CIDR/Aggregation/Allocation Policies

Sean Doran

SprintLink< /a>

• Stan Barber's Notes
• Feb 1995 NANOG Home Page

Stan Barber's Notes

NSFnet AUP make aggregation difficult (UUNET and BARRNET opinions).

When there were a few national registries, geographical aggregation worked. Now that there are many and this is causing problems. Since there are still many 16Mb routers still out there and since there is no way that many can replace their routers before a year has past, CIDR fragmentation is becoming a big problem.

1. How to distribute ip addresses? Provider based addresses tend to work until a customer changes providers. Does the provider allow the addresses to migrate? Would that really solve anything? Why bother? It will prevent big problems during route flaps and it will make these small routers last longer.
2. How many addresses can one get? IPv4 space is not unlimited. Sprint is projecting exhausting the space in 18 months or so. Eric Carroll says that we are rationing ip addresses now since he says that the InterNIC is requiring engineering plans. Eric Bennett from AADS says that the InterNIC is doling out address space and it important that many providers appreciate the complexity of managing space effectively. Vince Fuller says that some providers will allocate address space in an unwise manner. Yakov asks what the US can do to come up with a consistant fashion. Mark from the InterNIC said the NICs are trying to work out a consistant set of policies for address allocations.

   What to do about subnets of Class C allocations?

   Rwhois can solve part of the problem. Bill Manning said that some of the InterNIC staff has been saying that SWIP should be used instead of Rwhois. Mark says he will talk to the InterNIC staff.

3. What are the factors that justify the allocation of space to an organization? Elise says that paying the registries and the class of membership would determine how much space an organization would get. Peter thinks that liabiliy is not a problem if folks are doing things effectively. Mark Fedor argue that folks will pay anything to buy IP space. Peter argues that folks have to be able to add registries so that folks who are unable with existing registries can create one they are happen.

   Eric Carroll argues that folks are already unhappy because folks are in a rationing system and nothing will be fast enough or be enough.

   Peter Lothberg argues that addressability is not an issue. Automatic renumbering would solve that, but noone has that, yet.

   If there is not a uniform set of standards for how addresses are allocated, how will registries be able to allocate them in standard ways?

   The IANA has a draft out that addresss some of these issues.

   Peter Ford argues that RFC 1466bis acceptance should go through the full IETF standards track.

   Vince Fuller thinks that 1466bis should say that the addresses should be recoverable unless a contract sez otherwise. He also sez that if renumbering is required, there needs to be some kind of estabilished window during which the old addresses would work while renumbering was happening.

   Breaking up Class A and Class B networks is in the offing and we need to start moving our engineering to support true classless addressing.