These are unedited transcripts and may contain errors.

IPv6 roundup ?? 5 May, 2011, at 11 a.m.:

CHAIR: We are going to go ahead and get started here. Please have a seat. Enjoy it as we finish up the RIPE meeting here.

So, for this meeting, we didn't have an IPv6 Working Group, which you probably noticed. What we did, instead, is we chose to load up the Plenary topics with IPv6 content instead. This is a little bit of an experiment. It may also be part of a grand world scheme to make IPv6 just a part of our daily lives and nothing special that we have to talk about. We are not there yet. But that's part of the idea here.

So, what we are going to do this morning for this final session is, we are going to do a round?up of the IPv6 that we have seen throughout the rest of this week. So this is actually going to consist of a couple of different parts. We are going to have some presentations talking about IPv6 world day which hopefully you have heard a little bit about. We are going to kind of change the session here and we are going to do a panel discussion where I am going to talk to some people I have asked to come here who know a little bit about IPv6 and discuss some of the things that I saw come up during the week, but of course we'll have open microphones so if you have IPv6 things that you want to discuss, then we'll be able to talk about that at that time, too.

So that's kind of what we are going to be doing this morning. I'd like to start off and ask Andre Robachevsky about IPv6 world day.

ANDREI ROBACHEVSKY: Good morning. I am Andre, I work in ISOC, and this is a presentation to tell you about the World IPv6 Day that is sponsored by ISOC and that will take place on the 8 June, this summer.

So what's this? This is a 24?hour test flight, and the main idea is that main content providers will open their front doors to everyone who is able to connect using IPv6, and no back doors, although many of them have these back doors for quite sometime. But, also, the goal is to motivate organisations across the industry from content providers to ISPs, hardware makers, etc., etc. to prepare their servers for IPv6 and ensure a smooth transition.

It's important for the Internet industry to collaborate and test IPv6 readiness and it's not the first. There were events similar to that before and they were very, very successful and we hope this day is going to be also successful.

So, what are the motivations? Of course, well, the main motivation is hatching the egg for IPv6 deployment. It's about improving IPv6 connectivity. It's an opportunity to increase confidence for content providers. It's an opportunity to feel the real IPv6 traffic for ISPs. It has many other side effects. Is provides a target date of already planned roll?out. We heard things like, I don't want to have to answer why I don't have this. And it's, as an Internet, it's a never?ending exercise and that's just one example of such collaboration.

This is the URL where you can find more information about World IPv6 Day, and information you need if you want to participate. So for web owners or for hosting companies, this information is there, you fill the form and you let us know that you want to participate and please contact if you are willing to do this and help us making this a real global IPv6 day.

So far, we had a lot of response from all over the planet, really. Website owners, ISPs, hosting companies, exchanges, vendors, etc., etc. We are very happy. There is some, you know, buzz created. It's partly a PR exercise, of course, and we create a website where we'll be listing our participants who are running, where we'll be providing references to various activities taking part during this World IPv6 Day, measurements, for instance. And we really all run with interest from people all around the globe, which I think is a very good thing.

It's put some spontaneous efforts. Like regional initiatives in Japan, Slovenia, Sweden, Czech Republic, a lot of things are happening, so quite a buzz in the operational community.

So, I think, as I said, as a side effect, this World IPv6 Day, it gives many already?underway efforts a point of focus, which is also a very good thing, and this is just one milestone. It's not that we'll have IPv6, you know, completely deployed after this World IPv6 Day. Obviously not. So there is ?? this is one in the series of providing more incentives, kind of step functions, for further IPv6 deployment.

Well, what we hope to achieve is improve connectivity and large websites are looking into this and measuring how improved connectivity will serve or manifest itself during this World IPv6 Day. We will document all the sites that participate. ISOC will provide a dashboard of participating websites and their status. And, of course, longer termers permanently turning on IPv6 and we hope that many of their websites already have IPv6 on, and some of the websites that will turn IPv6 this World IPv6 Day will turn to be IPv6 available after that. And lots of measurements will take place. We know that RIPE NCC is planning to provide a comprehensive measurements of the Internet during this day, also Arbor Networks will provide some data and we'll refer to this afterwards.

This experience that we hope to collect during this World IPv6 Day, I will certainly serve as an input to improve operations.

And that is the end of my short presentation, short announcement, what World IPv6 Day is about. Thank you. And I don't know if you have any questions. I think we have sometime to address them.

CHAIR: No questions? Well, we also have an opportunity to discuss this a bit later during our panel discussion, so if you think of something as you are going through your e?mail, just keep it in mind. Thank you, Andre.
(Applause)

CHAIR: We have David Freedman, who is also going to be talking to us about IPv6 day, although from a slightly different angle.

DAVID FREEDMAN: Hello, my name is David Freedman and I work for a company called Claranet. I am going to be talking about the implications of World IPv6 Day for access networks and just before I start, I want to describe what I mean by access networks and what various types there are with relation to IPv6.

Having a think about this, I spoke with a number of people. And we have realised that there are actually four classes of access network. The first class is where you have production quality generally available IPv6, and you have it fully deployed, you have it as a product. Anybody can call you up and ask you for it and it's there. The second class is perhaps you have a partial deployment, perhaps people can ask for it; certain products have it available, others don't. There is a limited deployment. And the third class is where you have no IPv6 at all and your customers are relying on transition mechanisms that they may be using perhaps without their knowledge. And then there is Japan, which is slightly odd and out of scope for this presentation. And if you'd like more information about Japan, please feel free to speak to Randy Bush.

So, when is World IPv6 Day in the region in which the RIPE NCC operates? Well, World IPv6 Day runs from midnight until 23:59 UTC, and if you look at this small graphic here, you'll see that the countries in the region are around the UTC's time zone plus a couple of hours, at the most. So, think about your user base. Is it a mix of residential, business or both? And what will users care about during the day and what will the day look like and how will they experience brokenness? If you think about it what's going to happen? It's on a Wednesday, and Wednesday is a working day and there are going to be people that wake up in the morning and a lot of them may not use the Internet or they may use the Internet perhaps on a smart phone or other mobile device. Ultimately, they will probably go to work and spend most of the day at work and not get a chance to really use the computer and use the Internet until they get home.

I think I have to ask in this situation, what would my mother do? Well, she'd probably call me. I know that. But let's just imagine I didn't answer. What would she do? Probably reboot her router. That's the first thing. Whenever something doesn't work, a home page is set to Google, as far as she is concerned. If Google doesn't appear when she launches her browser, the Internet is broken. She calls me. The first thing she knows is to reboot the router. That usually fixes most problems. Let's imagine it doesn't. Then perhaps she'll run an anti?virus scan. How much time is this going to take? An hour, maybe. When she finishes the scan and it still doesn't work, what is she going to do? Head to bed. And then she'll get up in the morning and she'll call me and she'll say, "David, was the Internet broken yesterday?" And nothing will have happened. We will not have noticed. She would not really have cared. It would have been one short day, she would have spent a few hours, and that would have been it.

