Archives
IPV6 WORKING GROUP SESSION
14 May, 2013, at 4:00 p.m.:
DAVE WILSON: People of earth, your attention, please. We have a fairly busy agenda for the second IPv6 Working Group session this afternoon so we will get straight into it. Thank you very much. Aoife is doing the stenography this afternoon. We have Marco doing the job of scribe and Marco, another one, who is taking care of questions from the chat room. One quick announcement, I have been asked to say the PC elections close immediately after this session. So thousand is your last chance to get your votes in.
That old chestnut of the Working Group Chair selection, if ‑‑ the agenda first. We will go through the Working Group chair, Geoff is speaking about IPv6 who is and who isn't and then we have got a bunch of lighning talks to get through and we will get you out in time for 5:30 with a discussion session on v6 deployment, which will be led by Jen.
Please rate the talks, that is going to be very, very helpful for us and for the rest of the Working Group Chairs and indeed the PC throughout. You do not know how helpful this is, it's helpful to the presenters, getting feedback, are really gives us a picture of what is going on.
So, IPv6 Working Group Chair election, we have been over this a number of times in every single Working Group. And the reason we are doing this now here instead of in London, is the IPv6 Working Group got special dispensation to hold off a bit because it was changing all its Chairs. Now that we are here and settled in for a few months we are able to get started on that. And that gives us a distinct advantage because all the other Working Groups have done the work, so maybe we can just copy their homework. That is exactly what we did. We picked the procedure that was put forward for the Anti‑Abuse Working Group, because it looked pretty good and seemed to fit with the principles we were looking at. I will go through the principles and the actual text is on the mailing list and has been now for a couple of weeks.
Puts out a three‑year term for Working Group Chairs, who of course, can step down voluntarily at any time they wish before then. When it comes to selecting a new one, it mandates a call for candidates a month before the meeting itself and the selection takes place at the meeting. If there is only one candidate we will ask people to raise hands and say it's OK. If it's not possible to have ‑‑ ‑ to do it by acclimation, it will be secret ballot. Co‑chairs can be any number up to 3 and higher than zero. Anyone physically at the meeting has a voice. If we find ourselves in a situation where the Working Group wishes to remove a co‑chair there is a procedure for that. You can request a vote of no confidence. The procedure ask for the issue to be addressed first first, it can't be done a week before the meeting you can request secret ballot and anyone, again, anyone at the meeting can vote on that.
If we find ourselves in the situation where there is no chair, then we just find a candidate in the meeting or candidates, and there is an election right there and then. But then of course, they are expected to go through it formally either themselves or someone else in time for the next Working Group session. And that is it. There is text there about a staggered introduction but because all three are new in this case it shouldn't apply. We want to make sure all other terms don't expire on the same day in three years' time. I am sure we will do something about that. I have done at least and popped out the other side and been adopted it. Are there comments about that that haven't already been brought up in the mailing list. I see a few thumbs up ‑‑ three, two, one. In that case can I ask people to raise their hands as our Chair selection procedure. If you are in favour, raise your hand. Thank you. If you are against, raise your hand. I see none. We can consider that adopted, thank you very much.
(Applause)
Next up, I believe is Geoff, isn't it? Geoff. Thank you.
GEOFF HUSTON: Hi. I am with APNIC. And this is some work I have done with George Michaelson on measuring IPv6 and I want to give you an update on what we see and why we see it and perhaps introduce a little bit of economics into the debate about the underlying issues that are going on here at the moment.
This is the count that we see looking at v6 across the last couple of years. It's not quite the same as Google's. And part of it is we have tried exceptionally hard to reweight these numbers to actually account for the fact that the world is not Google‑coloured, oddly enough. I know it will come as a surprise to some of you. There are a lot of users in China and India, and their intensity of use of Google differs from, say, you here or in America or other economies. When you take a raw account of folk who go to Google's web pages, or as we get, a feed of data coming from where Google have placed on line advertisements, you are not seeing a uniform sample of the world; you are seeing a sample of the world of folk who use Google a lot, and it sort of misses folk who don't. And so we have done in this graph here is actually reweighted the numbers we get, based on the best estimate we have of users per country. Now, you and I both know that the stats coming from the ITU are pretty crap, but no one else publishes any data whatsoever and it's kind of the best we have out there. And so what we do do is weight the sample numbers against the population corrected ITU numbers, because the world keeps on having more people in it, and then pushing it back out this way. And so there are a few sort of issues about this. There is slightly more usage on the weekend than weekdays, something about corporate firewalls or whatever and there is a very small number of folk who are v6 capable, they will get it but when you give them the choice, they will not necessarily use v6. And if you have looked hard at the way your MAC works if you are running MAC OS, that is probably you, because MAC OS doesn't have a strong preference for 6; it uses a rather complicated algorythm that takes an estimate of what the round trip is likely to be to somewhere it's never got to and makes a decision based on protocol. I think I just said random, it makes a random choice to which protocol to use and uses it anyway. So that is why that is that ever so slight difference.
