These are unedited transcripts and may contain errors.

Plenary Session, 2 p.m., 1 November 2011:

CHAIR: Welcome back from the lunch. My name is Kurtis Lindqvist and I am going to be the moderator for this afternoon's Plenary Session. We are going to start with interesting short talk by Axel from RIPE NCC on what we are going to do with this legacy D4 space now that we are not using it any more. So, Axel.

AXEL PAWLIK: I am not sure that I can guarantee either short or interesting, but you know...

All right. Legacy space, we all have heard about that. We have heard about it quite frequently in our talks with governments who all assured us that this preRIR space was given out unfairly to the Americans mostly. And shouldn't we do something about this and anyway most of it probably wouldn't be used anyway and can't we just reclaim it. And we said, yeah, we can look at that, it doesn't save us from impending doom, only IPv6 can do this, but we can try. And we tried with quite some success, and it's still ongoing of course, but of course we can't get that much Address Policy back and it's all good.

The next question is from the governments and the police force and other people. Do we actually know who is using that stuff? And we say yes, well we do mostly. So I think this a little bit about that, legacy space.

The background: Legacy space, in principle, is defined as Address Policy that was given out by IANA prior to the existence of the RIRs. Now, at some point in time, the RIPE NCC came along and then quickly also the other regional Internet registries in that form. What we sort of inherited from the early days, from IANA there is the responsibility for administrative correctness of the registry. That covers all the IPv4 addresses, the legacy IPv4 addresses in the RIPE NCC service region. We are the Internet Exchange for this region so we should run a proper registry for this region. It's about twoandahalf thousand organisations that hold registry space here, and they, altogether, hold about fourandahalf thousand prefixes. As the slide says, it's one of the RIPE NCC's key responsibilities, as an RIR, to keep the registry up to date and correct, comprehensive, correct and accurate. That's how it is.

Now, what do we do about that? Obviously, now, we want to go and reach out again to legacy holders, not with the request to give us back the space, if you don't use it, but, basically, with an invitation, an encouragement to go and reregister or register and update their registration with us and also quite likely to become RIPE NCC members.

Another interesting activity, of course, quite a lot of them we know, even more we think we know, and then there is the rest. So what we do, very carefully also, is basically run a mini sort of pilot project where we go and contact ten of them who are already RIPE NCC members so we have a pretty good idea of what they are, and say, look, this is the situation, we encourage you to bring your legacy Address Policy to the RIPE registry, again check that everything is up to date, basically formally put it under our authority here.

Then, after that, after the first ten and taking into account how they react, whether this first contact attempt needs some brushing up and some improvement, we go out and contact the other RIPE NCC members who hold legacy address space with a similar approach.

That's easy. Then it becomes more interesting. Then, we try to find nonmembers, legacy holders not yet RIPE NCC members who hold large chunks of address space, larger than /16s or 16s, and then, in the end, we'll go and go after the small stuff.

Now, how do we do this? We have a large team inhouse and, of course, we have also worked with Rob, the Chairman, on how to do this, how to do it sensibly, how do communicate nicely. So we have, basically, a draft that is ready. There is a web page and also brochures and stuff, of course, also to contact. And we have a team in the [registration] services who will do this. We also know that, we have already started this, to some degree, there are a couple of those guys that we are targeting here. We have contacted them and I think about half of them said, yeah, that's no problem, we'll do this, this is a promising start.

Of course, we need to bring forward some benefits. Why would you want to become a RIPE NCC member if you are not? Why would you like to  why would we think that it's good for you to register that space and manage that space in our registry?

Now, you can update and manage it. We know who you are, that's a good thing. There are more services that you might have not known about, that you might have ignored. Of course, that goes for people that are not yet members, that you might have ignored, that you haven't seen. Have a look and you can go and use them. We can nicely update your reverse delegations as well. Of course, you are a good citizen and you provide uptodate information for network operations, basically that's what the database and the registry is for.

So you are a good guy if you do this.

Of course, you don't have to do this, we can't force anybody and don't want to force anybody, but, of course, there are risks in not doing this. Probably the worst is the risk of hijacking. If we don't have any ongoing contacts with you, it is easier, probably, to pretend that somebody is you and take your space away. We try to avoid stuff like that, but, of course, the more up to date your records, the more frequently we have contact with you, the less that risk is.

Then we say, well, if you don't do this, we have some database entry, we have, you know, some entries in the registry for your Address Policy, but that is  we freeze that. We don't know what to do with it. We don't want to run any risks that bad things happen so that what we'll do. We continue reverse DNS setup, but we don't want to maintain it, either. That's sort of  we can't guarantee it really, what's in there, and whether that's still up to date, depending on what happened to the address space.

Basically, those are the things. You might not be able to prove or to show that you are the legitimate holder of that address space. The records that you might have might be really old or they might not be there any more, so bring it in. Obviously, some of you will have questions about that. I am happy to take them. There is, like I said, a dedicated page on our web server to talk to this project and the benefits and the risks. You can also call us, like, on the telephone, remember those things, and, of course, we have a team of people here, and those people, go and talk to them if you feel this is talking to you.

And that is the end of it. Any questions? Any address blocks to be registered?

AUDIENCE SPEAKER: Hello, Stefan from AfriNIC. I received a letter telling me that I have two /16s, which I knew about, on how great it is to register data, and this sort of thing. Thanks for reminding. But it didn't mention at all a special issue for us; we were, a long time ago, a national Internet registry, this two /16 were not for our own use but for to be distributed to another organisation. Now we have a lot of /24 and in a lot of place, companies that are sometimes bankrupt, disappeared or difficult to get in touch and we  we  I mean AfriNIC  are not really sure what to do about it. We still have the /16, so, in theory, we could delay all the /24, but some are in use, some are not. It, clearly, seems too harsh. I also discovered something more funny: We are also the maintainer of a lot of /24 in other prefixes which were allocated by the French national Internet registry at the time. They are part of a /16 which does not belong to us, we don't really know what to do with it. In most cases, we are the maintainers and the RIPE is another maintainer, and more and more people are getting in touch with us. For instance, last week I had someone who had a /23 calling me because they wanted to use it, it was allocated for 16 years, they never used it, but now, of course, the /23 is worth a lot of gold and they were asking me how to revive it. So, I didn't find in your talk or in discussions with RIPE NCC Services, what to do in the special case of former national Internet registries.

AXEL PAWLIK: Right. In that case, I would ask you to talk to us, thank you, and help us find those people, if they are around. And if for a long time we can't find them jointly, I think the prudent way forward is to, you know, have a long list of prefixes that we  well, put in current time somewhere and reclaim after some amount of time. But, basically, the important things or the interesting thing is getting in touch to find those people.

