Dino, On Thu, Oct 6, 2016 at 1:50 PM, Dino Farinacci <[email protected]> wrote:
> > Absolutely! I heard quite clearly at the last IETF that there is > general desire for NVO3 to pick a single encapsulation. While the > sentiments on the list have been more strongly for just publishing > everything and assuming complexity can be handled later, that comes from > much less of the WG than the desire for a single encapsulation. > > Let’s try to reach concensus. Could we agree on one of these 3 options: > Please let Matthew and Sam, as the WG Chairs, determine what calls for consensus to make and what are reasonable options. > (1) Publish all 3 encapsulations as Informational RFCs. This makes the > working group look indecisive but at least the vendors can go to market > with what they choose with some acknowledgement from IETF. > I don't see them going through in the current condition. Geneve says carry along arbitrary and unspecified data; that's a security/privacy issue - maybe resolvable with text. Geneve says MUST use path MTU discovery, but header-size can cause issues to get to encapsulated packet header to return to the VM. VXLAN-GPE doesn't define what to do with its OAM bit nor is it clear that is sufficient. It's likely that more flags and extensions will need to be defined to handle aspects like security. > (2) Pick one encapsulation to be Experimental RFC and the other two > Information RFCs. This makes the working group look somewhat decisive but > at least allows the vendors to go to market with varying degrees of > acknowledgment from IETF. > Yes, picking an encapsulation would allow documenting as options not picked (with existing issues described) to go forward. > (3) Close the working group and keep drafts published with current status. > Let the solutions working groups work on the control-plane. Because if the > working group cannot decide on a simple encapsulation format, there is no > way there will be conclusiveness on a control-plane. > I am certainly aware that closing the WG is one of my options. Regards, Alia This is the only way to get some rough concensus to get out of this > stalemate. I vote for (1). > > Dino > > P.S. A gateway approach is a horrible idea. You will get bad solutions > (suboptimal routing) for customers because a standards group couldn’t > decide? This doesn’t encourage and promote interoperability and of course > pushes the IETF more so into irrelevancy. > >
_______________________________________________ nvo3 mailing list [email protected] https://www.ietf.org/mailman/listinfo/nvo3
