>> Iljitsch van Beijnum wrote: >> However, geographic addressing could give us >> aggregation with provider independence. > Brian E Carpenter wrote: > You'll have to produce the BGP4 table for a pretty compelling > simulation model of a worldwide Internet with a hundred million > enterprise customers and ten billion total hosts to convince me.
The problem with geo PI aggregation is expectations: if we set any aggregation expectations it won't fly because too many people have legitimate concerns about its feasibility. Personally I think that some gains are possible, but I'm not sure these gains will offset the amount of work required. My $0.02 about geo PI: a strategy change is needed. Instead of presenting geo PI as the solution that would give PI without impacting the routing table (this will not fly because there are too few believers and too many unknowns), present it as the icing on the cake of a comprehensive non-geo PI proposal. >> Michel Py wrote: >> That being said, I do acknowledge that larger companies such as >> global ISPs do have a problem with the RFC1918 space being too >> small. This brings the debate of what to do with class E, either >> make it extended private space or make it global unicast. > Eliot Lear wrote: > I think we bite the bullet and go to IPv6. I just rubbed my desktop lamp and I regret to report that no genie came out of it to grant your wish. > Screwing around with Class E address space at > this late date is counterproductive. Not screwing around with it means more dirty tricks and more double NAT. IMHO the strategy of keeping class E reserved hoping that it will accelerate the deployment of v6 is futile. > Iljitsch van Beijnum wrote: > [ARIN PI policy] > and now just as these efforts are getting close to paying off > (shim6) a small group of people decides to throw caution in the > wind and adopt a policy that opens the door to exactly the > problems the IETF has been working so hard to avoid. These people are tired of hearing the broken record "getting close to" that they heard consistently for the last 10 years. > Kevin Loch wrote: > In case you (IETF) didn't get the memo, the operational > community has flat out rejected shim6 in it's current form > as a replacement for PI. > This failure of leadership from the IETF to provide a > roadmap for a viable alternative to PI is a factor in the > support for going with the current technology. I have to agree with Kevin here. Michel. A little late and off-topic: > Iljitsch van Beijnum wrote: > Peer-to-peer isn't a good example, because of the high built-in > redundancy. Even someone who can only set up outgoing sessions > can run BitTorrent without too much trouble because there are > plenty of peers without NAT or portmappings of some kind (manual, > uPnP or NAT-PMP) that can receive the incoming sessions. When the > sessions are up, traffic can flow both ways. Technically speaking you are correct (this is called tit-to-tat I believe), but it does not work in the real world, reason being UL/DL ratios. Many "interesting" (use your own definition of it) torrent sites have dedicated trackers that enforce ratios; relying on the mechanism you describe can never achieve a UL:DL of 1. I do acknowledge that there are scores of leechers out there, but most of them sooner or later will open the ports in order to be able to seed completed torrents and up their ratio. Demographics show large number of teens who are technically savvy enough to get into the Web interface of their NAT box with admin/admin and open the ports. _______________________________________________ Ietf mailing list Ietf@ietf.org https://www1.ietf.org/mailman/listinfo/ietf