So, what sort of brokenness are you going to see? There is two real types. There is the issue of fake connectivity, where you get a well?meaning or ill?meaning RA, pages appear to time?out while it flips back to v4. Then there is the bad connectivity, so you have got bad tunnelling, things in the way. Then you'll get things like broken page loads and images won't load properly and the experience will generally be bad. Either of these two, my mother will look at this and say, "There is a problem with the Internet, it's broken."

So what can you do? How can you prepare? There are six main areas you can focus on. And the first and the most important of which is documentation. Writing everything down. Writing your processes down. For you and for your customers. Then you need to publicise this, both internally and externally. Everybody in the company should be made aware. You shouldn't have escalations, people calling you, there shouldn't be support people that have to raise issues about this on the day. Everybody should know. Your customers should know in advance. You really need to publicise the fact this is going to happen. It's also a good idea before the day to just take a look through your flow data. Look for protocol 41 traffic. Look for Teredo traffic, which users are going to be affected? Can you contact these people? How are you going to know?

And also the people doing auto?tunnelling, where are the relays? Where is your traffic going to be going on the day? How many of you that don't operate 6to4 relays know where your closest ones are. I can tell you, for a lot of people, there are only very few of these in Europe that are public, I think a handful, I can think of off the top of my head. Where is it? Is it your competitor's network? Set up a triage system. Look at, perhaps, ring?fencing some support staff that can focus on the support processes you have written up on the day. Use an IVR system, such as press 1 if you are having problems visiting websites today. And also promote this on your music on hold. So the messages that people lis tone while they are waiting have a recorded message that says today is June 8th, world IPv6. You might be experiencing problems. If you are experiences these sort of problems, press 1. Integrate the whole thing. Finally, collect information about how to work with operating systems [](froze) pieces of CPE. How many of you know how to turn off transition mechanisms if they are causing a problem in all the versions of the operating system your customers are are going to be using. You know, for instance, Microsoft have this great knowledge base article that they are readying up for the day which is already out there now, and contains a link to these, fix it, which is you ?? basically, it downloads an installer for your type of operating system which, for instance, turns off Teredo or turns off 6to4 and it's a one?stop?shop, you can send the customer to that URL and it turns them off for them, you don't have to talk them through the processes of going through all these steps and configuration and turning this stuff off.

I came up with an example process flow chart that we are going to use when you have an incoming call. And the idea really is that if you have prepared people well enough in advance and you have publicised this, then you shouldn't be getting the call in the first place. Let's just imagine you do. The call comes in. They get the music on hold. The music on hold perhaps directs them to the site where they can do some self?help and they terminate the call. Imagine they carry on. Who do you want to speak to, press 1 if you have having problems. It takes them through, if they press 1 or don't press 1, they go to standard support, eventually they are pulled through your dedicated team. They run through the series of tests, what does it say? Can you actually reach Google via IPv4, right this is definitely an issue. Then the advice everybody is giving, just upgrade everything. User knowledge?based articles if you can. Upgrade your operating system, upgrade your office suite, upgrade your browser. If that's not working, perhaps you are getting auto?tunnelling from the CPE. Learn how to turn that off. Build up a cache of information about various types of CPE, how they work, where these options are. All of this stuff is going to be publicised. Finally, if this doesn't work, learn how to turn off the transition mechanisms, send people to the knowledge?based articles. Okay. The ultimate goal is to get a happy or fixed customer at the end of this. Also think about how much time you are going to spend fixing these customers versus where then they are going to call in. If you think back to when I talk about what the business day is going to look like, the customers are are probably residential customers are going to be calling in the evening. Are you actually going to offer 24?hour support. Are you actually going to be able to field these calls and what are the call volumes going to look like when they come in, who are are you going to have working on them and what times are you going to need to deploy these people if you take the calls at all.

To wrap?up: Don't bury your head in the sand, regardless of your size or IPv6 maturity. Your users are likely to have issues if you provide Internet access. As I say, have a plan because having a plan is better than not having a plan. Telling your users in advance, helping them to prepare, is going to prevent those calls coming in which is going to save you time and money. And just, you can't sit back and relax. This is going to come at you, if you provide Internet access, this is going to happen. You may only get one call, but if you don't know what to do when you get that call, it's not very good.

And that's the end of my short presentation. Are there any questions?

CHAIR: Questions?

AUDIENCE SPEAKER: Jan, I know I should be on there with you, sorry about that. I suggest we make a poster out of the last slide and send it to all LIRs, so they can stick it on the wall in their help desk, okay.

AUDIENCE SPEAKER: Question: Would it make sense to actually have music on hold reference before the VIX day and have a reference to test IPv6.com in there so people can test and open a ticket before IPv6 day?

SPEAKER: This comes down to the publicity and how you want to publicise it. You can put it on your site on your music hold, you can do a mail shot to customers, whatever. You can ?? you know, there is lots of ways you can do this appropriate to your business. Getting people prepared is going to prevent these calls coming in in the first place, and, the less calls you have on the day, the happier I am sure you are all going to be. Lorenzo?

LORENZO COLITTI: I just wanted to echo a point that's perhaps not completely explicit in the slides. You really made it already. But IPv6 day is not about turning on v6 in access networks; it's about trying to get those broken users to understand that they have a problem and fix it so that we can all get on with deploying IPv6. And it's the users, manufacturers, it's the browser manufacturers, so really the most important thing is to be there for those users to help them figure out how they can fix the problem and the more users are fixed, the easier it is for us to go ahead and publish AAAA records.

SPEAKER: Yes, very important point. We are not asking you to go and turn it on specifically for the day. If you want to, that's great, please make sure that you know what you are doing, I am sure you all do. If you do so and you manage to break it, then you are going to cause people problems.

AUDIENCE SPEAKER: My advice was definitely don't rush the deployment just because you want to be ready on the day, because if you rush the deployment and it doesn't work, then we all get to ?? when the press gets to say, oh, this v6 thing doesn't work, we shouldn't do it, it's a bad thing, it's irresponsible. This is not what we want, right.

SPEAKER: But also understand the implications of customers asking you, so can I have native IPv6 for this day? Do you support it?

AUDIENCE SPEAKER: Benedikt Stockebrand, freelance IPv6 guy. In a previous life, I have been working for large scale and user ISP in Germany. This is one thing I really want to advise people working with this sort of customer again and that's just trying to deliver IPv6 to a couple of million customers at that particular day because it will just blow up your first level support anyway.

