On 6/17/2016 10:41 AM, Saumya Dikshit (sadikshi) wrote: > A new version of I-D, draft-saum-nvo3-pmtud-over-vxlan-03.txt > has been successfully submitted by Saumya Dikshit and posted to the > IETF repository. > > Name: draft-saum-nvo3-pmtud-over-vxlan > I'm confused by a number of things in this doc:
- PTB is an IPv6-ism, but the doc also refers in places to the DF bit. The relevant IP version should be indicated to clarify what protocol is affected. - the applicability claims relevance across protocols that don't use VXLAN, which makes no sense given the claimed focus of the doc. - the doc refers to places (ToR) and protocols (VXLAN gateway) as if they are synonymous. It needs to be more clear whether you intend place, protocol, or both - but "place OR protocol" makes no sense to me. - the issues raised here have nothing to do with VXLAN or NVO3, and IMO are endemic to all tunnels - which is why they are already addressed in intarea-tunnels. The need for this as a separate recommendation should be made much more clear, and the relationship of the recommendations here to those should be also made more clear. - the "error" described in the end of section 3.1.1 is actually a feature. When you ask the network for a path MTU, you're asking "what is the MTU of the path", not "what is the largest message that the path can handle without internal fragmentation". There's no way to determine the second, especially over a tunnel. A tunnel *is* a link - the whole point of a link is to isolate the upper layers from how the link needs to handle things. FYI, Fred and I have tripped over this issue many times before. We agree that "largest unfragmented packet over a link" might be an interesting question for a network layer to ask of a link layer, but ICMP doesn't do that - it reports only what gets over the network layer. - the doc ignores PLPMTUD, which is what is actually needed to overcome the limitations of ICMP and ICMP translation here. Joe _______________________________________________ nvo3 mailing list [email protected] https://www.ietf.org/mailman/listinfo/nvo3
