Actually, you got me thinking. Why not require anycast, indeed. I think at some point it will become necessary. Maybe as a SHOULD this time. The argument of 'operational autonomy' cannot be sacrosanct. If you agree to provide a service that the whole internet depends on, then you need to comply with a few requirements. If you can't / won't, well... you can opt out. Operating a root server is not a god-given right or burden. Just opt out if you cannot fulfil the requirements the whole Internet needs. Some things just come with the territory. If, on the other hand, you operate some random email server out there, then yes, I agree you are pretty much the king of your castle. You are free to not do DMARC if you don't want to. I do agree with Patrik that the enforcement part is not the IETF's responsibility. That lies elsewhere. But past failures in enforcement should not deter the IETF of setting the requirements the IETF deems necessary for the correct operation of the Internet. ~Carlos On 29/05/2014 18:24, manning bill wrote: > and why not require anycast? if your going to meddle in other companies operations, be bold! > > > /bill > Neca eos omnes. Deus suos agnoscet. > > On 29May2014Thursday, at 14:13, Carlos M. Martinez <carlosm3011@xxxxxxxxx> wrote: > >> I need to clarify something here (thanks TM for spotting this). >> >> I didn't mean to say that the IETF should require all root server >> operators to provide anycast copies. I'd very much would like them to, >> but there yes, I don't think the IETF can/should require that. >> >> My comment about 'this requirement is well within...' was only intended >> to apply to the IPv6 issue. >> >> regards, >> >> ~Carlos >> >> On 29/05/2014 17:18, Carlos M. Martinez wrote: >>> I think there is enough consensus saying that root server operators MUST >>> support IPv6. I think it's hard to argue that the Internet needs this to >>> move to IPv6, as otherwise we'll be saying that it'll be ok for future >>> networks to not be able to access some root servers, or putting the >>> burden of supporting all IPv6 on a subset of root servers. >>> >>> If you add that not all root server operators offer anycast copies, or >>> do it in a limited way, well, we could be putting the IPv6 internet in a >>> fragile position. >>> >>> IMO, setting this requirement is well within the core competencies of >>> the IETF. >>> >>> Then comes the question what to do (if anything) with those root server >>> operators who chose to ignore this MUST. >>> >>> IMO, This is probably outside the IETF's sphere, and it should be >>> possible to even say so in the proposed document. >>> >>> cheers! >>> >>> ~Carlos >>> >>> On 29/05/2014 05:24, Jari Arkko wrote: >>>> >>>>> I would like every A-M.root-servers.net have an A and AAAA record. >>>>> >>>>> I don't care how the root-server operators decide to partition to workload >>>>> among hardware. >>>> >>>> Yes, that is my view as well. >>>> >>>>> Over time we will need more v6 responders and fewer v4 >>>>> responders. >>>>> I don't think that there is, or should be, any requirement that v4 and v6 be >>>>> answered by the same system, and given anycast, they might even be in >>>>> different locations. >>>>> >>>>> I think that the current text captures this just fine: >>>> >>>> Agreed. >>>> >>>> Jari >>>> >> >