BGP announcements; Schemes for IPv6 address assignment for multi-homing are being considered to solve this problem; Trying to avoid announcements of small aggregates.
PTOMAINE - address space announcement consumption. Studying the patterns looking for possibilities for aggregation. Assuming changes to BGP and router code are in scope.
Bert Wijnen presented a quick summary of the NM WGs:
ADSLMIB - cruising along; Requirements work (from DSL Forum) is done already, and ADSL extension mib is in AD review. The ADSL2 MIB is in WG last call, and the VDSL mib in in progress.
BRIDGE - Keith McCloghrie on sabbatical, so Les Bell is taking over as WG Chair. WG needs to advance documents, but no new work is progress.
ENTMIB - Keith on sabbatical; RFC 2737 ready for advancement to draft standard. No new work, but new chair needed to work on advancement. [soon after the IETF, Margaret Wasserman was appointed as WG Chair.]
HUBMIB - completed all work items, and now looking to advance documents.
New IEEE MIB work starting; Dan Romascanu acting as liaison for IETF to IEEE in order to get access to IEEE documents; cannot redistribute them, but WG members can view documents to verify MIBs.
RMONMIB - several pieces of work completed. The WG wants to do SSPM - configuration of synthetic sources for performance management. High capacity RMON mibs are completed, and the slightly modified RMON and RMON2 modules will be updated with a new publication process. The HC-RMON document will describe the enumeration objects that were extended, and the updated MIB modules will be stored at IANA or some other suitable site.
Infiniband MIB BOF - IP over Infiniband device; also need another MIB to manage Infiniband network fabric. Trying to have 1 WG instead of 2 to do this work.
SNMPCONF - milestones are overdue. Good interaction with Diffserv. Original design of Diffserv MIB changed, which improved diffserv. We might consider modifying mibs to better meet the needs of policy mgmt. BCP needs more input, and the document may be helpful to correct problems in some existing mibs. Further discussion to ensue.
DISMAN - new work on alarm mibs in progress, which was started in the ATOMMIB WG, but moved to this WG to create a generalized (instead of ATM-specific) solution. New versions of the Script MIB and Schedule MIB will recycle at proposed.
EOS - WG will address the high priority SNMP items that have been postponed due to SNMPv3 work, such as OID compression, efficient set operations; the WG will consider new PDU types; want to improve performance of SNMP, but will not make changes to RFC 2571 architecture.
SMING - Chartered to merge SMIv2 and SPPI (COPS-PR) and add OO-based concepts to the SMI. Trying to create one data modeling language in the IETF for all mgmt info; need to use pragmatic approach and contain scope of work.
SNMPv3 - WG limited in scope to advancement of SNMPv3 RFCs to full standard.
There have been a number of requests for new SNMPv3 security approaches.
Documents cannot get past IESG without security. We want to have WGs define additional security mechanisms, but we have been pushing back in SNMPv3 and RAP. ADs have discussed with Security Ads, trying to get security extensions done in the security area. Security A-Ds should identify which security protocols to support; also need to discuss who should do the mapping.
POLICY - 1 RFC after 3 years; now doing PCIM extensions; mapping to
schema and QDEN moving along, and near WG last call.
RAP - COPS-PR RFC will be out any day; SPPI was in IETF Last Call, ready to be RFC, when SMING WG started; the Framework MIB needs to move to PS; The WG chair has changed, and new charter items are under discussion, such as COPS over TLS and COPS over CLS;
At this point, a discussion about too many security protocols occurred;
The following issues were raised:
- some do not want too many security transports on standards track;
- where should be done, OPS or Security area?
- distinction needed between defining new security protocols and defining mappings for a particular application over a particular (existing) security protocol.
AGENTX - This WG is planning to deactivate.
2) IPv6 MIB Design Team Report
The design team is chartered to look at IPv4-centric MIBs and determine how to transition them to use the InetAddress MIB.
Recently, new TCs were added to the INET-ADDRESS-MIB module.
The following issues were raised for particular MIBs:
- UDP-MIB: add connection table, some counters.
- TCP-MIB: add tcpListener table, more counters
UDP and TCP expressed connections different ways; tried to align.
- IP-FORWARDING MIB: what do modern ribs look like?
- IP-MIB: want to get rid of ipv6InterfaceIndex
Ipv6IfTable - need help understanding this table; more counters;
The design team is currently seeking implementation feedback from implementors of these MIBs. Send input to OPS area 'mibs' mailing list.
Refer to the following URL for more information:
3) Reusable Row TC Proposal
The need for a new type of RowPointer TC grew out of work in the snmpconf, diffserv, and policy WGs. The DS MIB separates data path elements (meters, markers, etc.) from the tables which contain the parameters on how these DP elements will operate. A template is the combination of data path elements; parameters are outside the template; multiple copies of data elements exist, but only one copy of the parameters may exist.
Duplicating the templates is a problem because it is unclear whether pointers are deep or shallow, and making multiple copies of parameters makes an implementation much harder to maintain. Row pointers may point to the same row, or to a copy of a row - it is important to know which.
The current proposal is the I-D draft-ietf-ops-rowpointer-tc-01.txt.
4) Q & A Session
The following issues were raised during this part of the meeting:
- IEEE vs IETF process differences
- criteria for starting new work is stricter in IEEE
- what is the A-Ds' vision of where OPS-area is going?
- COPS-PR and SNMP protocols (should they merge?)
No: letting parallel paths proceed when there is no clear path
- what's wrong with SNMP; why CLI is more prevalent
- agent implementation vary in quality; never perfect
- agent design getting too complex
- intentionally no certification process in IETF
- example of security area: setup industry council to verify implementations; SNMP vendors could do the same thing
- emphasis that not only large ISPs need net mgmt tools
- rigorous validation of net-mgmt is not done, as it is for security.
- net-mgmt is not as interesting or important to operators as other new features
- call for interest in new security work
- OID compression & OID suppression
- request for comments from operators on what the typical SNMP traffic mix from NOC to agents
- support for overall vision of OPS area and design team to form and guide new work
- suggestion to have these meetings every IETF, not every other, but skip the WG status summaries