Last year was kind of a good year, if good is, you know, a rise of 1.8%. This year hasn't been a good year because it's been a rise of a lot less than that. If it doubles every year, 1 .8%, 3.6%, by 2019 we will be there. Don't get too worked up because it's only doing 2% a year and if that continues, we will all be dead apart from you very young folk out there, because it's going to take until 2067. By then, your carrier grade nets will be nuclear powered because you are not going to survive any other way. It ain't going to work.
World map, we have all seen world maps, that is another one.
That is little more interesting because the number of countries above the average is actually really quite low. And there are some interesting ones like Ecuador. And Peru right up there, almost as if Telefonica using Peru as a lab. Nobody here is from Telefonica, are they? Sorry. The number of countries who are is quite small, some countries Belgium, an astonishing uptake, Germany too 19%, that is really quite amazing. We can drill down a fair way in the Netherlands, something bad did happen over Christmas, things got turned off a bit, I don't know why and then you turned it on again, thanks. There is no more roll out since December, not very good. So you are back to where you were, it's hardly exactly progress going on here. .and there is the US, and it's certainly a more optimistic view and that is largely influenced by Comcast but there is something else going on, too. Because what we are able to do is sort of rank these providers by their number of customers. And there is no doubt, in the United States Comcast is by and large the biggest provider out there in access, AT&T, MCI, these are all familiar names to all of us. So, that is, you know, the top 30 providers in the US. Who is doing v6? Seven. Not 30, seven. Just seven. This is kind of typical for most countries, that everyone isn't doing it, just a few folk are doing it. And then I started to think, well what about the rest of the world, how typical is that? How many v6 enabled ISPs serve the world? That is an interesting question. Because if around half of the Comcast users do v6 and maybe that is 20 million‑odd Americans, oh, AT&T, maybe that is 16 million; Deutsche Telekom, maybe that is 6 .5 million folk using v6. And so this is kind of the v6 population ranking by provider. And certainly, folk that would you expect to see there, are there. K D difficult. Who has been doing stuff in Japan, 26% of their network is running v6 with users, problems at and free been doing it in France for forever, and Telefonica Dell Peru, they are around with around one‑fifth. So this is not unusual. But let's add them up. And if I add them all up, 30 providers serve 94 percent of the v6 population. How many networks are out there? 50,000. So, who is part of that 30? One. Good. Or two. Three. The rest of you are part of the 49,700 or so or 9,030 who aren't because those do it and literally everyone else isn't. Which is interesting observation that this is a very small bunch of people doing a lot of work and you are leaving it to them. Because you are not doing anything that is visible. That's getting interesting. So, yes, 94% of the world's v6 sit behind 30 access ISPs. Wow. Why? Why don't 100% of Comcast users run v6? Because they Co. And why don't 100% of horizon mobile do v6? Because they could. Why is it only something like Google that manages 87% penetration rate in their network? Because they gave out new CPE, they didn't go and buy it at some local corner shop or Amazon buy some crap made by you know NetlLink or Belcon years ago and expect it to still work. CPE is a massive problem, not only is it the bot army that basically drives most of the world's malice this days, it also is the bot arm, that stops you from doing their job. And it's so weird because it's only 50 bucks a unit. Quite frankly if you are in the ISPU business the temptation to simply give everyone an OpenWrt box that actually has software that works, should be overwhelming, so why aren't you doing it?
Why do you do your best job and have users completely stuff you up by buying crap CPE? Why do movie makers make really, really excellent movies on phenomenon digital 7 K stuff and you watch it on an iPhone? Same kind of problem, they do a great job and then you stuff it up. On the ISP business you kind of get it to the front door and let customers buy CPE and ruin all your work and investment, you are nuts, completely crazy.
So, on to the next slide.
What can I say? This is a slightly different view because now I want to look at, we know there are 30 doing 94% of v6. But this is networking, and this is a network defect. Your job is to connect your users with everyone else's users. You have got to do much the same thing together. Because if you go off running an all Apple talk network you are dead. You can't run a a different protocol to everyone else. Out liars in this business die. So if you jump too soon or too late, you die. You have got to run with the herd. And this business works by mob menality. You do whatever else does, literally. But that is not quite true. You do what the influences do, you follow leads. So who provides influence in this business? Who drives everybody else? So this is my first modest attempt to try and figure out who drives this industry.