AUDIENCE SPEAKER: Okay. I will. Because I already filed a ticket with NCC about this, and the people with NCC I talked with, apparently are not trained in the issue of former national Internet registries, they don't know the case, and it's something that maybe will require a bit more training because it's only old people still remember 

AXEL PAWLIK: It's only old people and hair and stuff. Absolutely, no, that's right, and as we embark on the approach, we find those things. Again, you might have, or you probably have, some records, some of them are up to date. That's exactly the problem, if you can find the people, you know them, that's great, then we can get in touch with them. And others have disappeared and then we have to dig a bit deeper. But it's good to know that. Thank you.

AUDIENCE SPEAKER: Rob Evans from [] Jan /ET. We have also had some of these letters. I am curious what it means to bring the address space into the registry. Does it affect billing? Does it count as independent resources under 200701?

AXEL PAWLIK: Yes, no. Obviously, if you bring them into the  what we know in our registry, then we will count them as your resources, and according to the proposed charging scheme, that's slightly out of scope here, but of course those will become part of resources that we count for you, but it has already been said, for the next foreseeable future, we won't actually charge for that, so they won't be  we know they are there, but they won't be taken into account for the charging scheme.

AUDIENCE SPEAKER: But they will be taken into account, folks, as existing holdings when you are applying for new address space in that limited amount of time that we have to apply for new address space?


AUDIENCE SPEAKER: I think people should understand, from Axel's presentation, that the aim of the game is to get a dusty corner of IPv4 address space cleaned up, and it's old stuff mainly, so we  yeah, we might not have thought that when the RIPE NCC started, we had the notion of national Internet registries which was not really success, so it was stopped rather quickly, but, yes, they did exist, address space went there. If address space is being brought back under proper management, will people be charged for it? At the end of the day, yes, keeping a registry up to date costs money, and the more the risk you have in the registry, the more cost you incur. How much exactly, that is a matter for the membership and the Board to find out, but I think people should get used to the idea that a proper registry is beneficial to everybody on the Internet, and keeping a proper registry costs money.

AXEL PAWLIK: And among all of us we have agreed that this proper registry, an uptodate registry is one of the highest priorities that the RIPE NCC has, and the threshold, barring people from doing it should be as low as possible.

AUDIENCE SPEAKER: Dave Wilson, HEAnet. On the second bullet there on the slide that is displaying now, am I correct in saying that that is a change from the behaviour today; that, currently, holders of space that aren't RIPE members can update their registration and will cease to be the case?

AXEL PAWLIK: Actually, I am not personally quite aware of our practices there. I am sure that Registration Services can speak to that.

SPEAKER: Andre, RIPE NCC. At the moment, if you have maintainer, a valid maintainer for legacy resources, you can update your records in the database. Like Axel said before, one of the points is that, often, we do not know who those people are, so, yeah, that's going back to that. But at the moment, yes, it is possible to do so.

DAVE WILSON: So at the moment it is possible and it's likely in the future that this will change, is a [fine answer]. I have a followup question, I am not looking for the answer now but I will need a very good answer; lots of my customers have legacy space. They will ask me, "I was able to do this before and now I'm not, why has this had to change?" And I will need a good answer for that.

AXEL PAWLIK: Basically, what those people can do, if they don't want to become a RIPE NCC member because they don't know who we are and what we do, they can move under a sponsoring registry and, well, indirectly do that that way.

DAVE WILSON: I understand that and I have no problem in explaining the benefits, but the question I will get is: Why do I need to take any action at all? What has changed from before?

AXEL PAWLIK: What has changed is that, over the last couple of years, we have come under increased scrutiny, especially about the correctness of the data here, in general the idea of legacy space, and we need to be seen to do something about it, and the closer we are in contact with those people, the better it will be, directly or indirectly, through you.

AUDIENCE SPEAKER: Wilifred Woeber from a very similar environment like Dave, from the Austrian National Research Network, and I was expecting something like that to happen for a couple of years, I think it's not a big surprise. With regard to the implementation, looking back in history and looking back at the discussions that were leading eventually to 200701, I am slightly surprised to get that served as a fully cooked set of rules. So, the reason why I am standing here is that I think there are some pockets where the procedures as proposed here, has already started to be discussed in this environment, that they need a little bit of flexibility, maybe, or a little bit of finetuning, and, in particular, following up on Rob's, Rob Evans' question, like, I would probably, most probably propose to our customers to move under the umbrella of the existing LIR environment, instead of doing sort of the contractual thing directly with the RIPE NCC. But then this should not imply that nobody else in our network can actually get addresses unless we convince a university to give up half of their old class B because it would count on the usage rate. So I think there needs  there is a need for quite a bit of clarification and of, sort of, agreeing on how to really do it in the daytoday environment. Thanks.

AXEL PAWLIK: Thank you.

CHAIR: I really want to close the microphone after this because we are running way over time.

ROB BLOKZIJL: I have a followup remark on Wilifred's remarks. By the time we have cooked out all the fine details of these rules, and I fully agree with you, part of the exercise this week is that we identified the RIPE NCC members who had registered and who have legacy space and I have been  I found a nice letter in that pack and this is for us to learn how to do this and this is not going to be a mass mailing to twoandahalf thousand organisations. This is being handled on a very carefully, almost personal, basis.

Secondly, by the time we understand how this whole field is composed, there are no IPv4 addresses available for allocation any more. So your university should not be concerned whether it is counted as being under their control in case they do another application. I can guarantee the next application from your university will get an immediate reply saying, sorry, we don't have this any more. Have you thought about IPv6?

AUDIENCE SPEAKER: I just have a request, is that you work a lot better on your valuer position, because what I'm hearing is that the reason you gave Dave for doing this is that you have a problem, you are under increased scrutiny and how you are going to solve it is, I have to work and I have to pay. You have to align those things better.

AXEL PAWLIK: Certainly it's the old community that has the problem about the scrutiny. The selfregulation process, and all that, and we are just the top little bit that is seen, so we hear it and most of you hear it through us, but yes...

AUDIENCE SPEAKER: Niall O'Reilly, University College Dublin. We are one of the clients that Dave Wilson mentioned, and I think it's important that I shouldn't let this opportunity pass without mentioning that back at the  in the chicken and egg situation when the NCC was starting up, it was legacy holders like us with motivation to be good citizens that helped establish the credibility of the RIPE NCC by registering, way back then, our resources in the very small, at the time, RIPE database, and that shouldn't be forgotten.

AXEL PAWLIK: Thank you, and it's not, and I am sure that most of you guys don't have a problem with this. It's the people that have fallen off the plate a little bit that we have to find, that we'll say "and who are you?" Thank you.

CHAIR: Thank you, Axel.


CHAIR: Next up is George Michaelson, who is going to give us an IPv6 beauty pageant, I have heard.

