my take: the MIF problem statement draft about to be published as RFC) discusses the various scenarios. probably more exists. But the issue of receiving config info from multiple sources is not specific to dhcpv6. it is related to any protocol sending config info. so it is a problem of pppv4, pppv6, dhcpv4, dhcpv6, RA, maybe even others. Therefore, we need to find the right solution for all the cases together, instead of just looking at one specific provisioning protocol IP address family. In this context, it appears to me that the right place to discuss that is within MIF wg.
my 2 cents. Marc. Le 2011-11-21 à 16:45, Ted Lemon a écrit : > On Nov 21, 2011, at 3:36 PM, Thomas Narten wrote: >> Isn't this what DHC already does? I.e., you run DHCP on two >> interfaces, and get conflicting information. > > I don't know what the antecedent to "this" was intended to be. > >> The existing DHCP specs are silent on this, and the DHC WG has never >> been willing to specify how to handle this case. > > Right, because it's been our position that this is not a DHCP problem. > >> The case of getting conflicting DNS configuration information via DHC >> seems a realistic scenario to me. > > It is completely realistic. > >> How do you propose this get handled? > > I think the right way to handle this is the way that the MIF working group > has proposed: treat each provisioning domain separately. Do not try to > merge information you have no basis for assuming is valid across provisioning > domains (which is all configuration information). You wouldn't typically > send an IP packet with a source address for interface 2 out of interface 2. > Why would you query a DNS server out of interface 2 when you learned about it > through interface 1? > >> Are you proposing we solve it for the DHC/RA case, but not for the >> DHC/DHC case? > > I think that having two ways of configuring things is a bad idea, and hence I > claim that standardizing the RA DNS server option was a mistake, one which > unfortunately can't be undone at this point. I think that if we are going > to say that RAs and DHCP servers can both provide some piece of configuration > information, then we kind of have to treat them as separate provisioning > domains. But we can't really do that, because some information is provided > only in RAs at the moment. > > My main wish is just to not make the mess any worse; secondarily, I'd like to > do what we can to make it less bad. > _______________________________________________ > dhcwg mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/dhcwg -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