So it's the 30 largest ISPs by user count.
I told you China was big, 300 million people sitting behind China.net backbone and another 207 million behind China169, there are a lot of folk in that country, and that is the evidence.
Also in India, which is number 3, BSNL, there are a lot of folk living there and when they start rolling stuff out with 40 dollar handsets that is a lot of people and that is what we are seeing.
So the largest one in America Comcast comes in at number 4 by that ranking. I bolded up in that top 30 the five who were doing six, Comcast, AT&T Deutsche Telecom, TNnet in Malaysia and KDDI. You kind of think well, maybe that is the folk who should be, maybe they are driving this business, because just five. What a strange font ‑‑ it's the mouse ‑‑ just five of those top 3 largest ISPs have any significant v6. Now, those 30 actually service 42% of the world's population, down the bottom. This is a really skewed business, if we destroyed we were destroying the big telcos we have rebuilt them again. But perhaps we are asking too much of BSNL, perhaps we are asking too much of the developing world so maybe we should look at value. How do you value an access business? One way, which seems pretty logical to me, is by the purchasing power of your customers. The more money they have, the more value they have, the more value they represent. So why don't we multiply your users by the GDP per capita of the country you are in, that might get interesting. And let's call that net user value. And this gets really quite interesting, if we multiply that we got to 30 providers do 43% of the value of the Internet and just six have v6 deployment. If all those 30 did their job, the rest of you would have no choice, because they are the influences. And that is the list.
Comcast, if you multiply things out by GDP, are clearly on the top of that list. Their value or the value of their users is intensely large. China.net still has a lot of people, I know the GDP is lower, they still have a lot of people. There is another American in the third slot, AT&T, NTT in Japan, Deutsche Telekom in the sixth, you look, there are so few there doing 6 and that is 43% of the value of the Internet, the value of the users of the Internet. And if you are looking at who we should be beating up on, NTL Virgin Media limited, I think they are unit these days, orange in France, the old France Telecom, they are not doing anything and maybe they should, because if all these folk did it, the rest of you have no choice, literally no choice because these are the folk that drive the business. This is the influences of the herd. Because this is where the money lies.
So, the one thing that is kind of interesting about this, is that those 30 largest providers account for 94% of the user population and 28% of the value who are doing v6. So, v6, right now, is actually bigger than we thought. There is a lot of value in v6 right now, because some of those providers, particularly Comcast, represent quite a stretch dream of value. And what should come out of that, if I rank it this way, v6 currently accounts for not just a little over 3% of the world's users, but actually accounts for 7% of the network's value. Because what we have done is provision v6 to the rich. And that is not a bad thing. It's not a bad thing if you are trying to influence everybody. Because it's those buying patterns that matter.
So, what do we ‑‑ I don't agree with Google, I think if you take the Chinese into account it's about 4%, a little under. The significant volumes in about 20 economies but realistically that is very bulked up business, it's no longer the little entrepreneur driving it ‑‑ a subtly 30 service the bulk of the world but those 30 who service the v6 service 28% of the world and that is 7% of the world's value. If you are kind of looking at who you should be saying, why aren't you, why aren't they doing it? You know, I will be looking at the folk on this list who aren't, and kind of wondering why the rich businesses and they are business depends on having a future. They are the folk with the rich customers. And they are the folk who should be up there with a much denser list than the six or so I can see there. That is the update from May. With luck and with hope, that list will be longer by October. If the economics is right, yes, but whether you guys behave as economic rationalists or not, I will leave that up to you. Thank you.
(Applause)
DAVE WILSON: Any questions or comments for Geoff? Thank you.
JEN LINKOVA: Did not say anything about getting rid of the chair because of using ‑‑ this is going to be exposed to my slides for a while. So, lightning talk. Slides which you have seen in the breaks then about wireless network here were not exactly accurate because there is another wireless ‑‑ actually two wireless networks which I believe we all should be using. It's IPv6 only network available and still will be available until the end of the meeting, so you can see SSID, you can see the password which also is an SLA for that network. I would really encourage you to use it and tell me what is wrong. But I also would like to thank RIPE NCC, particularly Marco, Colin and Menno, who made it happen. And for the Cisco systems who helped us with NAT 64 solution so you guys could access those poor resources which don't have IPv6 yet. And v4 only. Just big round of applause.
