On Thu, Apr 15, 2010 at 2:31 AM, Alexandru Petrescu <alexandru.petre...@gmail.com> wrote: > Le 15/04/2010 07:55, Parav Pandit a écrit : >> >> Hi, >> >> As per RFC 2464, Link local address for Ethernet based interfaces >> are based on the EUI-64 (derived from the MAC address). > > Right, that is probably a very widespread way of how the link local > addresses are derived from a MAC address. > > There are many other ways, for example link negotiation with PPP IPv6 > which for interfaces not having MAC addresses, like a 3G phone... and > knowing that 3G IPv6 handheld computers (watches, phones, pdas, pads, > netbooks) are potentially very numerous one may think it could be as > widespread as the first. >
Just as a point of clarification, 3G devices do not use PPP generally. The process is described in RFC 3316 Appendix A Cameron > For these, deriving the MAC address from the link-local address (w/o > doing ND) - will not work. > > A case that comes to mind is 6LoWPAN's _some_ version of "6LoWPAN > Neighbor Discovery" which requires the end node to do just that: derive > the MAC address from the LL address. I believe this intention wrongly > aimed. > > One shoudl try to identify why would one need to reversely derive the > MAC address from the LL address of some neighbor? Is it to save > messages? We can discuss that and see that sometimes it's not needed to > save, better use point-to-point links. > > Why do you need to do this message-less reverse resolution? > >> I have #3 questions based on this. >> >> In this case, when one Ethernet based host(from its link-local >> source) tries to ping the other Ethernet based host, it knows the >> Mac address implicitly (from the Link local address). > > In this case, if the problem is ping then the solution may be > "ping -I iface" (-I specifies the outgoing interface). > >> 1. Why is it required to explicitly do the Neighbor discovery for >> the link-local addresses? RFC 4861 says to do the neighbor discovery >> even for link-local addresses. Correct me if my understanding is >> incorrect. > > IMHO, one aspect is that it requires so in order to have the most up to > date info about the other node. Sometimes nodes change their LLs and > their MAC addresses (ifconfig does that). > >> 2. Does it mean that in Ethernet networks, interface can have only >> one Link-local address? If not then we violate the RFC 2464. > > I think I can do "ifconfig eth0 add fe80::1/64; ifconfig eth0 add > fe80::2/64; echo $?" and see success, I suppose - have you tried? > >> 3. Does RFC 4941 Privacy extension for autoconf apply to Ethernet >> interfaces? > > Well yes I believe so, I suppose I have it running on my PC right now. > > Alex > >> >> Regards, Parav Pandit >> >> >> >> >> -------------------------------------------------------------------- >> IETF IPv6 working group mailing list ipv6@ietf.org Administrative >> Requests: https://www.ietf.org/mailman/listinfo/ipv6 >> -------------------------------------------------------------------- >> > > > -------------------------------------------------------------------- > IETF IPv6 working group mailing list > ipv6@ietf.org > Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 > -------------------------------------------------------------------- > -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------