GEORGE MICHAELSON: It's always a killer getting the afternoon slot after lunch because you have all eaten a lot and feeling very tired and sleepy. So, I'd love to say that I have a really exciting movie or some intensely beautiful images. I actually really like the images, but I have to tell you it isn't quite what it seems.

So, just to be clear to the sisterhood of the RIPE NCC, I actually think this was a very bad choice of words, and I certainly didn't mean to imply anything at all about the nature of our industry. It was just a joke about the comparisons that we are going to have to draw about relative performance in network technology. I actually think IPv6 is very beautiful, very strange but very beautiful.

Okay. So there is this thing going on that we love collecting data, all of us do it all of the time, we collect or access logs to our switches and routers and we collect the past configurations in rancid going back 15 or 20 years, we collect web logs, we collect DNS logs, we collect NTP logs. Everyone of these laptops open here has logs going back probably in a time machine for the last five years. Now, APNIC is no different; we have been collecting data for a very long time. We have DNS data from the birth of the DNS system in registry control. We were looking at IPv6 back in 2002 because of the registration system for 6to4. Geoff has BGP data along with people like Susanne [Hers] that goes right back into the 1980s.

Now, if you just sit on this data, that's hoarding; it's not actually constructing anything useful from it. So, what we are essentially trying to do is we are trying to look at the vast massive data that we have available to us that we actually think is unique to us as registries, and find a way to get some value out of it to put back into the community to inform the decisions that the community has to make about this transition to 6.

So what we think we have is this system which is a really effective basis for performing v6 measurement. Now, this isn't something new. We have been working on systems that measure v6 for a long time. I mean, we had all have gone into our web logs and done a relative count in 4 and 6 and said, wow, we are taking amazing hits on 6 and done stuff like this. But when we moved into this situation of using active measurement instead of passively looking in the web logs, but actively engaging with the users' software and [tickling] a little bit more out of it, we actually got to a point where we could discover a lot more about the capability. And we have been doing this long enough that we think we now have a volume of data that's going to be informational on trends. It's not just that we can put up single value statements about this, but we can actually drill down and get some finergrained information out of the data.

To be specific, we think we have somewhere around 50 weeks, although there are some hedges on accuracy for the earlier collection, which is an amalgam of a methodology using Java script that you will have heard discussed several times. The RIPE NCC is very active in this field and there are many people who have been using this as a measurement. We have also been working on some flash that I am going to talk about a bit more in a minute. And these measurements, they check a number of different behaviours in client systems.

They check dual stack preference, they check IPv6 only, but they kind of go a little further because they can even use direct literal v6 to make clients that would suppress effect because of happy eyeballs or because of decisions about not using network stats. They can drag a bit more traffic out. And we have managed to get the volume in our particular cycle of measurements up to this number, somewhere around 250 to 300,000 measurements per day, so that's the number of end hosts that are being measured. I'd like to say that it's a high level of coverage. I have put up this figure of about 100 that have got sufficient detail to be statistically useful. That's kind of a rough number. In practice, the problem is that, for fractional counts, you can't rely on a small sample set because you might only have one or two IPv6 events set against the total population of fetches. So, to be useful, you have to have a lot of measures. We think there is somewhere around 100 that are starting to get to a useful level, and out of 249 economies that the UN recognises as entities for formal recognition with a twoletter code, that's really pretty good coverage. We actually go quite a bit further than that. For instance, there are 28,000 people that live in the Aland Islands at the north of the gulf of Finland, and we have seen eleven of them. I don't know how many IP addresses there are in the Aland Islands. We have seen four hits from the Vatican. I don't think, on that day, that the Pope's iPhone was v6enabled, but we have measured traffic across the spread fairly widely and we have certainly gone to all economies; it's just that we haven't necessarily gone far enough to get statistically useful data.

So flash: The Java script active member is about asking people who are website owners to include fragments of code in their website that will encourage the users to do some fetching and we then receive the results of those fetches through tricks in the DNS that stop caching from happening, and that can inform us about the relative delay and behaviour of the system, and that's great and we love it, and if any of you are interested in doing the Java script measurement, either with us or with the RIPE NCC, get on board because there is a wealth of data there. But the thing is, it's kind of  the thing is that it's kind of, it targets the people who are motivated. You can have weird things happen. One day, out of the blue, we got a millionandahalf hits because a major news corporation added this onto their front page and we were getting one hit for every person that refreshed on that news site, and they were big and we didn't plan for those. There has been telcos in eastern Europe that have added us and suddenly provided a huge amount of data about one economy. We had this thought: What if we made an investment here? What if we bought the eyeball? And if it turns out that you can actually bid a really low price in the flash advertising community and the ads will say back to you  the ad sellers will say, back to you, you aren't going to get any clicks on this, are you sure? We said yeah, we're sure, and you don't get clicks, but you get a lot of eyeballs because they want to get that money, they want that 10 or 20 dollars a day, and the way they get it is by getting as many eyeballs as they can to get fractional dollar amount for the eyeball. So if you see these ads, don't click, because every time you click, it costs us more money and we count less people. That's why we kind of made them very pedestrian. The first one was a little too flashy. You know, it's kind of poortrash cousin. The bottom one, grey is the new black, it's European, very neutral. If I had known orange was this year's signature colour, I would have done an orange advert, but we don't really want your clicks. There is also a problem that there are some operating systems that suppress flash, so we do understand that we are undercounting certain categories of behaviour. If I observe that this is iphones and Android and make the second observation that I am not aware of any 3G network which has enabled v6 in the network segment and that they only do it in WiFi, I think it's okay to understand that we are, maybe, not seeing those people in these measures.

There are people out there like you, Randy, who would sit on it and click all day.

RANDY BUSH: I would not, and also, I have Flash turned off.

GEORGE MICHAELSON: Okay. Valid point. Okay, so we think we have data and we think we have a methodology, so this poses that really important issue: What is the question?

So, Geoff characterised this question as the word "lumpy". I mean, we have all seen the singleline figure. The world is at .3 percent. How even is that? Can we drill down a bit and understand a bit more about the breakdown? Is it necessarily the same everywhere? What have we had this ability to understand geographic differences and that is where the RIR system, and you guys, keeping your data accurate in registry and in WHOIS, that's where the value came in. I'd love to say this is all our work, and of course it is, but it's actually all your work, because the more data you put into WHOIS and you put into registry in terms of accuracy of your records, the finergrain detail we have about disposition of the resource.

Now, we actually got someone to look at this because there is obviously that question. We can't be as good as the GO providers, we can't be as good as MaxMind. If you go to the economy level, we are. We are up in the high 90 digits of accuracy for the economy of disposition of most of the addresses. So as long as you stop at the economy level, we have really, really good information in our systems, which is an amazing thing.