(Applause)
So, I just wanted to let you know that v6 actually works, but before I tell that you it works, I realised I probably need to explain to some of people in this room how exactly this network works, because I had a conversation yesterday in the lobby so I had to explain kind of general design.
So, it basically for v6 to v6 traffic is simple, go into Internet and so on. If your machine has only v6 address and you try to access example.com which has only v4 it's going to be a problem. So when your application ask for AAAA record for example.com, the magic DNS 64 server finds out out that there is no v6 address for that resource so it supplies you with crafted special v6 address which actually includes and code ‑‑ address of the resource and the magic router provided by Cisco systems does the translation. So for your device just v6 to v6 traffic and for v4 Internet is just normal v4 packet between NAT 64 translator and remote site.
So, quite simple.
Because we did not actually communicate it properly before the meeting, so unfortunately not so many users on that network. Let's see actually how many. Thanks Colin for providing me statistics. So four‑week 85 unique clients (full week) which is about 5% of the total unique clients seen on the conference network and the peak we reached on ‑‑ yesterday, I don't have statistics for today, so yesterday it was about 4% of total clients, Wednesday graph like this about 5 p.m., we got about 30 unique clients in the same time. Apparently I can see some left over at 8 p.m.
Actually, I think it's the first time when we try to get data on what is exactly might get broken, we use it. So thanks for ‑‑ to everyone who replied to me and told me that things are working fine or if things are actually broken.
I am collecting a kind of ‑‑ I hope next time we have least‑of‑known issues for this network.
But in general, could I see two types of problems: First, your application is assuming too much, and I assume that OK, there is no v4 address, there is no internet connectivity so i‑message from time to time, not all the time, on iPhones, try to fall back to SMS assuming there is no connectivity. Actually ‑‑ a couple of years ago the bug was closed I still need to follow up on what is going on. It's not happening all the time. Similar issue for Air mail. Another problem is when your application is not asking for AAAA so DNS 64 could not do any magic tricks, ask for a record, receive a record, could not reach IPv4 address. It's happened for drop box, the stop application as ‑‑ for Microsoft link and for some RFC ‑‑ messages clients.
So please use it. Please let me know if it works or not, because we would like to see more users, more problems and probably at some point one day oh, wireless networks should be just v6 only, but to get there we need to fix all those problems which I mentioned and others I have not mentioned because I don't know about them yet. That is it from me. Questions?
SPEAKER: Aaron Hughes, 6 connect. If you wouldn't mind sharing your configs for the 6to4 for Cisco and DNS servers could people can set up their own experimental networks and provide feedback.
JEN LINKOVA: Good idea. I think it's more question to RIPE NCC who are running this thing.
What I would like, by the way, to mention, we got a small group of volunteers to probably write a kind of best practices for running v6 only networks. We had some discussion this week, and I do hope that we might get some information to share on the next meeting. Because, apparently, yes, there are some things you should do, things you should not do when you are doing v6 only like conference networks, yes.
SPEAKER: One other suggestion, for the next RIPE meeting might I suggest you make the RIPE meeting the primary one 6to4 and make a secondary network.
JEN LINKOVA: We have had this conversation many times with RIPE so it's RIPE NCC call. I would love to do this, but I am pretty sure to do this we still need to fix all those lovely applications which are broken, so if you have support contracts with some of those companies tell them this stuff does not work on your network.
MARCO HOGEWONING: If I can, what Jen said, we would love to show up if we are sure that you and all the people around you attending these meetings, all 700 don't come and complain to us if things are broken, so indeed, like Jen said, fix the bugs and we are happy to show up.
Blake Willis: Aaron, to answer your question, I believe it was at the RIPE at Athens, Andrew Yourtcheno gave a ‑‑ if you look for the Cisco specifically iOS XR ‑‑ I am sorry, iOS XE, NAT 64 configs and search for Andrew's name you will find the presentation and it tells you how to do that.
Another question for you is, do you have plans to make a name and shame list of applications that don't work on that network that could be made highly public visible somewhere?
JEN LINKOVA: There is ‑‑ there is WIKI page which lists some known issues. I am going to come ‑‑ make a list based on this week, to find out what are those problems might be solved before the next meeting. So I do hope that by, what is it going to be, November, we might yet get a list of applications and I do hope they are going to be shorter than list I have now, but who knows, maybe it might be longer. We will see for some bugs like ‑‑ we might just need to talk to vendors and get them fixed because probably nobody complained to them. I suspected for example, for Dropbox not many people actually tried, so I reported this, we will see how it goes.
