All: Directly over Ethernet is the lowest overhead encapsulation (less than UDP) and does not come for free with the direct over UDP proposal. In addition, supporting NSH over VxLAN is also an L2 encapsulation. Both of these support MAC Chains as in https://datatracker.ietf.org/doc/draft-fedyk-sfc-mac-chain/ as well as NSH tuple match and MPLS forwarding. At very least we would need both direct Ethernet and direct UDP to cover all the proposals.
The NSH header is a tag header used to steer packets to intermediate destinations (which are SFs). The proper layering is between the L2 and L3 headers not at L3 or L4. Cheers, Paul -----Original Message----- From: Dino Farinacci [mailto:[email protected]] Sent: Friday, November 06, 2015 7:04 PM To: Surendra Kumar (smkumar) Cc: Bottorff, Paul; Lucy Yong; Alia Atlas; [email protected]; Larry Kreeger (kreeger); [email protected]; <[email protected]> Subject: Re: [sfc] comment on draft-kumar-sfc-nsh-udp-transport > On Nov 7, 2015, at 10:51 AM, Surendra Kumar (smkumar) <[email protected]> > wrote: > > That doesn't mean DNS is spec'ed to run over VXLAN or MPLS. But DNS is > transport independent. > SK> It is spec'ed to work on IP/UDP, understood and makes sense. It should be spec'ed to work ONLY over UDP/IP. Everything else comes for free (or at the cost of what these NSH packets run over). Dino _______________________________________________ nvo3 mailing list [email protected] https://www.ietf.org/mailman/listinfo/nvo3
