Hello everyone, It has been quiet for some time now, so I think we should start to discuss the final /8 proposals again. In part 2 of the discussion, we discussed how many addresses an LIR should be able to get from the final /8. I'll try to summarize that part of the discussion. Please correct me if I am wrong or if I missed something: Most people preferred option b: All requests are downscaled by a certain factor Some people preferred option a: Everyone gets one (and only one) fixed size block There were comments that IPv4 would never run out, but let's assume that it will run out. In the worst (best?) case we won't need this final /8 policy, so it doesn't hurt to be prepared when we do need it. Because most people preferred option b, let's discuss some of the details to see if we can make that work. If we are going to downscale address space requests, how should we do that? Some issues: - If a company demonstrates the need for a /20, we could give them (for example) a /26. Smaller organisations would end up with even smaller amounts of address space. Currently prefixes longer than a /24 can not really be routed on 'the Internet' (for some definition of 'Internet'). Do we expect this to change, or do we have to make the minimum allocation size a /24 so that LIRs receiving address space can use it? - Do we want a maximum allocation size? This might be useful to prevent a few large companies to grab most of the address space, but it might also be unnecessary. - Do we want to make exceptions for certain situations where downscaling is not possible? For example a server farm with many HTTPS sites can't use less IP addresses than the number of websites (with the current protocols). I would like to leave PI assignments out od the discussion for now. I think we have enough PA related questions to discuss, but don't worry: we will get to the PI discussion later. As before: please stay on topic. This discussion is hard enough as it is :) Thanks, Sander Steffann APWG co-chair