SPEAKER: About making the primary network v6 only, I think that could probably happen if you implemented 464 on that, obviously talk to Andrew about that later. But it would require an architectural change of what is in place now. We can take that off‑line. It is IPv6 only on the client. Anyway, sorry, we will talk about that later.
SPEAKER: IPv6 on one of my phones and I was planning to turn it off and it didn't break, I didn't really notice I was using that, so I guess that is a success. Still for those folks who found something doesn't work, maybe you need a little paper signed by the RIPE NCC taking ‑‑ not that it will help it, but like a journal they can come in and write in or maybe leave e‑mail, something doesn't work or ‑‑ so people can just kind of lodge not just a complaint, keep a record of what specific ‑‑ something very specific that people can add to like other simple resource.
JEN LINKOVA: So far the reporting work quite simple, it wasn't so many issues.
SPEAKER: If you didn't scale, right.
JEN LINKOVA: I do hope not going to see so many issues so, it's not going to scale. I hope we see less and less issues. If stops scaling I will think about how we can make it better, so far I don't think it works spending time on inventing problems when I could do it faster.
SPEAKER: Sharing specifically DNS server access and translation stats, it would be interesting to see how these issues play together.
JEN LINKOVA: You mean you want publically accessible 6to4 ‑‑ it actually was one of the topic I would like to bring at the end of this session, so we might discuss it later. Let's just postpone the discussion.
SPEAKER: You are thinking about that, cool.
JEN LINKOVA: I want to see opinions on this why.
Marco from RIPE NCC: I have a question from the chat and Sasha, next meeting stateless ‑‑ NAT 64
JEN LINKOVA: I need to address end devices in specific way. So, I need state because if your device using like privacy extensions, how ‑‑ I am not sure how can I do stateless.
SPEAKER: Stateless is one ‑‑
JEN LINKOVA: It doesn't ‑‑ I am pretty sure ‑‑ I am interested to see how it works in real life and useful for people, right, and I am pretty sure stateful for people for conference network enterprise network, might be much more useful, actually, than do stateless. Stateless is good idea for data centres when you can just, how you number your devices for uncontrollable devices my opinion is stateful works pretty well and, you know, it works, I don't want to touch it.
SPEAKER:
JAN ZORZ: Firstly, the answer to Dimitri's question, if I understand he asked for publically accessible NAT 64. I am running three of them, they are publically accessible whoever wants to use them can just set the DNS server to one of three that are listed on the page. And I think Lorenzo used it to test the 464 Etisalat implementation.
And second thing is, maybe, Jen, you can ask your colleagues internally when they are going to implement the 464 ‑‑ the C lath activation, not only when you do IPv6 over mobile part but when you activate the IPv6 only wireless connection. That would help.
JEN LINKOVA: I am not ready to comment on the second part because I have nothing to do with all this end user devices. I am a plumber, I am looking after pipes. But thanks for saying that you have publically accessible solution. Now, you probably might need to scale it after we advertise it in this room and the room is full. Big boxes. Great. So, please, next meeting, come and tell us how much traffic you see now.
JAN ZORZ: OK.
JEN LINKOVA: Please use it. I really would like to see the reports.
DAVE WILSON: I actually have a question on the topic, but it's for the Working Group as opposed to just yourself, Jen. And I am not going to hold things up trying to get into this discussion now because we have to stay on track. I do want to ask why are we doing this? Because when the subject come up with the RIPE NCC a few weeks ago they asked us Working Group Chairs, it's going to take sometime and effort and money to make this v6 only network work. Is it worth it? And I went, I have no idea. I don't know what deploying it proves. I can take some guesses but they are only guesses.
So my question for the Working Group is: If we have this network here running at the RIPE meeting what are you getting from it or what would you like to get from it? Can we make it primary network next time is a great question because there is answers to that and we can start thinking about what those answers are and I think it's worth enumerating that to see what the users of this network, which is you guys, not really me, what what would you need in order to make this your primary network and start thinking about how do we get from where we are to where we need to be. I would love to see that discussion on the list.
JEN LINKOVA: If you have any type of feedback just send e‑mail to the mailing list or just to Chairs, because we really like to hear it.
SPEAKER: I think maybe instead of calling it IPv6 only, just call it RIPE ‑ IPv6 only, just kind of, you know, in the middle of experimental there is a slim difference.
JEN LINKOVA: First make everything work fine and then we can thinking about polishing names and titles. I am a simple network engineer. I want to make things work. Marketing is another thing.
(Applause)
DAVE WILSON: We are right on time. Next up is Nathalie with lightning talk, lost stars, thank you.
