Christian Seitz <chris@in-berlin.de> wrote: > What do you think about Geoffs presentation? Did the Internet change and do > we have to adjust our expectations? Do you agree to the fact that we now have > "A Network of Names" and are using a lot of CDNs to bring content to the End > Users and therefore perhaps do not need to do more IPv6 migration or do you > think we all should find a way to somehow speed up the IPv6 migration around > the world to enable end-to-end communication between all devices again? I can't argue too much with Geoff's results. This is not the first talk of his that points to this result; it's been a trend of many such results in a variety of different ways. They don't make me happy, and I know Geoff isn't particularly happy the results, so I feel sad. I wonder (hope?) that this might be a blip in a much longer history (yet to be written) where end to end becomes more important again. If it makes sense due to Moore's law to push content to the nearest data center, maybe it will make even more sense to push it even closer to me. But, on the intellectual side, I think that there are some things that we can do that would be useful. Specifically, let's put away our emotions for a moment and imagine that this CDN-focused architecture was always our goal, and consider how we would really do this properly for IPv4. For instance, 1) 0.0.0.0/1 is for eyeball networks only, and is repeated RFC1918-style into every jurisdiction, or part of a jurisdiction. (forget about how we get there) 2) 128.0.0.0/2 is for content. Not sure if it's globally unique or not. 3) 192.0.0.0/3 is for global infrastrastructure. [4) IPv6 is for global infrastructure] This would be a table-top game, and the goal is to find out what works, where the fillable gaps are, and what gaps can not be filled. What do we break? I imagine this to be weekend-before-RIPE hackathon-like (OARC-like) event. -- Michael Richardson <mcr+IETF@sandelman.ca>, Sandelman Software Works -= IPv6 IoT consulting =- *I*LIKE*TRAINS*