Internet Architecture Board


IAB Minutes 2009-01-14

Home»Documents»Minutes»Minutes 2009»IAB Minutes 2009-01-14

IAB Teleconference


1. Roll-call, agenda-bash, approval of minutes, administrivia

1.1. Agenda

1.2. Attendance

  Loa Anderson (IAB Liaison to the IESG)
  Gonzalo Camarillo
  Stuart Cheshire
  Lars Eggert (IESG Liaison to the IAB)
  Aaron Falk (IRTF Chair)
  Russ Housley (IETF Chair)
  Olaf Kolkman (IAB Chair)
  Gregory Lebovitz
  Barry Leiba
  Andy Malis
  David Oran
  Danny McPherson
  Dow Street (IAB Executive Director)
  Dave Thaler
  Sandy Ginoza (RFC Editor Liaison)
  Kurtis Lindqvist
  Lynn St. Amour (ISOC Liaison)
  Lixia Zhang

1.3. IETF 74 Plenary Planning

There was a short discussion of the planning for IETF 74 technical plenary. Loa and Andy will be attending the MPLS-TP meeting next week, and will use the opportunity to introduce the plenary idea to potential speakers. All other IAB members were requested to review the IETF 74 plenary material on the IAB wiki (plan, goals, etc).

The group will revisit this item during the 1/21 business meeting.

2. JFC Appeal

After substantial discussion of the points raised in the appeal, the group decided to uphold the decision of the IESG. It was agreed that a few of the points raised by JFC warranted IAB clarification; Barry will write a draft response and distribute it for review by the IAB.

Note: Liaisons dropped from the call for this agenda item, but then re-joined for the NAT66 technical discussion. Russ also recused himself from this topic due to his IETF Chair role.

3. NAT66

The bulk of the meeting was devoted to continuing the technical discussion of NAT66. Dave Thaler introduced the topic, and proposed three questions to guide the discussion:

a. What are the issues that v6-v6 translation might address?
b. What, if anything, does the IAB want to say about these issues? c. How does this guidance, and the use of NAT66, relate to the UNSAF considerations?

The first issue considered was host/network re-numbering. Lixia had suggested previously that the re-numbering problem could be resolved by giving everyone provider-independent address space. (i.e. move the problem into routing). Andy noted that even if such an approach could work, standardization and deployment would take time and would therefore not address the immediate needs motivating NAT66. Gregory stated that re-numbering was a valid issue, and that small businesses and home users might not have easy access to PI space. Referencing Andy’s comment, Dave Oran added that NAT looks “deceptively” simple as compared to the RRG proposals, and that the IAB should not pre-judge other solutions in this space.

Stuart raised the point that NATs often prevent inbound connections, and that this difficulty is often given less weight than the re-numbering part. Dave Thaler noted that this is only the case when there is address sharing, and that much of the NAT66 discussion has focused on 1:1 translation. Stuart then described a typical small-network scenario where all devices are addressed via DHCP. Dow observed that there are two main solutions, PI space and DHCP, and two primary use cases, large networks and small/home networks. Dave Thaler noted that, in his understanding, there are a sufficient number of cases that currently fall into the gap between these. Stuart suggested that the IAB document could simply describe the trade-off: “NAT could make re-numbering less onerous, but on all the other days where you are not re-numbering you have the other problems that result from NATs.”

The discussion moved to various approaches to name translation, and Dave Thaler used HIP as an example (though HIP is used for a somewhat different purpose). Starting from the HIP reference, Dow asked how many of the addresses of interest in the “avoid renumbering” situation were actually being used more in the context of an identifier than a locator, and whether this was a use that should be deprecated anyway. Dave responded that there were a small number of systems that fall into this category, such as inventory systems, configuration management systems, etc., but that some of it was also topology related or involved routing. Dow noted a possible trend (enabled by link-local addresses) to move away from global addresses for the network infrastructure as well.

Dave Thaler observed that there seemed to be IAB consensus to say something on this topic – that re-numbering needs work, but that we don’t want to constrain the solution space. Gregory added that the community was seeking concrete direction. The group discussed the properties of 1:1 translation, and whether guidance on this part of NAT66 would be appropriate.

The board then considered the issue of topology hiding and host counting, and debated whether these were real or perceived benefits of NAT. Host counting is related to whether hosts share an externally observable address, leading the group to discuss IP address scarcity as a driver for IPv4 NATs, and how v6 NATs differ in this regard. There was general agreement that address scarcity is not a primary driver for IPv6 NATs, even though some business processes today might not easily support allocation of subnets (instead of individual addresses) to end customers.

Gregory described that an apparent goal of host hiding was to obscure the role of individual hosts, such as a finance server, that might be a target for DOS. Gregory and Dave Thaler will work offline to better describe this issue. Dave will make some minor edits to the document and then submit it as a -00 draft. The IAB will continue the NAT66 discussion at the 1/28 meeting.