NATHALIE TRENAMAN: This presentation ‑‑ this presentation is ‑‑ my name is Nathalie, I am IPv6 programme manager for RIPE NCC, and this presentation is called lost stars. It's not a about David Hasslehoff. It has actually something to do with IPv6. And some of you might have heard ‑‑ who of you have actually heard of IPv6 RIPEness that we have? Oh, I am so happy to see this, that is a lot of hands. I was about to explain the whole thing in in‑depth, now I do the summary, for those who don't know, it's a methodology that we use to measure IPv6 among our members, so we have around 12,000 members in RIPE ‑‑ of RIPE NCC, and basically, we give them stars, hence the name of the presentation, if they have certain achievements unlocked with IPv6. So, for example, you can get the first star if you get an IPv6 allocation, then you can get additional stars for Reverse‑DNS, route object and if you actually announce your IPv6 prefix in global routing table. And if you have four stars, basically you get listed on RIPEness.ripe.net as an LIR and we still do send free T‑shirts to motivate people to get the four stars. In case you have the four and you did not get your T‑shirt yet, come and see me afterwards, we are happy to ship it. That was the pitch for the RIPEness.
We have the fifth star before somebody wonders, Nathalie there was something with a fifth star, yes, there is ‑‑ this is, by the way, the chart of the distribution of the stars amongst our members. As you can see, still, 25% of our members don't have IPv6, and 21% has all four stars. And a subset of that has the fifth star.
Now, you can gain stars by getting this information ‑‑ by getting these things, getting your allocation, setting up the Reverse‑DNS but you can also lose stars. And this is what I want to talk to you about because this is kind of important and I think it's serious. The R and D team, our research and development team, was doing some other research and as a by‑product, they discovered that over 400 of our members stopped announcing their IPv6 prefixes in the last 12 months, and that is close to ‑‑ well, that is about 4% of our member base. And they just not stopped announcing the allocation and started announcing more specific or something or vice versa, they completely were lost in the global routing table. And then when we start looking at these numbers, surprising part was that 70 LIRs stopped after more than three years of announcing. So, that is quite significant, and the numbers, there is no trend in where these members come from or how big they are; they are small members, they are big members and to show you a bit, I don't know if you can see it, otherwise you can download the slides, you can actually see that some of them even decided to stop announcing their prefix after seven years. And these numbers are ‑‑ I was actually surprised to see this high number of members stopping announcing IPv6. And then I thought, what is going on here. This is important. And I would like to know what is going on, what is the reason. Maybe they are getting ready for production. They were just announcing it for tests and now getting ready for production and we will see them soon but maybe not. Maybe there was a TCAM issue in the past and they maybe they did some troubleshooting and switched off IPv6 because it wasn't in production anyway, maybe there was ‑‑ we don't know at this point, and I would like to know. And, offer help if needed, maybe they need training or something else, maybe I can speak up for them and bring them in contact with people that can help them further. Anyway, you get where I am going.
The thing is, as you can understand, it's contacting 400 members, might take sometime, some effort, and I was just wondering do you think it's worth it to do this or do you see the use or maybe not? So I would like to invite you to come to the mic and tell me what you think about this. Thank you.
SPEAKER: I have one idea about rather than ‑‑ CCS, France ‑‑ so I have one idea about how ‑‑ why do some provider may retire their announcement, for example, some providers have a /29, which is more than a /32 and they may have decided that part of that /29 to be used internally, not to be announced over the internet and still leave part like /30 or /32, to be announced over the internet. So for me, I thought about this situation in a previous company and it may be a legitimate way of doing it. So, if you look after more specifics, what is the result.
NATHALIE TRENAMAN: No, like I said earlier, it's not more specifics; they are completely gone from the global routing table. Thank you.
SPEAKER: Peter he is letter: First comment is yes, knowing why would be very interesting. Second is I wonder if there is a correlation with those announcements going away near the 512 on IPv4 network when we got to that point and TCAM tables start being balanced in weird ways.
NATHALIE TRENAMAN: Thank you very much.
SPEAKER: You should or somebody needs to understand what is happening here. Regarding paperwork, maybe some ‑‑ disappeared, bankrupt, did you check this paperwork?
NATHALIE TRENAMAN: They are open and existing LIRs, so they didn't close or merge, yes, thank you.
SPEAKER: IPv4?
NATHALIE TRENAMAN: I didn't check because that is not part of the RIPEness stuff to see if they also had IPv4. I am not too sure how that would be relevant for this research but I ‑‑ if you think that has value I would like to look that that, yes.
SPEAKER: What I meant is that maybe they stopped announcing everything so maybe they also stopped announcing IPv4 so it would be good to exclude that.