Keep in mind there are people who use both the Internet and e?mail. I suppose you know these sort of people. If they run into problems and they call up the first level support, that can be a complete disaster. If you are a large scale ISP, especially with end users, try to somehow deploy IPv6 incrementally to a reasonably small number of of users at a time, and take particular care about your power users, because those are are the ones who will show up on June 8 and they want IPv6 by then somehow but they are much easier to work with than your unsure mother or father. So, don't rush it at a large scale. It's a good thing to have a number of websites and services coming up with IPv6 but success is an entirely different business, especially with private end users.

AUDIENCE SPEAKER: Martin Leavy, Hurricane Electric. So, don't underestimate the amount of v6 traffic that could be generated on world v6 day. Probably in the next couple of days on some mailing list close to home, some stats on a one?hour test will get published by one of the original players that was right at the beginning of the announcement, and the traffic levels are non?trivial. That's for a v6 guy like me, a really good thing. [](froze) the focus, which is excellent, Dave, on the failure side, should not under?shadow the success and that, you know, I am giving a ra?ra here but that's what I am going to do, it's a ra?ra to all those people that thought up this day and got a lot of people on this band wag. That is not a culmination but just one more step on the work that we have done. Just a simple ra?ra statement. But let's look on the positive side of this as well as fixing, quite correctly, some of the on day issues that will go there. There will be real traffic, and it will, I think, surprise a lot of people, which is a good thing. Thank you.

DAVID FREEDMAN: I think Jan and then Alexis.

AUDIENCE SPEAKER: I have one question about IPv6 day. What do we expect from this day? Should we make it look good or should we try and find failures and try to fix them later and make it look bad? What do we expect from this day?

AUDIENCE SPEAKER: Lorenzo. As somebody who was involved from the beginning, I think really we made it clear, right, this is a test. We are sort of inflating the lifeboat that's supposed to get us, to take us away from v4 Internet which is sinking and we are actually seeing, like, is the lifeboat going to float? Okay. So ?? and so, I think we should just look at the data and, you know, report on what we see. I mean, if the Internet melts down, it will be bad, because it means that, you know, this v6 thing that we thought was going to work, is not just going to work, or we have to fix lots of things, and so my personal opinion is that not a lot will break, there will be some breakage. We have pretty reliable data on that, maybe some other things will fall over under load. Some other things will be deployed in a hurry and they'll fall over as well. But I think it will be you know mostly just another day in the life of the Internet, right. And if that happens then we can say okay, it it was just another day in the life of the Internet and you know it wasn't so bad after all. And so, I don't think we should go out of our way to spin it in any particular fashion. I think we should look at the data and be reasonable and if we really get to a place whereas we expect, 99.97 percent of users don't notice, then we say, hey, 99.97 percent of users didn't notice, that's good. So ?? and I think that will be huge if that happens, it will be huge in injecting confidence into the people that are very sort of worried about this and very sort of wary about we can't turn it on, it's going to break, we can't publish AAAAs, so I say just look at the data and report objectively on what we see, and I am sure that the PR departments of various participants will be putting out whatever message, you know, they want, but I think, as a technical community, it's our job to sort of present this objectively.

AUDIENCE SPEAKER: [Alexis] from Finland. I just wanted to congratulate you on a very constructive and positive, good presentation without any whining about how this doesn't work and how that doesn't work. Because we have had a dozen of those this week. Personally, if you take any given time, at least 0.03 users, percent of the users on the global Internet, have basic problems with their basic IPv4 connectivity. So I am not that sure we should really even be worried about as many users or less users having problems with IPv6 connectivity.

DAVID FREEDMAN: Interesting point. Rudiger?

AUDIENCE SPEAKER: Rudiger Volk. Well, okay, David, first, thanks a lot for this presentation. I am extremely happy that we have this kind of information put up in public, at least at this point in time. It would have been much preferred if we had started to discuss this somewhat earlier, because did everybody hear and understand Lorenzo telling us, well, okay, he is going to attack the actual network operators? And by the way, and by the way, and by the way I remember, I remember Lorenzo telling us that half a percent of users getting brokenness is something that makes him not to enable a service because that's not sufficient. The remaining half percent of customers that might get a problem is, well, okay, 5, 6, to get numbers for some people. Well, okay, who has help?desks to deal with that?

Well, okay, sorry, I am complaining and shouting often and frequently around here. What makes me actually almost most concerned is that, well, okay, there is kind of a mixed message about the whole thing. Andre was telling us, well, okay, this is all glorious and lots of things will happen in addition, and well, okay, the marketing departments love that and start to do showcases and stuff, and it appears that there is indeed a danger that a lot of the operators, in fact, miss the message that there is an attack coming, an attack that, well, okay, exposes problems that have been around and not taken care of for quite sometime, but to a large extent, also, due to the people who are unleashing the problems on that day in a focused manner have been protecting themselves and the rest of the world are from those problems for most of the time.

AUDIENCE SPEAKER: Lorenzo: Let me take that. As regards, earlier, the 0.03 percent of users have problems with their Internet connection at any given time anyway, the response from our site reliability team is that yes, but those new 0.03 percent of users have problems all the time and they don't know how to fix them and they only have problems going to Google. So, that is why we are careful about this.

As regards the attack, well, I disagree that this is an attack, and, in particular, I think that you have two choices and you can do either but not both. You can choose to say that this is an attack on your users, or you can choose to say that we are dragging our feet on v6 deployment. You can choose one of those two. You can say Google is not deploying v6, they are slowing down the Internet, or you can say Google is deploying v6 and attacking our users and that's bad but you have to choose one of them but you can't consistently say both. Okay. So that ?? I think ?? I think it's not really a vary assessment to make.

And finally, as regards to ?? Google is protecting themselves, they are withholding AAAAs and how they are attacking us, it is very different for everybody, I think, to have ?? to expose a small percentage of users to breakage for one day, that's why we publicise six months in advance, and to expose users to breakage on an ongoing basis. I think that's very different.

RUDIGER VOLK: Well, okay, Lorenzo, let me clarify my intention. I am not really that interested, and it's not so much fun to try to blame someone for something very specific. What I am trying to make very explicit is that someone is deliberately creating a singularity in the operation of the network globally and that usually has to be dealt with as an attack, and well, okay ?? and not just kind of as a neutral threat of vulnerability exposed, it's deliberate. The core operation that Andre was praising in high voices to deal with this adequately in advance certainly could be improved and, well, okay, I am making noises to help make sure that people who will see the singularity actually are aware and not lulled by having the marketing departments using nice showcases.

AUDIENCE SPEAKER: Lorenzo: Agreed. That was also the reason why we tried to give as much of a heads?up as we could. At the last RIPE meeting, we, in Rome, I was on the stage and I said, look, this may be happening, we don't have any commitment, and somebody from the floor asked what is the commitment, I said there is no commitment, but I said, look, we are thinking about doing this heads?up. Of course, there is lots of things we could have done, but I do think that it's ?? well, I do think that it was, you know, it was conducted in a responsible way. That's all.

