1. Roll-call, agenda-bash, approval of minutes, administrivia
Loa Anderson (IAB Liaison to the IESG)
Marcelo Bagnulo (incoming IAB)
Lars Eggert (IESG Liaison to the IAB)
Russ Housley (IETF Chair)
Olaf Kolkman (IAB Chair)
Jon Peterson (incoming IAB)
Lynn St. Amour (ISOC Liaison)
Dow Street (IAB Executive Director)
Aaron Falk (IRTF Chair)
Vijay Gill (incoming IAB)
Sandy Ginoza (RFC Editor Liaison)
John Klensin (incoming IAB)
Lynn St. Amour (ISOC Liaison)
The IAB used webex during this call for both voice and screen sharing. After initial setup, everything worked well and voice quality was generally good. This was the first time using webex as the audio bridge for an IAB meeting. The usual IAB jabber room was used instead of the webex chat function.
Since the IESG slate was received from the NOMCOM this morning, the board decided to schedule the last 30 min of the call as an Executive Session.
There was a short discussion about the date and location of the IAB retreat. The dates of 27-28 April were selected. The location will be either Ann Arbor, MI or Ashburn, VA.
Loa noted that the IAB will need to select a new liaison to the IESG, given that his term is ending in March. He further suggested that some overlap would be of benefit, and gave a brief description of what the role involves.
1.4. Meeting Minutes
The board approved final minutes for the following meetings:
Dow will post the minutes to IAB web site.
2. NAT66 Techchat
The group entered into a long discussion on the NAT66 topic. Dave Thaler began by summarizing the discussion so far, to give the new IAB members context. Marcelo remarked that it would be good to understand the differences between NAT for IPv4 and NAT66. Lixia noted that some remarks could be added to the current draft, but that a full comparison would be better as a separate document. She also thanked Marcelo for his comments on multi-homing, and will expand on that topic in the next version.
There was a discussion of 1:1 vs 1:N translation, and Dave noted that the current draft does not call out that aspect of the design space very directly, but that it is something of a sub-case under translation. Stuart returned to the question of “making NATs transparent in the architecture”, arguing that guidance of that sort, without specific details on how to accomplish it, would be less beneficial. It was thought that most IPv4 NAT designers also probably attempted to make their devices as transparent as they could. This turned the conversation back to the perceived need for NATs, and several members argued that it was actually “ease of management” rather than a strict “shortage of addresses” that led to early single-address-per-customer business models at ISPs. Dave Oran noted that IPv4 had no easy way to do prefix delegation. Lixia pointed out that, for the most part, IPv4 NAT was not “designed” (at least as a part of the architecture), but rather that it showed up and the Internet accepted it.
There was a short discussion of techniques for “requested” transparency, A+P, and the recent SHARA concepts. The conversation then returned to whether or not the document should elaborate on 1:1 vs 1:N schemes. Dave Oran added that there are other aspects of the mapping function that might be worth talking about, such as whether the transformation algorithm is secure. The group then considered the use of PI space as a solution, and Olaf suggested that some of this is related to policy guidance for RIRs. This led back to a discussion of business models for ISPs, and Lixia pointed out that since many sites multi-home, even the use of PA space does not solve the routing scalability problem. Dow attempted to summarize this tradeoff as one with a cost at each end of the range: NAT has a cost in interoperability and reachability, while PI and/or multihoming without NAT has a cost in the global routing state. Olaf inquired about the transfer of costs between parties. For example, the current text mentions that sites *not* deploying NATs should be unaffected by sites that choose to. Olaf asked if a similar statement could be made about the use of PI space – namely, should sites who choose to solve their problem through routing (vice NATs) be required not to incur costs on other sites? Dow suggested that the cost transfer is somewhat different in these two cases: with NAT it is often from one end user/site to the other, while with routing it is from one end user/site to the commons of the routing infrastructure. Lixia further clarified that the issues of transparency to the users and the cost to the network could largely be separated. Stuart added that NATs also add cost and complexity to software development, since applications now need to be able to deal with NATs.
Lixia felt that in general the IAB should steer clear of policy guidance, but in this case the there is difficulty because the availability of PI space has such a close tie to the architectural model. The discussion moved to multi-homing via routing vs. NATs. Andy described this tradeoff in terms of monetary cost; you can get similar, but not identical, functionality from the two approaches. Multihoming via BGP provides more power, but it is also more complex and is generally a more expensive offering from an ISP.
Dave Thaler asked the group whether or not to remove the section on host counting. The group considered whether it was part of the IAB role to “actively dispel myths” about perceived benefits. It was not clear exactly what level of customer interest there is in preventing host counting. The discussion closed with the key question of “whether to make translation a permanent block in the architecture?” The group will consider this question, and members were requested to send additional text to Dave Thaler and Lixia capturing the meeting discussion.
At this point in the meeting liaisons and incoming (new) IAB members dropped from the call.
3. Confirmation of the NOMCOM IESG Slate
After discussion, the IAB unanimously confirmed the IESG slate. The IAB thanked the NOMCOM liaison, Dave Oran, for the excellent job he has done. Dave will forward the IAB’s decision to the NOMCOM chair.