This archive is retained to ensure existing URLs remain functional. It will not contain any emails sent to this mailing list after July 1, 2024. For all messages, including those sent before and after this date, please visit the new location of the archive at https://mailman.ripe.net/archives/list/[email protected]/
[address-policy-wg] RIPE Access Policy Change Request to allow allocations to critical infrastructure
- Previous message (by thread): [address-policy-wg] RIPE Access Policy Change Request to allow allocations to critical infrastructure
- Next message (by thread): [address-policy-wg] RIPE Access Policy Change Request to allow allocations to critical infrastructure
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Joao Damas
joao at psg.com
Tue Jan 20 16:50:03 CET 2004
Coming in late but let's see if I understand this: The problem has to do with the fact that there are requests for a /24 that plan to host only one address (maybe 2, one for a server and one for a router) and that if the requester tells the truth, it does not qualify. Note that from a routing point of view, if you look at an anycasted network from the outside, it is indistinguishable from a multi-homed network so in itself the routing is not the problem, it is the fact that you are asking for a prefix to host only one active IP address but you need it to be routable and today that means a /24. This goal could very well be achieved by keeping the addresses of the servers you are already using and migrating all other active IP's out of the /24 that contains the address you want to "anycast". After all, none of the root servers that are anycasting today have renumbered (one that is not anycasting today will renumber soon because they can't apply what I just described, which was unfortunate). The problem is then with obtaining a /24 for each anycasted site, not for the service IP but for the "management" IP, something that allows you to uniquely identify the different incarnations of the the same IP, geographically distributed and only accessible through networks other than your own (for the operator of the anycast service, anycast is indeed different from multi-homing). Depending on the requirements for independence (in its various aspects), you may use an IP inside someone else's PA block for that, if you buy hosting from them, for instance, or you may need to ask for a separate /24. It is this last part that may pose a problem within current RIPE policies. Note that the IPv6 root server policy has nothing to do with anycast but rather with routability of a prefix containing only one active IPv6 address and that there is no RIPE IPv4 root server anycast policy in spite of there being anycasted instances of root servers hosted by European organisations. Joao On 16 Jan, 2004, at 15:37, Gert Doering wrote: > Hi, > > On Fri, Jan 16, 2004 at 11:18:29AM +0100, Andreas Bäß/Denic wrote: >> RIPE should be able to allocate a single /24 IPv4 and a /32 IPv6 >> address >> block for the operation of anycast DNS servers if: >> - the operator serves a TLD zone >> - the number of (planned?) nameservers would lead to a truncation of >> the >> authority/additional section in a DNS delegation response containing >> the >> NS RRset in [draft-ietf-dnsop-respsize-00.txt 2.9] > > I would support that. > > Yes, I know that it's a very special case, and it could be done by > a "swamp" (or a random PI) /24 just fine. But I'm very much in favour > of > having official ways to do things that need doing - and I'm convinced > that this is a good thing to do. > > As for the impact on the routing table: this new policy would not make > more or less impact than just announcing a random /24, but it would > allow for better documentation what it's all about. > > As far as "other" anycast deployments go: there have been some voices > on the list that "we must not do a policy that's too narrow minded" - > yes, > that's true, but we don't want a "permit all" policy either (well, we > *could* do that, but there does not seem to be consensus to do that). > > *If* other people show up, and say "we want to do this-and-that, and > have a specific need that cannot be done in the current framework", we > can always adapt the then-existent policies. > > > (This is speaking as myself, and as APWG co-chair). > > > Andreas: I would appreciate if you could give a short presentation > (5 mins?) in Amsterdam, repeating the reasoning behind this proposal, > to stir some discussion. > > Then we can have a final call for consensus some time after the > meeting. > > Gert Doering > -- NetMaster > -- > Total number of prefixes smaller than registry allocations: 57882 > (57753) > > SpaceNet AG Mail: netmaster at Space.Net > Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 > 80807 Muenchen Fax : +49-89-32356-299 > >
- Previous message (by thread): [address-policy-wg] RIPE Access Policy Change Request to allow allocations to critical infrastructure
- Next message (by thread): [address-policy-wg] RIPE Access Policy Change Request to allow allocations to critical infrastructure
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]