AUDIENCE SPEAKER: I have a question on Jabber, unidentified user. What will happen on June 9? What are the failure rates on that day and will you instruct your staff to deal with those, too? I am thinking still of AAAAs and the configuration that is around?

DAVID FREEDMAN: I think what we anticipate is for people to either remove the AAAAs, from which I understand Google will say they will certainly be doing that for the main website. Please, somebody from Google correct me if that's not the case. Others may choose to leave this on. Once we have the support processes in place, depending on how the day goes, we may or may not leave the triage team in place, we may leave the IVRs and MOH in place. It's all going to depend on the day. I don't think we are going to get into the situation where we are going to want to spend lots of money ring?fencing support resources for this. We shouldn't have to. But once we have got the experience of doing this, I can't see it being a problem to leave some resources in place for the next, for the following few days, if people are going to leave their AAAAs in place, I have no objections.

AUDIENCE SPEAKER: Benedikt Stockebrand again. About considering this day sort of like an attack, yes it is. Basically, what we are doing is we are kicking certain people where they sort of hopefully notice before something worse happens. So, yes, in a way it is an attack but it's only to prevent things getting worse. It would have been much easier if we had done this like three or four years ago, but then we wouldn't have had this necessary support from a number people like Google, for example. It is a good thing we do it this way and it gives the ISPs especially a chance to deal with this situation on a fixed day, just like you mentioned, get everything in place, get the support beefed up. This is really what we need. And we do have a singularity which is ?? we are running out of addresses. That's the actual singularity. This IPv6 day is just kind of a side effect on that.

NIALL O'REILLY: A small comment. Niall O'Reilly. A small comment on the singularity metaphor. We are not dealing with a point singularity. If we were, we could navigate around it. We are dealing with a fault line and what we are doing on World IPv6 Day is building a bridge across that. Whether the bridge will hold the load, whether it will be just business as usual or what, is what we'll find out and we'll come back, sooner or later we are going to have to build something to take us across that fault line, and the sooner we start, and this is a good start, I think the better.

AUDIENCE SPEAKER: Alexis again. Talking about attacks again. I'd rather call this a counter?attack than attack, because the operator community's refusal to adopt IPv6 for the past decade can be viewed as an attack to people like Google or Microsoft or other innovative companies out there who have IPv6 products lined up just waiting for it to happen.

RANDY BUSH: I was not going to speak, but, Alexis, bull shit. IIJ has had IPv6 turned on for its customers commercially since 1997. For IPv6 day, we are going to have to turn it off. We are going to have to block AAAAs at our consumer customer facing DNS servers. Somebody shouted why? It's brilliant. This is true, by the way. This is the likely solution, that all the consumer ISPs are are going to hit. In Japan, the structure is stratified, the layer 2 and below providers, NTT east and NTT west who essentially own all transport are not layer 3 providers. So, in my home, when I plug into NTT's CPE, I get a v6 address. It goes nowhere outside of NTT's TV and VoIP network. It's a dead end. So, if Google hands me an AAAA, I am going to bang it against my TV set. How do I get out of this? Tough question.

AUDIENCE SPEAKER: Alexis: Actually, I understand that NTT has been working on distributing address selection policy by either HD CP v6 which I understand they are using on those lines. So, if you accept the HCP v6 and use that address selection policy you should be able to use the other prefix you are receiving from your own network.

RANDY BUSH: You understand future science fiction. These are are the same people that told you IPv6 was widely distributed in Japan. It isn't and never has been. There is more IPv6 in Europe today than there is in Japan.

AUDIENCE SPEAKER: Eric. With respect to the changes that only applies to things that actually have a 40, 44 table that is anything not Mac at all. Let alone can receive updates and can make policy changes which is basically almost anything other than say, Windows 7, or may be Vista, your phones, whatever sort of device I says ma may be able to do IPv6 on the Wi?Fi, do not have the ability to receive 3044 policy updates in any fashion whatsoever.

CHAIR: Thank you very much everybody.

(Applause)

At this point, I'd like to ask a few people from the community to come up and have a seat behind me here, so this is Eric, and Jan and Andre and Randy Bush and Mischa, if you'd all come up here. We'll get started here.

Now, we'll just spent a little bit of time talking about IPv6 world day. I'd like to you cast that out of your heads. We can talk about IPv6 world day now too of course, but the idea for this panel discussion is to talk about all the v6 topics, especially ones that came up during this week that we find interesting. So, if you have seen panel discussions in the past, you may have seen people give a short introduction of themselves and give maybe a short slide presentation or that. We are not doing that here. I find that incredibly boring so I am vetoing that. Hopefully we can get straight into some interesting discussion.

So, if people in the audience interest things they want to talk about than that's certainly appropriate. I am not here to control the discussion or anything. Neither is the panel up here. We are just here to kind of facilitate, have an interesting conversation, maybe recap and clarify some things that came up and hopefully it's interesting for you guys.

So, I am not really sure if any of you have anything in particular you want to start off with. If not, I can ask a question. We will do a brief round of introductions.

ANDREI ROBACHEVSKY: I work for ISOC, and I guess I got on this panel because I made the presentation about World IPv6 Day.



ERIK KLINE: Eric Kline, Google, I work on IPv6 sizing, a lot of internal software.

JAN ZORZ: Jan Zorz, Slovenia. I am probably here because of 501 stuff.

SHANE KERR: And because everywhere I look for v6, there you are.

MISCHA PETERS: Mischa Peters, engineer for A10 Networks.

RANDY BUSH: Randy Bush, Internet Initiative Japan. Sort of ops, sort of research, half?assed job of everything, excuse my rudeness, I have a down server.

SHANE KERR: If you need to use the projector for that, we can follow along.

RANDY BUSH: I'll take you up on it because I bet you there is somebody with better FreeBSD out there than I have.

SHANE KERR: Well, I guess maybe this is a question for operators, someone asked me, brought this up early year during the week and we had our v6 exhaustion so, we are out of v4 addresses at IANA, we are not quite there yet ?? has anyone actually seen anyone effects of that on their networks or on their processes or in anything at all?

JAN ZORZ: I have seen a change in my schedule, in my calendar.

RANDY BUSH: There are less slides of the Geoff's classic v4 is going to run out. I have seen less of them. So, run out is good.

SHANE KERR: Geoff, do you want to respond to that?

RANDY BUSH: He has got a nice new slide set which shows by the way estimates of when different RIRs are going to run out, so we can have a new form of fear and panic.

SHANE KERR: I guess maybe in our region it hasn't really hit home yet because we haven't had our last /8 yet. I guess in APNIC that happened a few weeks ago to it's a bit more real in their region than it is here.

RANDY BUSH: No.

