Re: [44net] directly routed subnets

An AS is from a formal point of view just a number that you need to peer with other ASes as an independent network with your own policies, including routing policies, ethics, acceptable use policies, etc. There is no extra hardware. You use the AS-number when setting up external BGP peerings in your border router(s), in which you tell your peers what routes you want to export and whether you want transit or not, while they tell you what routes they want to import and whether they offer transit or not. This is what is currently done in the AMPR border router at UCSD of of their AS (AS7377). And by making a traceroute to ampr.org = 44.0.0.1 you can see that AMPRnet has its transit via CENIC (Corporation for Educational Networking in California), the californian regional research and education network backbone. If peering would be allowed elsewhere than out of the UCSD AS, you would either have to accept the policies of another local AS in your area, or use a dedicated AS for AMPRnet. The former will have an impact on what you can do and not do that would not necessarily be compatible with the goals you would like to achieve with AMPRnet. The latter definitely sounds to me as the rational thing to do if the intention is to develop AMPRnet as a global ham resource with its own policies. This is exactly what the AS concept is for and I do not see any problem for AMPRnet to get an AS-number. Regarding tunneling, you have to differentiate between the external peering (interdomain routing) and internal peering/routing between AMPRnet subnetworks (intradomain). if you allow external peering anywhere, the need for tunneling to the one and only border router at UCSD to go to Internet disappears since each subnet can peer with any other local network, even at a local Internet Exchange Point, and get transit to Internet locally. This means that a specific subnet will not depend on the global Internet connectivity to get to a local destination outside AMPRnet. The idea is to keep local traffic local. The challenge is still to get all AMPRnet subnets connected iternally, intradomain. But isn't that exactly the challenge that we want to keep? To create a global ham network in parallel, but peering, with Internet based on radio links? It will most likely be easier to set up local islands than getting global internal connectivity, but that is already the case. The challenge is somewhat more manageable if we can control the routing policies and all intermediate nodes. We can use dedicated terrestrial wireless links at different amateur frequencies, amateur satellites or maybe even some dedicated wired stretches (god forbid :-) if nothing else is available. On stretches where we by no means can come up with a dedicated link, we can still tunnel.... The internal intradomain routing can be set up statically or using an intradomain routing protocol such as OSPF in each AMPRnet island and, if needed, export routes between islands using iBGP as opposed to the eBGP functionality used in Interdomain routing. Regarding how we are going to pay for this, I would leave that to local agreements about peering and transit. I do not know exactly what agreement is currently into effect regarding the connection between AMPRnet and CENIC, but I have good hopes that we can come up with an ethics and acceptable use policy that would be acceptable for the Swedish University network to accept peering and maybe even transit of AMPRnet, at least if the transit comes via a member university taking care of the policing. The cost would in that case most likely already be covered by the university membership fees to Sunet. The extra cost induced would be very marginal and, depending on the AUP, the purpose would already be supported. This remains to be discussed if we enter this route though. A local ham club or individual without any contacts with a research or higher education institution could go via their commercial ISP or check if there is a local Internet Exchange Point where AMPRnet would be welcome to connect, and cover the costs, if any, themselves. Those that cannot find an affordable agreement can still tunnel to Internet via their more fortunate colleagues... In my mind, an important internal challenge for AMPRNET is how to avoid the fragmentation of the AMPRNET address space that would result from allowing address allocations that are used only for Internet contact without any ambition to connect internally. In such cases I have a hard time seeing the reason to use AMPRnet addresses at all. A way to meet that challenge could be to require motivations for using AMPRnet addresses, require progress reports and make all allocations time limited, and I guess that the coordination of that should be part of a delegation agreement to be signed from the AMPRnet root and down the delegation tree. Bjorn On 2012-03-21 03:49, Lin Holcomb wrote: