Hi David, Paula, colleagues, Some comments on the IPv6 policy draft document and the IAB comments. Hopefully these can be considered by the registries for the next revision. I decided not to take the document and analyse sections of it; rather I think we should collectively step back and look at the problem we are trying to address (no pun intended). What the document does not cover sufficiently: - Internet Exchange Points - newcomers to the IPv6 transit service - what is meant by a "customer" - what is meant by "transit" It is my understanding that the address design of IPv6 intended a subTLA (and later, a TLA) to be allocated to transit providers only, not leaf providers. The document doesn't declare this. A transit provider is a service provider who accepts routing information from a customer AS or peer AS, into his AS, and then announces that routing information to another customer AS or peer AS. Those of us who configure routers and run ISP backbones use this definition. It is my opinion that these service providers are the ones who qualify for subTLAs. "customers" are service providers who connect to transit providers as defined above. Service providers could be enterprise networks who have their own AS appearing in the Internet routing table for multihoming or management purposes. They could be ISPs as we know them in the commercial world, or they could be educational establishments; all have their own AS. These "customers" would be the organisations I'd envisage qualifying for NLA address space. These "customers" would assign address space to the end users. Using these definitions, look at the numbers. How many transit providers have 100 customers in the relationship as defined above. Well, a straw poll on the routing table shows very very few. 100 is too big, and I agree with the IAB that 10 may be a more realistic number. Secondly, IXPs. These are not mentioned, but I think we need separate paragraphs to cover IXPs. I'd suggest that an IXP would qualify for a subTLA provided they have a public document showing a memorandum of understanding, or a signed agreement, that the peering relationship between service providers at that location forms an IXP. Maybe a minimum of 10 service providers could realistically qualify such an operation as an IXP. Same barrier to entry as "transit" providers. 100 is definitely too high. Newcomers. Work needed here, and probably the most difficult one to properly define. Some businesses may wish to only offer IPv6 transit services. If they can show business relationships to provide transit services (defined above) to "customers" defined above, they should qualify for a subTLA. (By "relationships" I mean signed contracts or commitments from potential customers to "buy" transit services.) Build in safeguards re deployment etc, but allow for up to 2 years. We aren't switching IPv4 off one day, and IPv6 on the next. Now to a reality check. Using my definitions above, I analysed the current IPv4 Internet routing table. It makes interesting reading! - There are 4619 ASes originating networks in the Internet routing table today, 5th March. - There are 1608 ASes providing transit services in the Internet today. This is the number of providers who are providing connectivity to ASes originating networks. Rather than constructing complex rules about allocation to existing providers, we have the solution above. 1608 ASes are providing IPv4 transit services, so 1608 ASes qualify for subTLA space. Give it to them now, and they have two years to do something with it. If they don't do anything with it, revoke it. Second, the ASes which are originating networks are the service providers who will receive NLA address space from the subTLA holders. Some originating ASes will be transit providers too, but then this is why the hierarchy has been built into the address scheme... Wasting address space? I think not. 1608 transit ASes is 1608 of 8192 possible subTLAs. 1608 routes in the IPv6 routing table is tiny compared with the potential 16384 defined by RFC2450 in FP 001 IPv6 space. We'll need significant growth of the Internet to use all 8192 subTLAs. Yes, that is a huge amount of address space relative to the IPv4 Internet, but it is still only one of 8192 TLAs in one eighth of the total IPv6 address space. People "dusting off" ASes in their back pocket? Well, you need to do a lot of dusting off to become a transit service provider, especially if you have to show that you are providing transit to at least 10 other ASes. No danger there, provided "transit" is defined. People rushing to get ASes so they can get a subTLA? Well, same applies. You need to provide transit service. If newcomers sign business agreements or contracts with 10 other ASes (same barrier as existing service providers), then they too should qualify. And remember, the registries don't assign ASes unless you are multihoming between different ISPs. I hope this is useful and can act as a reality check - I hope we introduce a system whereby service providers from all parts of the industry and the world are encouraged to deploy IPv6. The document in its current form is an excellent first step at tackling the issue but I hope it can be enhanced to encourage further development and deployment of IPv6. best wishes! philip -- -------------------------------------------------------- Philip Smith vm: 6178202 ph: +61 7 3238 8202 Consulting Engineering, Office of the CTO, Cisco Systems --------------------------------------------------------