JOAO DAMAS: So regarding effects, yes, actually there are quite cruel but yet unconfirmed that is from the provider itself, information come out in Spain, for instance, that at the beginning of June, that one of the big ISPs over there is going to start using CGN for their new customers and using as reason the fact that he can not get ?? never mind that he can still ?? that's not, apparently, what the real ?? it's being used to curtail service.

SHANE KERR: When you say CGN, you mean they are giving out private address space?

JOAO DAMAS: The users will no longer get access to all the ports in the public address that they get. They will get a public address but they won't be getting access to the 65,000 ports any more.

SHANE KERR: You are actually sharing an address with another user.

AUDIENCE SPEAKER: Mark Townsley, Cisco. One thing I have been seeing it more publicly available evidence of IPv4 address trading.

SHANE KERR: Okay. I guess there was some news about that in ARIN region especially.

RANDY BUSH: Two things that are actual data points. One is, of course, the big Nortel Microsoft noise. The other is John Curran published on the NANOG mailing list, maybe on the order of a week or ten days ago, the change in the rate of applications to do a transfer. And there is a good spike at /24s, interestingly enough. And wondering what NCC is seeing? Is NCC seeing A) a big grab for the last of the v4 space? You know, an upswing in requests, and how is that distributed across to ISP size? And secondly, are you seeing requests for transfer changes?

AUDIENCE SPEAKER: We have not seen an increase in the transfers, any requests for transfer of resources so far, and with regards to resource requests, it's the growth table as it has always been. So nothing particular, no extreme spikes.

SHANE KERR: So no panic here yet.

AUDIENCE SPEAKER: Apparently, not yet.

SHANE KERR: Okay. All right. We can move on maybe to something else. So, I looked throughout this week to try to find v6 topics that might have raised a lot of discussion, and one thing that raised a lot of interest and discussion, which surprised me a little bit, was the 501 BIS work. The reason that it surprised me is there is probably as many IPv6 recommendation documentation as IPv6 users out there today. There is a huge number of resources available, you know, going back 15 years or so, about how to do your v6 migration, and all this stuff, and now we have a document coming out ?? being published in the RIPE community, put together by Jan and Sander and, all of a sudden everyone, is jumping on it. And I don't know, Jan, why do you think that this document has gotten so much interest when there is already all these others?

JAN ZORZ: Because it's not 2000 pages long. And it tends to be practical and, you know, where where there is ?? built inside a big company or enterprise, those procurement guys, they can't read RFCs, they don't want to read 2000 pages documents. So I have been told so. They finally managed to find something that is short, fairly short, usable, they can do copy, paste, and be done with it, and they, hopefully, can request and buy the equipment that does v6 somehow.

SHANE KERR: Okay. I think that seems fair to me. Maybe I can get a quick show of hands from people in the audience to see, have you used this 501 document? Have you passed it on to your vendor? If you are a surrender has somebody come and give this document to you? A few at the back. Okay. I know Jan has got a lot a direct feedback from people.

JAN ZORZ: I recently learned that it's also advertised by APNIC, RIPE 501.

SHANE KERR: That's interesting.

RANDY BUSH: It's gaining legs in the NANOG region, too.

AUDIENCE SPEAKER: This is Mark again. And I hope, you know, 501 clearly exists, and if the RIRs are, you know ?? so we have, USGV6, and, as a surrender, we pay a lot of attention to that. That's the US recommendation because it's directly tied to RFP. First thing we looked at when we saw 501 was okay, can it fit in as a subset? Because, what we don't want to have to do, it costs money to actually verify and qualify and make sure that we can put whatever little logo is on there and that takes away from all the v6 resources, we don't want to have to pay attention to a does of these, right. And I understand you couldn't just like adopt a DOD recommendation in the operator commuter whole hog, 501 exists, it seems to be very well written and I think it's great if the other regions take it, because I don't want five, right. At some point, we won't ?? it becomes self?defeating because there is so many, right.

SHANE KERR: The great thing about standards being there is so many of them.

RANDY BUSH: Oh, my God, you mean the five RIRs are actually going to do something the same.

SHANE KERR: Let's not go crazy. Okay, well, that's actually, I guess, a good recommendation. Did you already have plans to coordinate with the other RIRs with this document?

JAN ZORZ: Not yet, but apparently ??

SHANE KERR: Apparently, now. Okay. Moving right along here. This is a question I think mostly for the audience. As I mentioned in the introduction to this session, we didn't do an IPv6 Working Group this week. Do people miss that? Is this a model not only for the v6 but also for the other Working Groups? Because we are trying to kind of look at the idea of not meeting all the time if we don't need to. So would people have preferred that we had an IPv6 Working Group this week? If no one says no, we may never meet again.

JAN ZORZ: Well, at worst, and I am speaking because I really felt the need of it, when I am doing this, when we are doing this 501 thing, we need feedback from IPv6 Working Group, because it's published in v6 Working Group. So do I ask questions to a Plenary or do I ask questions to v6 Working Group?

SHANE KERR: Right, yeah, that makes a certain amount of sense because documents are really Working Group group items, but not for the whole community. It's a work item, yeah.

RUDIGER VOLK: I am not a regular attendee of IPv6 Working Group meetings; there is usually a lot of other stuff going on. This time, actually, I had hoped that a smaller circle doing a little bit more focus discussion in preparing operations for IPv6 day.

SHANE KERR: Okay. Is that Hans Peter?

AUDIENCE SPEAKER: Maybe you should even take the step of removing v6 into the Plenary because that's the hottest topic, to move v4 legacy Working Group.

SHANE KERR: That's not as crazy as it seems. Marco, do you want to give some of your crazy ideas about this or are are you not ready to go public yet?

MARCO HOGEWONING: Speaking as one of the co?chairs of the IPv6 Working Group, yes, there is this crazy idea now I have been sharing with with people in the hallways about this, yeah, crazy ideas maybe in Vienna to schedule, well, working title, the legacy protocol taskforce together with a group of people and actually focus with the future of legacy protocols, what's needed in the sense of maintenance there and actually have a v6 hand?off to the real world because it's deployed by now. So we start focusing on what's needed regarding DNS registration and all basically the redistribution of the leftovers of v4 or returning space and give v6 to the rest of the world to play around with.

JAN ZORZ: So you are thinking of IPv4 switching off Working Group?

MARCO HOGEWONING: No, let's call it maintenance, because I think, for the next 20 years, there will be a need for some guidance on the coexistence between v4 and six, but as v6 is getting more and more mainstream, the need for a specific Working Group on this topic might disappear and it might flow into the other Working Groups, like DNS and Address Policy, to take care of what's happening in v6.