So, another way of putting this is: Have we actually succeeded in communicating with people about this? I mean, it's not just the question how much 6 is out there, but are the programmes that we are doing at a national level or at a regional level, are they working? Are the communications working? What actually happened on World IPv6 Day? Did people see anything?

So, we are quite interested in this idea of the level of effectability of the deployment. Now, a previous version of this pack, I have to admit, was very flawed. It just focused on this thing about capability and observation was made, it's basically just counting Windows, which is a beauty pageant I don't want to go to. It's not interesting. If you want to know about that, what you can do is look in your own web logs. So we are not talking about capability of v6 here. We want to talk about actual preference. What we found is the relative IPv6 capability at the economy level, and this is, we think, a very real figure. We have managed to suppress the tunnelling and get away from that, and we even think we can get into the preferencing inside the dual stack. So, we do consider people like Hurricane Electric or 6 RD that have done infrastructure, as interesting, and we don't want to discount them. We are not interested in that capability. It's about enabled, capable and delivered, and we have gone to this idea of the relative measure, how are people comparing to the world average?

So, here we go: Now, this gets to a very difficult place, because we actually try quite hard not to make comparisons because it's not the nature of the business. This is a community that has very big players and very small players and it doesn't always feel very collaborative to make these comparisons, but this is really important stuff, you know, this is not something we can actually afford to ignore any more. We have to look at this. And you may notice on these graphs that there are some very clear things happening in terms of winners and losers, if I use that as a metaphor, differences are emerging.

Okay. So, here is the headline figure. The world is on 0.3. Well, this is our take on the 0.3 figure. So what you are looking at here is a graph that's in log scale. We are engineers, we can count in LOGs and the red line is marking that 0.3 figure. If you go to the left of the vertical blue line, that's the historical data that we are less confident on. That blue line is World IPv6 Day, and I think you will agree, there is an interesting uptake at that point, and it's possible to say, maybe worldwide, we actually did make a difference. Things kind of settled down after that and I would say there is a fairly strong signal coming through our data the world is actually sitting on this 0.3 figure.

Okay, so now let's ask what happens if you look at the per economy distribution? So this is the scatterplot of all the economies that have enough samples to be statistically meaningful, and I think you will agree the variance is about 0.3, is huge. We have got a variance that's between 0 and 10% and 0.01%. So we have a lot of variance. And this is actually a variance at the economy level. It's not that they're hugely variant signals, they really are different.

Okay, so let's look at this at a regional level. Now, this plot is using the UN regions, which are well understood. They don't align completely with the RIR regions; they are subtly different. They are used by economists, strategic planners, by the OECD and the G8 and the G20, and we have made a decision to use this because we want to have information that can get beyond the technology space and into the regulatory space. We need to have this conversation with these people.

So, if you go and look at the regional level, you get this sense that, actually, at the regional level, we are grouping fairly well around the world figure. Now, there is one standout that's below the line. I hope you can see there that it's Africa, and that's coming from around four to five hundred thousand samples a week. I am confident this is an accurate reflection. This is not to impute failure on the part of the African community. This is an observation about the reality of the customer side. Think about the CPE issues in Africa. I think think about the GDP per economy in Africa. They are struggling with issues there that are beyond the capability of the RSP to resolve. There are some amazing upswings in this data recently. The other thing to notice is that North America, which is often talked about about as the standout behind the v6 wheel, it's actually looking fairly ordinary compared to the rest of the world. It's interesting.

So Europe: I mean, I know my audience, I did this one at the LACNIC meeting and I talked about the South American economies, but I have redone the slide pack to talk about Europe. You can see Europe is consistently showing above world figures. It's quite comfortably now into the solid single digits and it's probably going to stay there, so, yes, we have actually seen some local change. If you do the breakdown into the subregions of Europe, you actually start to see a bit of difference emerging. Northern Europe and western Europe are quite strongly ahead of the other regional economies. I think you could also argue, if you look at that blue vertical line, there is a very strong uptick for World IPv6 Day. So it's arguable to say that the communication activity really did have an effect there.

So, drilling down into western Europe, that top line, that brown line, that's France, so if [Renne Deprai ]is here, or if anyone from Freenet is here, I think you can feel fairly comfortable that your technology is showing some strengths. The groupings in western Europe actually are a significantly higher level and you will see that they are well above 1%. So the adoption level in western Europe is really very, very high compared to the rest of the world. If you make a comparison with east Asia, you get this huge variation. The top line is Japan; the bottom line is Korea. Now, that is surprising, but if you think about the age of penetration of fibre to the home, highspeed cable to the home, a lot of this rollout was done before v6 was documented in [DOCSIS], so I suspect their CPE is actually not v6capable and that what we are seeing here is that their investment in home network delivery was like first adopter effect. They are going to have a repricing issue when they want to do the v6 deployment.

So, to end, v6 service actually is quite lumpy at the economy level. There are really quite big variances at economies and regions and there are variances inside regions and you can't necessarily make simple assumptions about what you know about an economy in terms of its GDP and automatically get to the right figure for v6 penetration. You have to know a bit more about what's going on. We are able to see some signs that we are measuring ourselves and that probably means our advocacy is working because when we do things that relate to 6, we see effects in the graph. Yeah, they are small numbers, but we are actually getting a message out there. But we have got a long way to go. Where we going with this? Our intention is to carry on doing this. We actually now have a resource online which is in its very early stages but it provides full breakdowns for every economy and also for some organisational structures like the OECD and the G8, the G20, and our hope is that we can get into a dialogue with strategic planners using this data and this resource and communicate with them. We are making all of the anonymised graph inputs available, so if you want to work on this and put it into your own Powerpoints, that's fine and we are trying to add some other stuff that's going to go down to the network provider level and maybe the prefix level, but maintain some degree of anonymity. It's a very strong issue for us. This is a longterm commitment we are making to a measurement activity. I suspect you'll see Geoff, because he is the real economist who is actually going to say some things about this. You are going to see a lot more from us about this, because we think this is a strategic planning input which is going to add quite a lot of value into the communication.

That's it...


CHAIR: Questions. All right. Thank you, George. Next is going to be a panel discussion on the IPv6 CPEs, led by Andre.

Let me invite the panelists to the panel, to the comfortable couches or armchairs, and this panel is about IPv6 readiness and customer premises equipment, consumer electronic equipment, and, well, we know that the long and windy road of IPv6 adoption is full of the circular dependancies when content providers don't see much sense of deploying IPv6 because there is no one to use it and ISPs do not see much sense in deploying IPv6 because there is no content and also because the users are not capable of using IPv6 even if they deploy it in their core infrastructure. So, in this panel, we are going to talk about this last metres of IPv6 deployment and, in particular, about CPE equipment and CE equipment and its IPv6 readiness. We invited four vendors of CPE CE equipment in to this panel, and let me start introducing them. I will not introduce them myself, but I will, rather, introduce them by way of asking them to make  Hans, you are the first, I guess. I will introduce our panel by way of asking them to make a short presentation about their products and the readiness of their products.

