On Jun 8, 2011, at 7:05 PM, Tim Chown wrote: > On 8 Jun 2011, at 21:19, Keith Moore wrote: >> >> Nor, bluntly, is it about a few big content providers or whomever else you want to label as important. The internet is a hugely diverse place, and you don't get to dismiss the concerns of people whom you want to label as red herrings. Again, 40-something percent of the IPv6 traffic that is observed on the net today uses 6to4. That's about as much as Teredo, it's a hell of a lot more than native v6. As long as 6to4 is one of the major ways that people get IPv6 connectivity (and it clearly is), it's premature to declare 6to4 historic. > > You see 40% of your IPv6 traffic as 6to4, we see rather less than 1%. Our observation point is as a university on an academic/research network that is native dual-stack. We probably have most of our IPv6 traffic come from other universities around the world, who are also most likely natively connected. Hence little if any need for transition methods. This may be different to your scenario, of course, but it is hopefully a future that will be more widespread in time. I'd love it if we all saw a lot more native IPv6 traffic soon. Just to clarify, the 40% is not from my measurement. It's an approximation to figures I've seen quoted elsewhere. Like you, I'm sure this figure will vary from place to place. I haven't tried to do any measurement myself, because the amount of traffic is not a good indicator of overall usefulness. On the other hand, if any transit provider anywhere in the world is seeing 40% of v6 traffic as 6to4, that is a pretty good indication that somebody (besides myself) is using it. For that matter, the very fact that operators are observing problems with relay routers is another indication that people are using 6to4. Why would they be using it if they didn't want to? I realize that some platforms enable 6to4 by default, but not all of them do. And I've already said I support having hosts and routers ship with 6to4 disabled by default. > We did use 6to4 in its router-to-router, site-to-site flavour many years ago while a project called 6NET ran, but have had no use case for it since. Perhaps it would be useful to see your use cases more clearly documented with examples. I've already given examples. People keep looking for more specific examples in an argument where any specific example can be dismissed as irrelevant. It's not any one specific example that matters, it's the fact that people are using 6to4 and there's not an obviously better replacement that's available to them. > The problem is that 6to4 is unfortunately also harmful to real users, at least the ones that don't want to know about IPv6. It will continue to be until we can be confident no vendor anywhere has 6to4 on by default, won't it? NATs are harmful to real users too, and they do a lot more harm than 6to4 does. When will we deprecate them? When will we declare them Historic? It's misleading to talk about only the harm being done by 6to4 without acknowledging the benefits of 6to4 or the lack of a suitable alternative. And to say that 6to4 does harm is misleading. Is it 6to4 that's doing the harm, or people who advertise routes to relay routers that don't function properly? Why are people blaming the 6to4 protocol for configuration errors made by network operators? > The question is whether Historic stops knowledgeable people like you using 6to4 safely in your own context/community, without affecting 'normal' users. Does it mean 6to4 off be default, or 6to4 removed from product? Historic doesn't stop someone who can write his own code. But if it results in implementations removing support for 6to4, declaring 6to4 as Historic will stop people who use those implementations. >> It's not one versus the other. 6to4 is helping to promote ubiquitous IPv6. > > The other view is that 6to4 is delaying ubiquitous IPv6 deployment, by adding brokenness. Geoff's stats illustrate that very well, though those are not based on vanilla 6to4. I disagree with that assessment, because it's only considering the case of using 6to4 when IPv4 would work just fine. That's not an appropriate metric. Nobody who has native IPv6 connectivity needs to use 6to4 to reach native IPv6 destination addresses. But a deeper problem is the notion that a single set of address selection rules will work well for all, or even most, applications. Keith _______________________________________________ Ietf mailing list Ietf@xxxxxxxx https://www.ietf.org/mailman/listinfo/ietf