DANIEL KARRENBERG: Let's just remind ourselves that RIPE Working Groups don't have to meet at every meeting. There is a mailing list and they can work between meetings. If it's about ratifying 501 BIS, that's better done on the mailing list than at a meeting, so if the question is should we meet again, yes, maybe, but you don't need to wait for the next meeting to get some work done.

SHANE KERR: No, clearly not.

MARCO HOGEWONING: I was basically about to make the same comment, please post a draft text of 501 BIS as soon as possible to the mailing list, that people can have a look at the text and comment on it if you are looking for feedback, I think that's the best way ?? it's an extensive document, it takes some time to read, and make up your mind.

JAN ZORZ: Well, we thought of doing it in a different document, but, on Tuesday, we received the comments from the community that the community wants the 501 altered and modified and publish it under the same name but under a different number, apparently. So, we will have to do some work and implement the new document into old one, so, probably, it will take some time.

MARCO HOGEWONING: I can understand, but be aware the community is bigger than those people in this room. There are other people out there who are not making it to a meeting or not watching the webcast and might still be interested in it.

JAN ZORZ: Okay.

SHANE KERR: Wilfried.

WILFRIED WOEBER: Wilfried Woeber, one of the co?chairs, and usually being faced with the decision to split myself in two halves with the parallel stream. My perception this week was it was one of those really nice situations where the IPv6 Working Group had enough time and didn't have anything to run against it.

SHANE KERR: That's true. I also felt the same way. I am often torn with the split streams. So I was glad to have it in the Plenary, but I also realised that a lot of other content didn't make it to the Plenary this time because of that.

AUDIENCE SPEAKER: Alexis. It occurred to me that since this 501 is a very good document and if there are ?? in some countries, there are some sort of authorities that have an interest in promoting IPv6 adoption, I suppose one tool for them might be to promote the document or even send it out to certain types of businesses.

SHANE KERR: Yes, that makes sense to me. Okay.

RANDY BUSH: I think it would be a little droll to disband the v6 Working Group right as we are about to actually maybe hit an inflexion point in the deployment of IPv6.

SHANE KERR: You disagree, Joao?

JOAO DAMAS: While we are on the topic of documents. 501 is great for ISPs to talk to their vendors. When they want v6, would there be scope for there to be a document on ?? sort of lightweight document on a review of the implications of the different things that vendors are trying to sell to ISPs and their consequences on the network and the user experience, so CGNs, and what if you go 6RD versus actually using native v6 and this sort of overview, possibly with references, it doesn't need to be exhaustive, as a guidance for the people who get bombarded by their vendors, especially to help the engineers, because vendors are smart, they go to their managers and then the engineers get stuck with this shit. So, something the engineers can then use towards the upper layers, and then, separately, something that would help out the users in figuring out why their users experience is worsening, and use that, something they can refer to in firms, you know, users tend to sell ?? in providing, perhaps, guidance on which ISPs suck less because of what they have done in this time.

SHANE KERR: So, I guess, as far as a document, you are kind of talking about the document version of a few of the presentations that we have seen this week, which makes a certain amount of sense to me because you can't sort of slide ware to your boss because all you see is these slides with some pictures and they don't actually mean anything.

JAN ZORZ: We need to make sure this document doesn't look like an attack to vendors.

Just to clarify, at A10 Networks we provide equipments to do IPv6 migrations and we actually look at documents that are around and I already forwarded the 501 document to our engineering team so we can have a look at it and see what our customers are looking for.

SPEAKER: I wasn't against it. It was more about the choices, or the consequences that technology choices imply rather than vendor bashing, then everyone can use it.

MARCO HOGEWONING: It's Marco. This time speaking as in the RIPE NCC training department, we are internally building with some documentation on guidance on which technology to choose, that will be published in the near future. It's almost done. And this is done as an internal stuff at the NCC, so it probably won't be a RIPE document. You probably see traces of it appear on Labs, but actually is important, and while doing this, of course, we can read the standards and, of course, we can read the manuals and understand what's going on. What I am really interested in is actually feedback from people who have deployed all these fancy transition techniques. So far, the only thing I have seen in the world that has a substantial amount of users is the classic 6 and 4 tunnels using stat EIX tunnels, and there is this huge, basically, 6RD deployment in France from all the other stuff that's out there. And we see little or no traces of it actually being used in the operational field and it's really hard to provide guidance to people, and I also see this on a day?to?day work, while doing IPv6 training to LIRs. Every wants to know what should I do, but we can only offer them the theory and it would be so nice if somebody can actually deliver some practical statistics on what's used, how good is it really if you deploy this to 10, 20,000 users? It's a completely different environment than testing this in your lab with with 10. Also, usually you test with more text of of users that know what's going on. And as Dave already said, what happens if you shove this technology to your mom? Who is she going to call?

LORENZO COLITTI: Lorenzo. I definitely agree that this would be useful actually. Thinking about it, there are some things underway in the IETF but the time lines there are much longer there, so ?? and so, and there is, for example, there is a draft on impact of large?scale NAT on the Internet, and that took a long time to get done. Even the theory would be, I think, would be better than nothing. Even saying, writing up a comparative document saying these are are the options, there is 6RD, there is Dual?Stack Lite, there is native v6 if you can do it, and saying, you know, for each one, pros, cons and harbour blockers. If you approach the transition, I think lots of people just get lost in all the options, right.

SPEAKER: What we notice as a more core vendor is that the problem currently lies more on the CPE side than on the core side.

LORENZO COLITTI: No, it doesn't.

SPEAKER: We are seeing it does, more the DS?Lite size. Large?scale NAT is easy an it's not a migration technology.
Lorenzo, we agree, and, therefore, we need to document, you know, the harm that it causes and, you know, the drawbacks that it has. I think the IETF has done a good job on that already. Although it a very long and lengthy document.

JAN ZORZ: I think we need that A plus B stuff out as soon as possible.

MARCO HOGEWONING: Well, I think also author of the CPE survey and CPO support is getting better. A lot of vendors build stuff but nobody is test it go and that's the feedback I keep getting from CPE vendors, maybe mark can add something to it. They put it in there but they need back on what's actually working and not. They can't just write bug?free code. Everything here who has any background in engineering know that you will encounter problems in your first releases and you have to cope with those, but if nobody tells you where the problems are, you are not going to fix them yourself and find them and feedback is the most important part. Tell us what you are using. Tell us what actually works and what not and people can focus on the problems.

LORENZO COLITTI: Feedback is blocked. I have a CPE at home. I am one of the two users in ASN fan Cisco, or maybe in the US that that is v6 from Comcast. The other one has broken v6, by the way, so she has to disable v6.

SHANE KERR: We have some engineers that have Comcast v6, there are not just two of you, maybe three or four.

LORENZO COLITTI: My special status is degraded. Drat. I do have a CPE at home. It's on the residential user.

SPEAKER: You are talking about native v6.