HANS LIU: Hi, this is Hans from [] dealing, and, today, I would like to talk about our IPv6 CPE device developed and deployment.

In [] dealing, we have been participating and we have been developing our IPv6 CPE devices since 2003, and, as you can see, we got our first IPv6ready logo Phase One and Phase Two in 2003 and in 2006. Later, in 2008, we shipped it out, our first IPv6 CPE ready routers to the retail market with our IPv6 ready logo and the debugs. Last year we started participating IOT test events to make sure that our devices and our implementations has no IOT issue with the others. So we went to cable labs and the past test event layer and this year, this May, we also have our embedded DS CPE server to 35 by IPv6 logo programme. It is by far the first embedded IPv6 DHCP server less 35 bit programme. So far, we have more than 23 IPv6 routers and the SS point available.

This May, we also joined layers that test event hosted by UNHIOL. It is a test event lab, do tests based on RC6204, which is basically a requirement to CPE routers and the test run which them is based on that. We are very proud to be the first router that passed the test.

So, let's take a look at what we support. We support of course the tunnel, 6to4, tunnels. Connected dual stack, DHCP, DHCP 6, PPP, and we support DSLite. In addition to the wind type, we also support, as you can see here like simple security, ULAs, cooperate with DYDNS, and MMR  device web page measurement web page login.

In addition to the functionality, we also push very hard to the performance, because I know that performance is somehow an issue for people to hesitate to go on to IPv6, although we know it's not true, and some our highest models can now reach 2 gig wire speed IPv4 net and IPv6 routing.

Regarding to deployment. We cowork with ISPs very hardly on a field trial and we also drive IPv6 to retail market and we want to promote it to IPv6 home users directly. So, if you see IPv6 ready logo on our box, it means that the device is IPv6 ready and you should give it a try.

Since 2008, so far we have already shipped out like 10 million IPv6 ready devices in the market. So, the length of CP device should not be a reason or excuse to hesitate an ISP to move on to IPv6.

Here are our router device that support the IPv6, and here are our new ones coming. Also, the SS point, and we also do power line device. So in our power line the WiFi router, we also support IPv6 a layer of available in a market right now. That's my presentation. Thank you.


CHAIR: I wonder who comes next? Okay. It's you, Ole.

OLE TROAN: My name is Ole Troan. I am working for Cisco, I have been doing v6 since 1998. I am getting a little tired of waiting now. So hopefully, I think we have reached, what did you say, George, about 0.3 percent? So that's what I have achieved in 12 years, so perhaps you shouldn't listen to me.

So the Cisco strategy is really we want v6 on every application, every device and, you know, every service. And yes, there are gaps in that, but we really have gotten pretty far. We have finally got all the links [] cyst stuff going since May this year. A combination of all the software and some of the boxes have their own distributions for that. We have the small business stuff do v6, you know, so there is a difference here between retail routers and, you know, SP managed routers, the latter ones doing TR 69s for management docks and stuff. Those in the Telco home business unit, they support v6, cable home networking to v6 and we have all the good old stuff running IOS, also I joined in 1998, we had a decent v6 implementation then running on IOS, so we have, you know, what is it? I think we had the first routers out in '96, we have 14 years of v6 experience. And, yes, sure, there has been some improvements since then.

So, Andre asked me to see what trends do I see in this space? And certainly I see  I am mainly, now, involved in lots of architecture stuff and also ITF BBF things and the innovation in, you know, CPE style features, features, changes required on the CPEs are increasing and, you know, it's off the map. That was a pun that you probably didn't get because I posted a draft called map just the other week, which is a mechanism to do stateless v4 or v6. That's what you do with your leftover v4 deployments that you hopefully have very soon. There is a lightning talk later today on that if you are interested, but it demands on the CPEs are increasing as a consequence of that and we are innovating like mad. I think we are averaging about  well, I didn't dot numbers, but it's, you know, about one new proposal every week in, you know, IETF Working Groups like Software and Behave and others, PCP. We have, you know, the buffer guys running around getting  really wanting CPEs to be fixed, we have the home networking group that is, you know, wanting to build much, much more advanced home network where you couple lots of these CPE routers together and where you want them to you know automatically discover edges and do prefix configuration and stuff. Also, the last trend I see is is that the v6 deployment trend. That's you guys. I mean, that graph, which is in Norway and I think it matches with George's numbers, isn't very impressive. I mean, this is not exponential growth. This trend doesn't take us anywhere. It's certainly the case that the CPEs are there. Don't use that as  my job, I work with Mark Townsley, and our charter in our little group in Cisco is to do whatever it takes to remove any roadblock there is to v6. And v6 CPEs used to be a roadblock. It isn't any more. In the last year, yearandahalf, has certainly changed that. So we are done. I mean, I am the editor of RC 6204. The requirements. I am the editor of BBF TR 1204, which provides the same for service provider routers and they are all done and they are products. We are doing lots of new stuff. I am an engineer, I'm most excited about all the new stuff. But that's going to take a while and, you know, that's certainly not something that should stop you from turning on v6 in these boxes today, they are fully capable of doing that.

Let's talk about the challenges when we have questions I think.

CHAIR: Thank you, Ole. If you can click through, Eric, then you'll hopefully get to your slide.

ERIC VAN UDEN: Good afternoon. My name is Eric. I work for [AVM], we are a German company and we produce DSL devices for Internet technology, but, as well, [EuroDOC SIS] and LDE devices, and in this case I want to tell you very short and very brief, in a nononsense way, what we did with IPv6, and I absolutely agree with my neighbour, why don't we move.

We started twoandahalf years ago about router force implementation of v6 into our products. In real life, it's part of our standard firmware since 2010. For this, for AVM it was very important we use our own stack so the FRITZ boxes are based on a Linux base but we use our own IP stack to get a more flexible and more speed up with connectivity.

You can't do this without the help and the support and the knowledge of an ISP and in this case we had the good XS4ALL, who encouraged us to speed up with implementation of IPv6, and together we learned a lot about the pros and cons about implementations IPv6 into CPE in real consumer CPEs. And I think, and you will hear more about this in the afternoon, and I think this brought us both in a very strong position with implementing IPv6 to the end customers.

The other approach that we see the last months coming up very rapidly is the implementing of the dual stack light. Dual stack light is needed for the poor ISPs. They were available already, they don't have blocks available any more and they want to still growth. You can use DSL light, by using the IPv6 connectivity and a tunnel protocol call the DSL light to get the IPv4 back online. The DSLite is implemented in all our products but you see most active from the cable environment.

