!! Proposed Addressing Structure The address ranges to be allocated are in the * 172.16.80.0/20 (172.16.80.0-172.16.95.255) range for routing nodes * 10.10.0.0/16 (10.10.0.0-10.10.255.255) range for client nodes. See (http://www.freenetworks.org/moin/index.cgi/NetworkAddressAllocations freenetworks) for more details. ! Routing Nodes Routing nodes are IP routers in distinct locations (e.g. separate nodes on the node map) that * agree to meet Melbourne Wireless group consensus on the minimum requirements on routing nodes, such as ** you aim to have 24/7 uptime * are prepared to forward traffic on-behalf of peer Melbourne Wireless nodes to other peer Melbourne Wireless nodes * run OSPF as the wide area network (between nodes) interior gateway protocol, in a configuration prescribed by Melbourne Wireless group technical consensus Routing nodes will be allocated one address from 172.16.80.0/20 range per __wide-area__ wireless interface. Along with this will be an allocation of one or more /28 subnets from 10.10.0.0/16. This address space can be further subnetted to supply connectivity to (local area) wireless and wired client networks that form [stub] networks from the node. Reachability to directly connected 10.10.0.0 /28 subnets will be advertised through OSPF. ! Client hosts If you do not meet the above requirements, for instance * you don't want to run OSPF but are a distinct node on the nodemap * you have just a laptop that you want to connect via wireless to a real routing node * you have additional routers in a home network connected to your routing node then you are to use addresses from the 10.10.0.0/16 space. Essentially only nodes that run OSPF across the WAN will use 172.16.80.0/20 addresses. If you fall into the first category above - you don't want to run OSPF on your node - then you can only usefully use one WAN link to a true router-node. You must negotiate with this node owner to obtain 10.10.0.0/16 address space (she might for instance request another /28 on your behalf). The router-node might statically route this space to your node, or they might run RIPv2 with just you for instance. ! Picture Will draw picture to show this in action... ! IPv6 The IPv4 allocations will be mimicked in the site level addressing part of an IPv6 /48 prefix (ie. bit positions 49-64) for those nodes that desire IPv6 address space. For the WAN addresses, this address will be fec0:172:16:xxxx::1/128 and for the clients subnets [[2002/fec0]]:????:????:xxxx::/62. This leaves space for future allocations 'over-the-top' when (http://www.faqs.org/rfcs/rfc1918.html RFC1918) IPv4 address space is available to us is depleted. For example, if a node is assigned '''172.16.81.33 and 10.10.66.240-255''', the corresponding IPv6 addresses/prefixes are also assigned: '''<48bitprefix1>:5121::1/128 and <48bitprefix2>:42F0::/62'''.