NATHALIE TRENAMAN: Ah, no, I didn't check that. Good one, thanks.
JEN LINKOVA: Thank you. And now, Jan.
(Applause)
JAN ZORZ: Hello. I am from Slovenia and this will be very short. Basically it's a question. But first, some history.
In the past, we, us, our Working Group and we as the authors, removed one of the speed bumps in deployment with publication of RIPE 554, that is the specification for IPv6 requirements when you are ordering the equipment, this is now widely used by vendors and by basically by lots of people that are buying the equipment. That was actually told to me by the operators and by some governments that they need. So we removed this speed bump.
The second one was published in cooperation with the pick‑up task force in IPv6 Working Group since the last meeting, it became RIPE‑631. Why we did this, because we heard from the operators that the next speed bump, that their help desks don't know anything about IPv6 so we put together the troubleshooting for the help desks, I also hear from the operators, many of them are implementing this in their help desk and also using the tool that comes with it. So, this is now done and my simple question to you is: What is your next speed bump? What do you think we should work on? What would you need to advance your IPv6 deployment to your residential customer and basically enable it by default? So please, come to the mike, the more suggestions that we get, easier we can do something about it. Thank you.
BENEDIKT STOCKEBRAND: Another Working Group Chair. From what I have seen over the last, what, almost a year, why do IPv6 deployment at enterprise rather than ISP or content provider perspective, there is very, very little best practice from an end prize point of view. Good news for me because some customer hired know bring in ‑‑ if you want more enterprise to see get interested and put demand ‑‑ or create demand for the ISPs to provide IPv6, I think they are pretty important to take care of, even though they are not that strongly represented in the RIPE community, but that's, in my opinion, one of the things where things actually have to happen. Other thing; promote DS‑Lite. Because it hurts so badly and it's sometimes even hurts, you know, top managers with their home internet access that they realise something has to happen somewhere.
JAN ZORZ: Are you OK?
BENEDIKT STOCKEBRAND: I am, I am, I am actually dead serious about that. For last about year‑and‑a‑half, in pretty much every single training I have done on IPv6 I had somebody asking in the first five minutes do we talk about DS‑Lite? Yes sure. So what cable provider are you with? Cable Deutschland. And yes, I know about the problems and I can tell you how you fix at least some of them, by switching to IPv6. And those people actually are very important because they spread the word in their enterprises. It sounds completely crazy but that is how it is. And if people only go to the doctor when they can't stand the pain any more, well, we will have to hurt them.
JAN ZORZ: Yes, I am with you.
BENEDIKT STOCKEBRAND: It's not as funny as it sounds.
JAN ZORZ: Thank you. So, I also got one suggestion from somewhere saying lots of people don't understand the mess about the stateless and stateful autoconfiguration, can you please document that so people would understand what to use and what to do. Do you think this would be maybe useful ‑‑ I don't know.
JEN LINKOVA: You mean dep ‑‑ yes, let's do this. Promote the slide and deprecate ‑‑
SANDER STEFFANN: I want to second that comment about enterprise. I think the ISP networks have already progressed a lot, I think enterprises are a big problem in all kinds of fields and not knowing how to deal with SLAAC and the management not knowing how to deal with routing and update their home‑made software, there are many, many areas inside the enterprise that need help, I think.
JAN ZORZ: OK. So who would like to help with this?
JEN LINKOVA: We have got a volunteer. Great.
SPEAKER: What we see ‑‑ what I think we see in big organisations like big banks and ‑‑ big companies, is that they have always worked behind NAT and behind big firewalls and all that kind of stuff and they are kind of afraid to basically open up the network for ‑‑ because ‑‑ they think everyone has to be public IP space et cetera, et cetera, et cetera, and I think that is one of the main issues for big companies to roll out. And then, one second thing: Spotify does not work on IPv6 only.
JEN LINKOVA: If it was a bug report please accepted me an e‑mail, otherwise I will immediately forget about it.
SPEAKER: Erik: Regarding the enterprise deployment, just remind you that together with couple of friends of us we wrote an RFC on this, enterprise deployment guidelines that includes some words at least presenting the pros and cons on DHCP V SLAAC and ULA and that is not useful ‑‑ so it may be very useful to restart something here.
JAN ZORZ: Maybe we can translate that into the language that people understand.