We are not there yet. We are still working and it is also for AVM very important that we implement a lot of features and stuff, but must be available for the end user, must be simple for the end user because you can't expect that the end user can understand everything what is needed for IPv6 and our approach is to make it very simple, just a few clicks he is online and he is not only online under, let's say, on a speed way but he is also online in a secure way and that's why we make it very simple and we invested a lot of time and money to keep this IPv6 implementation simple for the end customers.

Our new thing was also requested for our customers and the implementation has some stuff based on, still working on the IPv6 in the LDE environment. Provisioning is done, VoIP is done. And I think a very big approach, big decision was made a few weeks ago that we said okay, when the FRITZ box is active in the Netherlands, it's connected to an XS rollout. It's always IPv6, so we switch on IPv6 native  I think this is a good approach from X 4 all. This is how we want to go forth and enable IPv6 in all the boxes in the field. XS4ALL is using the FRITZ box as their standard device.

There is a white paper and I had it on my presentation, but if you want to read more about all the stuff we have done, we have a web page,, where you will find all the information, everything we did for implementing IPv6 to our products.

Thank you.


CHAIR: And Carl, please.

CARL WUYTS: I work as a system architect in the networking division at Technicolor. This is formerly known as Thompson, is one of the market leaders worldwide with over one hundred million sold gateways over the last years. We are headquarters of technical, and DSL is situated in Belgium, where I am at.

Andre asked me to pinpoint on four major topics here is the CPE ready? How did we do the development and the deployment? And what about the user?

Well, simple, same as Hans said, we are targeting the IPv6 ready logo Phase Two, and in fact we did it with the current firm wear on our DSL devices. So it's really the one you are got in front of you as a customer. So we think this was an important thing.

The feature was very straightforward. We focused on packet forwarding, so no services so we started with normal dual stack with PPP on top of internal mechanisms in place. BGP, DNS, all the normal straightforward things you might expect in a packet for the kind of deployment.

The important thing there is that we did it all by ourselves. Initially we were looking at taking something from open source. But the legacy v4 deployment, so we decided to dot same for the v6 and for a few obvious reasons, the disadvantage there is that you are a bit slower in the market because you need to do things yourself. But the big advantage is that everything is owned by yourself. You are not struggling with GPL, you can keep everything in kernel if you want to, instead of moving it to user space. We have a very low memory footprint. It's the v6 part is less than 1 meg on a flash now for the whole v6, so it's quite good. It's also reflecting on the performance of course, and if somebody asking us to add something extra, well we can do it ourselves because it's owned by ourselves.

The deployment, we have changed there. Initially, we were showing slides like we do now to the customers saying, we can do this, we can do that. But now, in fact, we threw away the slides and we say we just come to you, plug in your DSL cable into your device and we show you how easy it is to get it working. We have, because it's our own development, we have made everything configurable, so we have a parameter nearly for everything, meaning that when we go to the customers, we preconfigure it for his needs and, in fact, he can just plug it in and work, no matter whether it is 6RD or DSLite, whether he wants to do it static or dynamic with options. We can choose and just plug it in. For our legacy users, it's important that he can upgrade their device to v6 without having to reconfigure everything again. It's quite normal, I think. A typical trend we see is that it's now present on every tender you get. It's a must, while before it was an option. What we also see, that you talk to customers last year, they said, well, we have v4 addresses for another four years. If you talk to them today, it has decreased to maybe one year left, so you see the gap is closing pretty fast.

The final point there, the user readiness, and, well, it depends a bit on what you understand as user, or target customers is the residential CPE user, so, in fact, the user at home. And the question there is: Does he need to know the IP address? In fact does he know what an IP address is or should he know what the v6 address is? I think not. He wants just to user box and surf the web and keep doing it for many, many years. So for us it's not a target to teach the end user what an IP address is. But you just keep him on the web.

Now, to convince the ISPs for this, because our customers are ISPs and Telcos, we give them some examples. For instance, if you want to convince your customer what IPv6 is, showing a real life example like plug your iPhone in, your iPhone is IPv6 capable, it can connect so it can make a link with real device and for an end user an IP address which is something virtual for them. With regards to the NATTing, same thing. We see a lot of DSLiting, whatever it is referred to, it's a bit dangerous there because you might introduce a worse user experience than before because you share addresses. It depends on what the user wants to do with it, but, again, make him aware of what he can do, what he can tell to his customer and try to guide him there, instead of just tell him, do this or do that. Another trend we see is that we have a strong go to the 6 RD and DSLite. We have some customers that don't really know it yet and others who [deposed] everything. We have all DSLites, A plus B, everything which is available and present and you have to mark compliant or not compliant. So I think our role is there to stop throwing all kinds of protocols in and make a decision and pick a few of them.

Thank you.


ANDREI ROBACHEVSKY: And thanks to all of you for this introductions. Before I open the floor for questions, let me pose a few questions myself.

The first question would be: Well, did all this development happen as a response to a demand? And where this demand came from and where this demand is coming from and what are the trends in this demand if you can elaborate on that?

OLE TROAN: I can go from a standards perspective, perhaps, that  so this was a demand, you know, DOCSIS cable guys came together and wanted to do v6 and they did DOCSIS 3L and specified a CPE in a specification [lee] router which came out in 2008, or something, so they were the first ones and they certainly created this package of DOCSIS 3 where v6 was a mandatory part. That was certainly one part of it. The programme forum which specifies the architectures for DSL and [EPON] networks came out with this TR 124 I2 last year which also specified, you know, how v6 CPE should look like if you wanted a CPEmanaged CE, a TR 69, bells and whistles. And then, later on, the IETF, in an effort, together with the cable labs and the  because some big broadband providers wanted some common functionality on retail CPEs, so the ROC 6204 is really focused on the retail CPE. So that kind of drove some demand at least from the standardisation side. They knew what was supposed to be built.

CARL WUYTS: For us, it was different. I think we just started with it like two and a half years ago, approximately. We saw the increasing demand in the field, but there was no really something pushing it that way. But we were aware that we needed to start in advance to get to a decent point within a decent time frame. So we were not driven by DOCSIS or DSL or a specific customer, but more in general that we saw a  indeed, you can see the charts with the full completions. So an at a certain point in time you need to get in there and start doing it.

ERIC VAN UDEN: For us, we started to implement this twoandahalf years ago, we demonstrated on the Cbit. In that time, luckily for us, they had more or less the same rollout plans, so we could speed up each other and bring to a successful implementation into our products. So, first of all, it was more internally and to look and see how is it working, and then XS4ALL came to us and said we want to go active with it so that helps us to speed this more up.

HANS LIU: I think that covers everything.

ANDREI ROBACHEVSKY: Another question would be: What kind of challenges do you see? Do you see this kind of chicken and egg problem in this field or any other challenges?

