[address-policy-wg] IPv6 PI for HOSTING
Lorenzo Colitti lorenzo at google.com
Fri Dec 25 03:14:03 CET 2009
On Fri, Dec 25, 2009 at 12:09 AM, Masataka Ohta <mohta at necom830.hpcl.titech.ac.jp> wrote: > IPv6 will be ignored, because IPv6 is, technically, useless, which > has nothing to do with IPv6 address allocation policy. > > Because IPv4 NAT can be fully transparent end to end, it's fine to use > port restricted IPv4 with NAT. Well, but there are lots of things you can't do with port restricted IPv4 and NAT. For example, run a peer-to-peer program with 1000 simultaneous connections. Run itunes and Google maps on or your computer, your girlfriend's computer, and your phone at the same time. Run proper VOIP and skype call without using third parties, which degrade your call quality if their internet connections drop packets. On the server side, IP address sharing on a large scale means lack of geolocation, and lack of geolocation means no location-aware services. Search for ramen from your hotel room and expect to find places nearby? Who knows, you might find something in Hokkaido. Want to stream a baseball game in San Francisco? You might not be able to, because the licensing for that content requires you not to be in San Jose and your IP address might mostly be in use by people in San Jose. Trying to go to a website? Better hope your IP address neighbors aren't sending it too many queries, because otherwise the website operator might block your IP address for excessive use. Without IPv6, there will be more and more pressure on port space as user numbers continue to grow, so these problems are likely to get worse and worse. What happens then is anybody's guess. Maybe the quality of Internet connections will get worse and worse and you'll need to pay more for a better connection? Maybe the only reliable services will be the services of the ISP you subscribe to, so we'll go back to a walled garden model? > It's impossible because attempts of optional headers, path MTU > discovery, stateless autoconfiguration, aggressive introduction > of multicast etc. to make IPv6 better than IPv4 have totally > failed only to make IPv6 and its operation a lot more complex > and a lot less consistent than IPv4. As someone who has designed and operated IPv6 networks, I can say it's not more complex, it's just different. In some ways it's easier, too: for example, the large address space eliminates allocation problems and allows addressing plans that are much easier to manage. Autoconfiguration is very useful, and so on. But of course, if you already know how IPv4 works and you don't want to learn anything new it will seem complex. As regards it being impossible, I disagree, and so would the hundreds of thousands of users who use Google over IPv6 every day (assuming they knew they were in fact using it).