It's true that the address (as the role of identifier) is hard to be in accordance with the topologically aggregatability requirement. However, it is still workable to be aggregatable in the overlay or virtual topology, e.g. DNS or home agent in MIP. As for DHT, is it workable to support global-scope resolution for the flat id (or host- granularity PI address) based on the assumption that every host will need a unique entry in the DHT system? I believe it's a big difference in the scale between the above resolution infrastracture and the current structured P2P system, like Skype. Besides the scalability concern, there is also a business and trust model issue, that is, who will run this infrastracture? Xiaohu
_____ 发件人: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] 代表 Heinrich Hummel 发送时间: 2008年8月20日 16:56 收件人: [email protected] 主题: RE: [RRG] Renumbering... Tony Li wrote: I think the real question is whether aggregates are required at all. If so, then I think we have an issue that leads us down the NAT path. Now, if someone has some mapping technology that does not require aggregation at all (DHTs?), then there is a possible alternate to NATs. In answering: This is exactly the key point of my approach: forwarding without being based on aggregates of Unicast user addresses. Address aggregation is a problem just because of the huge quantity. Additionally, it is principally wrong in view of user mobility and eventually router mobility. Heiner