LORENZO COLITTI: And the CPE works when it does native risks, does it work? Hell, I don't know. I mean, has anyone deployed the dual stack? I can try one in my lab, but really it's, you know, what I guess what I am saying is that some access network should run some trial that actually ?? but yeah, but we have been saying that. But I do think that even in the trials, I think the theory is better non nothing. Listing the options, because you go to a network and you say you do need to do v6 and they say but what do you need to use, Dual?Stack Lite, 6RD, native? I don't know, I think it would be useful. I have this thing about short documents and I think that you know if it's too long they won't read it and so, just comparative summary or something.

SPEAKER: In a couple of months time we probably have some information on that on Dual?Stack Lite, for example, and 6RD. We are rolling out already some field trials or production even with with large scale NAT, but...

LORENZO COLITTI: I am talking about just a comparative listing of the options. These are the things you can do. They are on the table. You can say 6R is better, or whatever, you can say all these options and it's good for you if you have, for example, Dual?Stack Lite is good for if you you have a bullet?proof v6 network that you know, you can rely on to carry all your traffic. 6RD is good for you if you don't have a D slam that's not upgradeable until 2019. Somebody has to put them in one place that's easy to find.

MARCO HOGEWONING: That's exactly what we are trying to do. And it's a hell of a job. We tried sessions on white boards and we ran out of space almost instantly on the amount of options you have. As far as it goes about testing, I really ?? I come from a commercial background. I know there is this drive to keep your information to yourself because it's a business advantage and you have to realise what you gain by actually publishing data if you test stuff. I know people are testing it but they don't tell they are testing it. Please if you have test results, publish them and have the community benefit of the work you are doing and have us all benefit from the work you are doing, publish, I think that's the key. There are a lot of people sitting on the information where they don't publically say what is happening.

SHANE KERR: It's kind of a new paradigm, it's a way of thinking about things for network operators. Normally, you don't go to your lab with a new roll?out plan, do a bunch of testing and then publish all that, right. You use it. So, I guess you are calling for a slight change of thinking from people. I mean, you can do both. But it's not something that you just normally think of as part of the way you do a network roll?put. And also, as far as the documents, it seems like there is some interest in this, so, maybe we can have a few people talk about what would be necessary to actually get the work done of creating such a document. I am thinking of maybe we can talk to some of the people who did presentations earlier this week and asking them if we can use their slides as outlines and just flesh it out a little bit into relatively short documents. We might be able to come up with something to answer that.

AUDIENCE SPEAKER: Alexis. In Finland, we have a couple of operators, very small operators that have on the order of hundreds of thousands of IPv6, domestic IPv6 deployments over DSL, and there has been an effort on the part of a lot of operators, both small and large, to deploy IPv6, but the problem hasn't been only the CPE, because, in Finland, there are two categories of how you can make, or deploy a connection, and one of them is just make the CPE bridge everything, so then their problem moves somewhere else, then usually it's the D slam that's the problem and that has been a lot trickier to fix.

When someone ? I think, Rob, it was you, who said that should be a list of all the options for operators on how to deploy, and that made me think that, basically, the best option for all operators is native, end?to?end, but that usually doesn't work, so what the operator is left with is what works with his or her network. So, I don't think it's really a decision over different technologies because there is usually only one technology for each case with the limitations. And if the operator doesn't know what that technology is, they usually go to their vendor and they tell them okay, this is the technology, and then they read up on it, maybe, and so on. But I am not against any of these lists, but, just, it struck me as funny.

LORENZO COLITTI: Can I comment on an earlier point you made? You said it's a paradigm shift because we need to ?? operators usually is used to doing stuff in their own lab and not sharing it. I would argue that there is a case to be made for better collaboration. If you look at World IPv6 Day, for example, you will see in the list, at the top, the people at the top of the list, you'll find Yahoo, Facebook, Google. These are what you would probably call competitors, even, pretty robust competitors, right. They have, in some ways, the same core business, and the way I think of it is that, you know, getting this v6 done is not about, you know, a competitive advantage; it's about building the new stadium in which you can compete against each other again. Because if we ?? if the Internet falls apart, then we can't compete any more. So the reason why I think you see those operators who are, you know, very, competitive against each other, on the same press release, that's kind of unprecedented right, can you find another press release where you have all those names on them? The reason for that is because we know that this is a shared thing we need to work together to preserve. So I would say, we all need to sort of go to our labs and figure out what's happening and then, you know, get over the embarrassment of the fact that you can't get it to work because there is loads of bugs and then sort of say, okay, well, there is lots of bugs, this is what we found, let's collaborate on it, that will head to healthy sharing of information, healthy sharing of vendor bugs because the classic excuse of somebody or a vendor or provider who doesn't have the resources to deal with the problem is to say nobody has reported this issue before, nobody is asking for it and things like that [](froze), so, by collaborating, you actually get to sort of shine a light on that a little bit and I think there is a case to be made.

SHANE KERR: I am going to be a little rude and take over to answer that, because part of the way the Internet has worked to date has been based on the basically unlimited availability of numbering resources. It hasn't been a part of your picture when you are thinking about rolling out new services or new ways of doing things. With v4 exhaustion and IPv6 not being very well adopted, that's no longer true. So you are looking at a world ?? and Geoff Huston, I think, has mentioned similar ideas many times, where we are looking at a world where new entrants have a really hard time. And that's not necessarily bad for the people that are big players today. So, I am actually really glad that the people participating in world v6 day are but it's not all the big players, right. I mean, a lot of them, I am sure, are not participating simply because they have no interest or don't have any time, but, possibly, part of the reason they don't have interest or time is because, well, they are in a good position right now. So, I don't know. I don't know. I hope that's not true. I just think that v6 is the way for us to move forward into a world where we continue to have new services. I mean, if Google couldn't have gotten a server of their own to host their services on 20 years ago, we wouldn't have Google today, and I would like for those kind of innovative services to be possible. I'd like for that to happen, but I don't actually think it's a necessary outcome. So...

LORENZO COLITTI: If you are an operator in that state you have to have a contingency plan just in case somebody else deploys this v6 thing and it actually works, you'd better be worried, right. Because if you are going to do pervasive NAT and your services suck and somebody else deploys v6 and, by some miracle, it actually sticks and you have no plan at all, you get a competitive disadvantage, right. So even if you are somebody who is saying I am going to do 9 layers of NAT and I am going to force things to go my way, you are then posing yourself at risk if you haven't sort of, if any of your competitors in the market don't see it that way. I would argue that even the people who are going to go NAT 44 all the way, I think that they, you know, may need some of this.

AUDIENCE SPEAKER: Some of us want a competitive advantage. Anyway, what I forgot to mention to Marco about these few Finnish small ISPs, Marco asked if somebody who has deployment experience could come and, you know, document or present or whatever. The tragedy is that they are small operators and they only have a handful of technical people and they are overworked and so on and they don't even have the are resource to send anyone anywhere. So...

