-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 01/08/2013 22:00, Rob Janssen wrote:
What we need is a request from a user to get one or more addresses allocated, accompanied by information like region where they live. The coordinator then must be able to look into the already assigned list of addresses and find an available address in the correct region (subnet), assign it, and return the address to the requester.
The question is should allocation be part of a regional sub allocation which itself is part of a sub allocation from 44.0.0.0/8?
The most efficient way is probably the way the RIRs have been handling out allocation in the last couple of years. It's probably just easier to determinate the network size and assign the first available network of that size. Eventually you could temporarily reserve the adjacent network of the same size so that the network could be extended to that size if necessary. An example:
44.a.0.0/27 for user A 44.a.0.32/27 reserved for user A extension or until exhaustion of the 44.a.0.0/16 block 44.a.0.64/27 for user B 44.a.0.92/27 reserved for user B or until exhaustion 44.a.0.128/26 for user C 44.a.0.192/26 reserved for user C etc etc
Once you have filled all of 44.a.0.0/16 you can start allocating the reserved blocks. However if in the meantime user B needs more addresses, you can extend his allocation from 44.a.0.64/27 to 44.a.0.64/26. The renumbering is pretty easy in that case, a change if netmask is all that is required (in this example). OTOH there is no link between a physical LAN and the addressing, a physical LAN might be using 2 or more subnets (44.a.0.64/26 and 44.a.0.192/26) for addressing.
Dividing the network into sub-areas will cause issues in case some areas have more needs for addresses while other sub-area networks remain almost empty/unallocated.
The encap file and the RIP44d deliver a full routing table for the AMPRnet with absolutely no route aggregation, so the routing table will always be the same size (e.g. their is no gain in routing table size).
CIDR has been introduced in 1993, we shouldn't make the AMPRnet "classfull" by sticking to multiple levels of sub-allocation.
But those are just my thoughts.
73 de Marc, LX1DUC