Note Taker: Reinaldo Penno 5 min: Administrativia, chair * Two WGLC on DS-Lite and 6rd * DS-lite updated to 04 due to comments. Version 05 to be published soon 5 min: 6rd: Ole Troan http://datatracker.ietf.org/doc/draft-ietf-softwire-ipv6-6rd/ * >100 comments received on WGLC * Published v.08 last week * 6rd operational document? * There are MTU issues that span many tunnel types and are not types to 6rd * Remi Depres: Question on Anycast and DF=0? In 6rd there should be no fragmentation. * Ole: Typo, should be DF=1 * Remi Depres: Do not understand the rationale the behind PRL option in DHCPv4 option * David Hankins: A client has to request the DHCP options it want to receive * Ole: I believe Remi is thinking about PRL as in NBMA networks. * Remi Depres: Possibly yes, is there any other use? * Alain Durand: Need to move to next topic * Fred Templin: Have a document in the editors queue about use cases for VET that might serve as template for 6rd use cases 10 min: GI-DS-lite: Sri Gundavelli http://datatracker.ietf.org/doc/draft-gundavelli-softwire-gateway-init-ds-lite/ * New version of the draft (-01) incorporates IP-in-IP and IP-over-MPLS * CID and TID as flow identifiers * Adopt as WG document? * Alain: This draft was presented in 3GPP workshop and there was a great deal of interest * Alain: Anybody object to take this document as WG item? * No objections * Gregory L: Multiple tunnels types creates a lot of churn in the industry: interoperability, deployment, testing, etc * Consensus: If there is agreement to narrow it down, it would be good. Comments the mailing list * Zhen Cao : Possible impact on mobile networks in terms of charging and billing because of encapsulation. * Presenter: Little or no impact. Need to understand the big picture of NAT solution and specific architecture * Zhen Cao: Impact of multiple shared addresses being used by mobile host * Alain: Need to move to next topic 40 min: PCP: Dan wing/Dave Thaler http://datatracker.ietf.org/doc/draft-wing-softwire-port-control-protocol/ * Alain: When transition talk started comments were heard loud and clear that users (though applications) need to be in control of their NAT * Dave Thaler (DT): I will talk about APIs. These APIs do not change since the one full IP model is maintained * DT: Two separate app scenarios: static vs. dynamic * DT: Example of NATUPnP library APIs on windows. General observations on NATUPnP. * DT: NATUPnP targeted to static port mappings, * DT: Example on Apple's DNSServiceNAT API. Supports both UPnP and NAT-PMP * DT: Apple's has a lifetime parameter and therefore design for dynamic mappings * Martin S: Purpose of the presentation? * DT: How applications use these APIs (there is a lot of confusion on that) and how they tie into PCP. * Dan Wing (DW): Applicability for NAT44, SNAT64, SLNAT64, DS-Lite * DW: Requirements: Simple, * DW: 'why not my favorite protocol?'. None meet the requirements. * DW: PCP basics: Lightweight, simple, not much state in either LSN or CPE * Gregory L: How likely the application will get the port it asks? * DW: Not sure of the question. NATs have their way of allocating ports and depends on address pool size * Alain & DT: Many applications and protocols already work around that. * Juergen Q: How PCP can guarantee that a PCP client can only open ports for devices on its home? * DW: Prefix validation. Ties into Embedded PCP Server model * Dave Thaler: Mapping APIs/protocol to PCP * DT: Applications should not have to known if there is PCP * Gregory L: Kludgy is a misnomer. Skype does kludgy things under the hood but works great * DT: Agreed * RD: A simple extension would be to ask for a port range and solve other issues * Simon P: What happens if you are behind three layers of NAT? * DW: We will consider this case * Stuart Cheshire (SC) : Each NAT along the path could remember state and avoid hairpinning * Gregory L: Doesn't all this state defeat the purposes * ?: Why not 'STUN'? * DW: We looked into STUN and rejected the idea. We can talk offline about it * ? : Why not 'DHCP'? * Alain: DHCP has other protocol limitations for NAT traversal * SC: We cannot use STUN because there is no lifetime * More discussions on 'Why not my favorite protocol?'... * Martin Stimerling: Issue with Cascade NATs. Need more investigation * Alain: is there interested to continue working on PCP? * Consensus: yes * Alain: is PCP a good general direction? * Consensus: yes * Alain: Anybody object to PCP becoming a WG item? * 2 hands * Alain: Who thinks this should become a WG item? * Quite a few hands. Alain will take with ADs * Adjourn