ERIC VAN UDEN: From my point of view, there is still some discussion that the people will say: Why do I need IPv6? This discussion is still happening, still. It's unbelievable it's still happening. They say we have enough IPv4 blocks, we don't need it. First of all, and we have some  of course, some talks about it. It might be true, but, on the other hand, if you don't move now, then it will cost you a lot of money when you have no blocks available any more. So, for us, it's always the discussion there is no need. Also, I agree with our neighbour, that he said a few years ago that we have enough blocks, and suddenly after one year, it's, oh, we maybe still have one year, so they are surprised by the speed of the phasingout of the IPv4 addresses. The second topic is, what did people say, we are a grade, oh my God there is no NAT any more because they think NAT is the security issue and is the security option to help the people to secure them from the Internet. It is not true but that is their mindset. But also, on that topic, we had some discussions to convince people, guys, IPv6 is not less secure or more secure than IPv4. You have  it's the same if you do the right things.

This was was one of the main things we saw in our discussion. On the other hand the implementation on the network side because of legal perspective was also very important but it's more related to the ISPs.

HANS LIU: I think also we sort of later, like, a standard unique and decent host implementation, like from our point of view, as the device vendor, I see like IOS support like stayed with the DHCP, Window 7 ghost is a very decent DHCP. V6 while  since the poor like the DHCP, and things like links has said to users to go for IPv6 although we claim that our CPE device, or even the routers also support IPv6 or IPv6 ready. I think like the real client device and like [NAS, IBCAN], like layer support on IPv6 will be a challenge for us.

OLE TROAN: I am with you there, Hans. Some of the host stacks have been problematic. The biggest challenge from a CPE point of view is you don't have access networks to plug into. And probably the biggest challenges we have seen is with access networks. You have hundreds of thousands of and ethernet nodes that don't support v6. And especially in these N to 1 VLAN architectures that people appear to like for some reason, you shouldn't do that, you want to run VLANs and this problem doesn't exist. So we vented things like 6 RD to tunnel across these access networks. Now, most products will come out or have come out in the access network on the peering side with support for, in all this layer 2 hacks, doing discovery inspection and DHCP snoopey and whatever else they do. But that has been slowing some people down, I think, but, you know, the biggest, biggest challenges is that you guys don't want to flip that switch. It's largely flipping a switch. Please do. And then that challenge is gone.

CARL WUYTS: I think one small thing to add is they expect a lot from the CPE. It's a small cheaper device which should really cost nothing if possible and then you see the end device likes Windows XP, although it's ten years old, it's still there massively. It has v6 if you enable it manually, okay, maybe a service pack can do that for you or some potential fix, but, for instance, the nod PCP clients. In order to cope with that we implemented the DNS proxy because then your CPE is the default gateway and we proxy the device to the network, but you see like this must be capable of taking care of the massive MIcrosoft deployment there, not giving support for those kind of things. Okay they might say move to Windows 7. That's a possibility. But we can not afford to say you are experience, not working.

ANDREI ROBACHEVSKY: All right. Thank you very much for your observations. I would like to open the floor for questions.

AUDIENCE SPEAKER: Hello. Jan George from Slovenia. When I go around the world and speak at the various meetings and events about IPv6, this CPE IPv6 question pops up every time and the operators are claiming there is no support CPEs are the problem. My generic response to no CPEs are not the problem. You just need to buy them. You just need to put some money in it. So, based on this, based on this, can you share the actual trends in demanding IPv6 from the field, you are CPE vendors, so can you share with us if operators around the world actually started to ask about IPv6? Thank you.

ERIC VAN UDEN: Maybe we have already  you told it. We see now in the RFPs is in high demand to have IPv6 available. We see also in the latest RFPs we received a demand we want to have DSLite. So, on new RFPs we are receiving and I think we will know the same RFPs are asking for IPv6 and asking for DSLite implementations. So, yes, there is  in fact I see from my point of view, I see more and more the demands of must have of IPv6 availability.

CARL WUYTS: For me it's similar, but it's going a step further now. We have some pilots running at customer networks, they are not just lab tests any more, but pilots. Usually starts within the employees of the ISP, so not a real open pilot, but we see more and more this and we see they are using native and 6 RD for the moment. Not many DSLites at the moment, but the request for that is rising quite heavily.

HANS LIU: Mostly like the same here, but where we got the RFPs, we see that it's interesting that some RFP asking for DSLite while they are asking for 6to4 and 6 RD at the same time. So my question is that I think maybe some ISP still have no idea what layer they are going to deploy layer IPv6 and how are they going to do that?

OLE TROAN: I don't think I have very much to add: I mean, just like Hans said, I think, two years ago, we saw RFPs with a v6 being the check box, and it obviously wasn't meant to work. Now the RFPs are much more detailed and it will be verified and tested and people really do care that you give them something that works. Not just saying that our v6 implementation didn't work two years ago. It obviously did. No one ever used it.

AUDIENCE SPEAKER: Hi, this is Shane Kerr from ISC. I have been not been involved personally but I had some involvement recently with working on open source CPE equipment. We were involved peripherally with the buffer blow project, which is an effort to kind of solve a problem identified of having buffers ruining your endtoend throughput and user experience. I mean that project is kind of evolving into an effort to make open source CPE images, firmware and things like that, so I can guess what the answer to my question will be from the guy from Technicolor, but basically, do you see any role for open source in this new world, especially in the v6 side of things?

CARL WUYTS: Is it my opinion?

SHANE KERR: Okay. Maybe I can ask for a more specific question. Is it possible to flash the firm wear on these new devices you guys are producing at all with thirdparty firmware, say?

CARL WUYTS: It depends. We have two strategies running. We have the embedded CPE devices where you can only flash your own software, but at the same time we are also working on the middle of our solution where we have the component where you can add components from third parties into it, either real components via an API or via an OPSI.

ERIC VAN UDEN: For us, it's more or less the same approach. We have the standard firmware is more or less closed, yes, you can update it with new firmware. But we have now also projects running in FHEM, this is in the multicab projects where you have part of the memory of the FRITZ box will be used for controlling lamps and a sensor, so over the [] multi/KA protocol based on the standard protocol like FX20 and OMatic and so on. We build with this experimental cases also running on IPv6 to control the lamps and to read out temperatures. So for us at AVM, it's under investigation, it's not a really clear decision at this moment what we do and what we have more and more open.

SHANE KERR: So that's a focus on open protocols, not necessarily open software, right?

ERIC VAN UDEN: It's a part. Let's say, for this you need to implement besides of the core, FRITZ box software you need to implement a software where you are able to control things, and this is a third party software. So, we we are a little bit moving in that direction.

OLE TROAN: Before I say, you should really fix your DHCP implementation, because it's not very fitting right now for, you know, small CPEs. But we'll take that 

SHANE KERR: I agree.

