On 19-Apr-19 00:23, Alexandre Petrescu wrote: > > > Le 17/04/2019 à 23:55, Brian E Carpenter a écrit : >> Hi Alexandre, On 17-Apr-19 21:41, Alexandre Petrescu wrote: >>> Brian, >>> >>> Le 16/04/2019 à 04:18, Brian E Carpenter a écrit : [...] >>>> I think I will drop this discussion until ipwave gets its two >>>> main drafts properly synchronized. >>> >>> I would like to ask you whether you disagree that we move the >>> appendix titled "ND issues in wireless links" away from the >>> IP-over-OCB draft into the IPWAVE Problem Statement draft? >> >> My understanding now is that the IP-over-OCB draft is scoped only for >> single-link subnets (i.e. does not even try to cover the case of >> hidden nodes). I think that needs to be stated very clearly, > > I propose this text in the Subnet Structure section of IP-over-OCB draft: > NEW: >> The subnet structure on which the Neighbor Discovery protocol (ND) >> on OCB works ok is a single-link subnet; the status of ND operation >> on a subnet that covers multiple OCB links that repeat the signal at >> PHY layer, or the messages at MAC layer, is unknown. OK. We could argue whether it should be "unknown" or "undefined" but I think this is a useful clarification, thanks. > > [...] > >> and I don't quite understand whether there is a non-link-local prefix >> at all. > > In my network of 3 cars no, there is no non-LL prefix at all between cars. > > In my deployment of 2 cars and an IP-RSU IPv6 4G NAT66 NEMOv6 there was > a ULA prefix between the RSU and the car, on OCB. In year 2015. > >> In a very fluid network situation, it isn't at all obvious that a >> useful non-link-local prefix can be established. > > I agree. > > The problem is probably less in the fluidity, than in determining authority. > > When there is an RSU one may associate authority to RSU that got it from > Cellular Operator who got it from the Default FRee Zone; but absent RSU > who should be trusted as authority? > > It is the same problem when we try to make security between several cars > in a convoy. We try to use quickly the OpenVPN software to achieve > security, but who's the client who's the server? Cars are created > equal. They are equally safe, even though some are more expensive than > others. We may end up with two openvpn tunnels between two cars, when > one is normally sufficient. > >> The draft seems a bit ambiguous on that point, so perhaps that can >> also be clarified. > > YEs, could be, in the Problem Statement draft. > >> Given those changes, I agree that moving the appendix as you suggest >> would be reasonable. >> >> By the way, where you use the word "global" to describe IPv6 >> addresses or prefixes, I suggest using either "Globally Reachable" >> (as defined in RFC8190) or "global unicast" as defined in RFC8200. >> ULA != globally reachable. ULA == global unicast > > Noted. OK Brian > > Alex > >> >> Regards Brian >> >> >