AUDIENCE SPEAKER: Jabber: I think the way the amount of different options for implement v6 for end users, 6RD, native etc. Is also confusing for CPE vendors because probably it makes hard to understand whether their implementation is going to work in the real world. That's probably why IPv6 is going to be implement splitted by having direct control by core user NCPs but it's proving extremely hard for the others.

ERIC KLINE: The same thing is true for mobile devices. People come to us and ask for 4 RD and can they also have dual stack in android and 6RD in android. Like, you know, what do we do? There is finite resources to put v6 into the thing. Which way do people want to go and everybody wants to run their network differently as well. So ??

SHANE KERR: What do you do?

ERIC KLINE: At the moment, nothing. We don't even have an IPv6 capable radio image from any vendor for any of the phones. So there is nothing to do yet. I mean it works on the Wi?Fi, that's not ?? I mean, that's ??

SHANE KERR: That's not getting a call.

ERIC KLINE: That's not PDP context and that's not a lot of things.

LORENZO COLITTI: Not nothing, sorry, 2234 supports v6 on the radio interface and the 3G interface but this is no device that ships this. However, ?? sorry, there is. You can take the latest open source code and you can do v6 on a 3G interface. Please let us know what the phone is.

AUDIENCE SPEAKER: The horizon HTC thunder bolt will run v6 over the air, and v4 over the air obviously. It works like a dream when there is LTE coverage in the city of your choice. Repeating what Lorenzo said, they implemented it by themselves. We have got someone at the back mike.

AUDIENCE SPEAKER: We have one more question on Jabber. He says the main question I'd have is, why don't the hardware manufacturers actually start doing some proactive development instead of people waiting to show up with requirements?

SPEAKER: Well, one of the problems there is that there are a lot of drafts currently, and what we are seeing is that no one size fits all, so there are a lot of different requirements for different companies. So, it's very difficult to be proactively implementing something that people don't want.

RANDY BUSH: This is a cycle we have gone through. For those of us who are not old enough to remember the NG Trans Working Group, we are all doomed to relive it. There are more transition mechanisms than there are v6 packets. And most of them are are are useless. And most of them that are labelled transition mechanisms are actually non?transition mechanisms. In other words, 6RD is not a transition mechanism. It is a way for me to give customers, access customers v6 without transitioning to v6.

SHANE KERR: Nice. Yeah.

AUDIENCE SPEAKER: [] moss an /SWAOESey. I have the impression here sort of polarised discussion between ISPs and content providers, and in between there is the poor end user who is asking or not for IPv6 and we can deliver him IPv6. And I also have the impression that there is a critical path between the ISP and the content provider and the end user I also which is the home CPE. I think that's the most challenging thing today to have a massive deployment of IPv6, getting a home CPEs with IPv6. I hear anyone ISPs saying, hey, I want to deploy IPv6 tomorrow, but, you know, my CPEs I depended from that vendor who don't even tell me what his road map. Of course, for companies who have IPv6, they have Cisco, Junipers, everything is fine, but the home net is really, are really poorly documented and when it is, it is of course 6RD, I won't say it's a bad thing, but it's a very good thing to start, so where are vendors and I think that the community should keep putting pressure on vendors for those specific home networks and saying, hey, you are maybe the weakest link in this.

ERIC KLINE: I think it's not great surprise that the edge is harder to update than the core. It's just simple mathematics, but I would agree.

AUDIENCE SPEAKER: Benedikt Stockebrand again. It does mean putting pressure on the CPE vendors. The first thing you want to understand the situation. CPE vendors work on a fairly small margin so, as soon as they start to put extra ram or flash into one of their devices, their prices go up and they lose business, and I mean big time business, sometimes it's a couple of cents that makes the difference between getting their stuff sold so the ISPs or having a competitor take over. So this is a very, very delicate decision for the CPE vendors, when to actually enter the market with dual stack capable devices. Introducing things like DS?Lite which put an extra burden on the CPEs make the situation actually worse. If you want to get the CPE vendors to move to what's implementing IPv6. Well, there is one thing at lease the big ISPs do is order that stuff big time so they have an incentive to develop, otherwise they risk going bust. It just doesn't help to keep picking on the CPE vendors.

JAN ZORZ: About CPE vendors, it happened with my travelling around the world about v6, I had a speech in grease and I influenced Greek CPE vendor to implementation v6 in their CPE. They do all sorts of DSL and fibre object particular stuff and they did it in three months, fully working implementation of IPv6 in CPEs. So, it can be done, and it's being done, so stop finding excuses.

AUDIENCE SPEAKER: It can be done, yes, but it makes the products more expensive and that's their problem. It's not so much the problem of implementing things. If you take a look at ?? in Germany AVM is a larger vendor, they have IPv6 support in their bigger CPE routers, because they have ?? they are big enough so they have a low end sort of product scale and they don't put IPv6 in there because it drives the price up. They only put it in the bigger ones. It is possible and in that case, they can do it, but if we are talking about the low end business mass market, this is a big problem for them.

SHANE KERR: They could rip all that NAT code out.

MARCO HOGEWONING: Co?author of the famous CPE survey. A question to the audience, maybe just do a hand raise. The current CPE server focuses on what's out there. Should we run a survey of what you actually want? One vote of support, I guess not.

ERIC KLINE: I think there should be some discussion about that at some point, Marco. Among other things, I don't think anybody has taken a whole bunch of IPv6 capable CPEs and chained them at home. Like, it's fairly easy to make that work with, you know, NAT v4 you just can't get everywhere, but the route at home and how to make the home work and look in some sort of arbitrarily complicated manner, has not yet been determined and I think it should be looked at. But you are talking about ?? you need to get v6 to these people so they can play with it, then they can decide what they want and see what's actually useful. The rest of it is just sort of blue sky tinkering. You have to have it to play with.

MARCO HOGEWONING: I understand. I was more thinking like, as an operator, you have all these choices and I heard yeah, we want this and had a want that, so maybe we just do it as a guide to the vendors, ask the operators like which technology you prefer. Discussion with vendors, is we only have this limited amount of engineering time and we can only do one or two protocols at the same time. So tell us whether you want DHCP P or PPP and we'll fix it for you but not at the same time.

ERIC KLINE: That is not largely dominated by the access technology.

SHANE KERR: I am just going to draw a line in the sand. We are a bit over time. So, I apologise for that. I am very interested in hearing this discussion continue, I'd like to thank everyone who came up here today very much.
(Applause)
And thank you all for sitting through our discussion here and I am, I look forward to receiving your e?mails on my v6 MX, so...

RANDY BUSH: I got the server back up.

ROB BLOKZIJL: Okay. Thank you, Shane, and panel.