OLE TROAN: Thanks. When is the date we get 

SHANE KERR: We are looking for sponsors for the project.


SHANE KERR: Please come and talk to me, anyone in the room, with big sackings of cash.

OLE TROAN: Obviously you can run open source on Linux, and I don't know what the product strategy he if there is one in Linksys, I certainly think that's a very good idea, you should be able to do that. I run DDRT and open  something on our system. Every Linux based distribution has lots of open source in it anyway. Pretty much is everyone is using it. With all the new stuff we are doing, the buffer blow stuff, home networking stuff, would I really like to see a reference implementation that was, you know, purely open source, you could put it on a D link or a Linksys or an AVM box and you could get more of the community involved because there is so much, many moving parts now that, I think it's a very good idea. I don't want to commit to anything from Linksys point of view. I really like this they are open enough that you can put things like zero ERT on them.

HANS LIU: I am a product vendor to you that I will comment to you I will provide simple everything, we are OS open to list open source and holding the space community stuff and actually our IPv6, we use Linux  in every layer. But however, like, I would like to comment that your current implementation it seems like looking for a 60 meg fresh size, why most of our devices go 8 or even 4, I think that would be a  if we like to largely deploy implementation.

ANDREI ROBACHEVSKY: I think it's Jim and then Marco.

AUDIENCE SPEAKER: Jim Reid, representing nobody but myself. First of all, I want to say thank you very much for coming here today. I think it's great that we have got some vendors from CPE equipment talking to this audience and I hope we are going to see more of you from other vendors as well and there will be a continuing dialogue and participation at RIPE. I think this is a very, very big step forward and I thank you all for making the effort today. That's a great step forward.

You have talked a lot about testing and responding to customer demand and in this case you are talking about your customers who are essentially the ISPs. But I wonder if you might want to approach this from a slightly different direction rather than just do top down, push technology to get this stuff out there, do a little bit of bottom up. Someone on the panel mentioned earlier about more community involvement. Well I would bet a fair bit of money that there are people in this room that would love to have experimental or beta tests versions of IPv6 capable CPEs that they would run at home or in their office, I am certainly one of them. And I personally am rather frustrated by this just now because ISP can't provide me with native IPv6, but the only choice they can give me for that is a rather expensive router and 1,000 euro for that. That's kind what it costs. I am not willing to pay that for the benefit of getting IPv6. But if we had a chance to play with some of the existing firmware or images we have got of the existing CPE kit, I would be quite happy to have a play with that, give you feedback, hack on it if that was at all possible. I understand there are limitations about burning flash images and all the rest of it, but I would love to get the chance to play with that kind of stuff. Give you feedback and see if the stuff does perform in the wild. When I talk to my ISP about that, they say oh, no networks, we are still waiting to see what vendor X is going to supply and from our own tests this aren't ready for prime time, so we are not going to give this to the customer to play with this. If the customer has a clue F you have some way of having that bottom up approach with more community involvement, I think this might also help because it can help put us to put pressure on ISPs that are dragging their feet but also the retail DSL market, we are also encourage us because we can say we have got IPv6, it works, it's cool, it does all this with high phones whatever else it is, and then their friends in the social circle will adopt the same approach. Let's try this super new router, put pressure on the ISPs and further up take up IPv6. I'd like to know what you think about that.

ERIC VAN UDEN: This is real life. My company, so we are already doing this. So, if you have also the implementation of the IPv6 within the Netherlands, we first had and we called a lab ware firmware available, so you could go to the website. You click on IPv6, they find you the lab firmware, we say, yes, it's experimental. It might not be 100% fleet. Please give us feedback and that is what's happening. First of all, we start with a small group of the ISPs. As you describe, what I believed was 20 or 40 people. The second step is okay, go live, to some customers who want to have this and the registration, they tested and sent us feedback and so you can go forward, and of course this firmware was also downloaded by other customers, no problem with it. We encouraged this, and this is the way we work, and also we do the same stuff for implementing apps on the iPhone and the apps on the Android phones, this the the way we work. First of all, we send out a lab version, people play with it, send us the feedback. The feedback is under investigation by our development and fix things when needed or do some new implementations. That is the way we work.

JIM REID: For instance, maybe you try professional documentation so you could get something like how to get an open WRT image onto your box, that might be useful to do, too.

ERIC VAN UDEN: This was already one of the questions, and so we are more and more opening our firmware, but there is no decision over what we want to do and what we will do on this open VRT. This decision is not made yesterday. [](significance).

ANDREI ROBACHEVSKY: Thanks for the suggestion and leave your business card anyway, Jim. Please short responses because we only have time for one question.

AUDIENCE SPEAKER: It's Marco, IPv6 Working Group Chair. When it comes to rollout, obviously the new boxes to IPv6. The biggest problem at hand is the install device, I have a question for you on life cycle. What do you think is a realistic view of when these old boxes will die of natural causes? We used to make the joke like there is another victim and there are another couple of customers online with IPv6 because their modems got fried. What do you think? What's a realistic scenario to have the install base naturally replace itself with v6enabled devices?

CARL WUYTS: I think it's kind of difficult to put a real figure on that. According to the ISPs, as long as possible if there is one bit free on the flash, it's enough to flash a new image with new features. So, for them, it's easy. Now, what we are now doing is that the real low end devices will try to get them out and replace them with some more future safe devices with a little more flash and RAM, but how long it will take, I don't know. We have millions of devices now in the field which can upgrade to an IPv6 firmware. They will last for quite sometime, I think. So it depends a little.

HANS LIU: It mainly depends on the fresh size, what we can put in there. Many we support like one year, if the fresh is still available for new features.

AUDIENCE SPEAKER: As far as you are concerned, it's basically not about replacing hardware; it's about getting people to install their new firmware? If I heard you correctly.

CARL WUYTS: The ISP deciding, not the people.

AUDIENCE SPEAKER: That depends on the model of the ISP.

OLE TROAN: In a recent case, we don't support old firmware. Buy a new box and that's up to the user in the end, so who knows? Right. This typically happens  it's a big big problem and it would only happen when you change service in some way, right. I want to have higher bandwidth or I want only you know small  my v4 address, then possibly something happens, but it's hard. I mean we have one solution, we could tunnel to your hosts, do you want that? By passing the CPE. I mean, we have massive amounts of innovation in that space, too. I think it's a terribly bad idea due to Teredo.

ANDREI ROBACHEVSKY: Thank you and I guess with this I'd like to thank you our panel for those interesting insights in this segment. Thank you very much.


CHAIR: That was that for the afternoon session. So it's time for the coffee break and after that it's the v6 session here and there is someone else in the room next door. AntiAbuse in the room next door.

(Coffee break)