BENEDIKT STOCKEBRAND: Once again on this. RFCs have some sort of reputation that they are pretty much incomprehensible to people in enterprise environment, so just it being titled on RFC and not dated it refers to whatever year is scaring a lot of people away no matter how much effort you put in that especially at an enterprise level. In ISP context you have a lot of really top qualified people running around. In an enterprise level you have to deal with the others, sometimes. And RFCs are frequently completely ‑‑ they are just ‑‑ it's a year and they stop thinking about it. RIPE documents are slightly better but even then we will still have to spread the word because RIPE is all ISPs and so on and but if we can get something like this done and get it promoted, I think we can actually go to the one group involved where IPv6 is lagging behind after the ISPs, after the content people.
JEN LINKOVA: I am glad that we actually moving to the very last part of this session, so I can give you five more minutes to think about what you might say, because we are going to have longer discussion on this topic, and meantime I think Nathalie ‑‑ Zorz am I dismissed?
JEN LINKOVA: Sorry ‑‑ I am just trying to think how shall we continue this discussion on what we have to do. Right. OK.
JAN ZORZ: Thank you.
(Applause)
JEN LINKOVA: So, Nathalie, welcome.
NATHALIE TRENAMAN: There is actually a little reason why Jan and I swapped slots on the agenda, because as you see on the agenda there, it was me and then it was me again and then it was Jan. But, there is a little reason that I wanted Jan to do his presentation first and ask for input and then I would ‑‑ I don't convenient slides; I just wanted to tell my experience after talk ‑‑ what the issues are at the moment, after doing a lot of IPv6 training courses and running into a lot of the same questions and every time we do a training course there are always two or three people in the room that say my CEO wants to know what it would cost to do IPv6. That is always the same, what would it cost. And I thought, OK, maybe we should do a document with economists, economists and ‑‑ economists and people who have more an idea about money. And then I talked to some other people with an MBA and stuff and they said you need to know what the cost drive is for IPv6, you need to know what ‑‑ where the money goes in an IPv6 deployment because then you can find out how much money and what ‑‑ and what deployment would cost. So that was a valid point. So I went asking, every person who did an IPv6 deployment around me, where did the money go for your IPv6 deployment. And they all said the same thing: Access providers, content providers, CDNs, all of them, said it went into the man‑hours of our developers, to adjust and write new code because we have a lot of IPv4 late ‑‑ we have a lot of IPv4 related code, we have billing software and we have got got internal/external software, etc., etc. So then I thought, OK, if that is where the money goes then maybe we should develop something and start talking to these developers, because apparently the ISPs now have the sense of matter to do IPv6, well according to Geoff, not too many yet, but still, hopes a lot of them are cooking in their network kitchens. But indeed, after talking to a lot of people I am not sure if the IPv6 is very, the sense of IPv6 is very much aware in the developer area. So, maybe we should put some time and effort in there and I would like to know what you think about that? That was it.
BENEDIKT STOCKEBRAND: The situation is worse. From what I have seen with one customer, you start to think about IPv6 and you have got, like, 6,000 software packages you are using and it took me a couple of days to sort these through. That is bad. And then the ‑‑ then there is the in‑house development software. And a number of things can happen. They have been using the wrong programme language and you have to fix all sorts of stuff to support IPv6 and then decide, OK, the ‑‑ what we use doesn't support IPv6 at all so we have to switch or whatever. And I have seen at least one case where they have been using Java, which you have to fix to support IPv6 Java programmes but they have been using a library for all their network access. And the question was, does this library support v6 or not? Fortunately, it did, but that was never why they picked that library. So, the question is this a major effort, we have to rewrite lots of our software or does it just work, is largely a matter of luck, of sheer luck, of picking the right libraries and right languages whatever and that what makes, from what I have seen in enterprise kind of context, just figuring these things out and figuring out how much of an effort it is about half of the work of deploying v6. It can be quite easy and it can be man‑years and man‑years of fixing your software, but it's completely unpredictable. So that is what is really scaring. If you ‑‑ this is going to take that long and cost that much and I need these resources. I need two or three man‑years to figure out, that scarce me.
SPEAKER: When I talked to my enterprise customers, I tell them I have no idea how much it costs because I have seen databases of IPv4 addresses as database index that might mean complete redesign of the database. But I start with telling them, hey, start saving money today, because all new software that is developed has to be IPv6 ready so go to your legal department and make sure that in all contracts you have from today, IPv6 is in there, and make sure that you can test the software, that the software development company maybe internally or externally is delivering to you. And this makes sense to most of them. They say saving money from today, well, it's a good idea, would be hard job to go to legal and explain what IPv6 is. But yes, we start, and then you have evolution of software, you turn down old software and have new software and this makes it less hard to do the transition because in your development cycle you can fix a few of those problems. But, saving money is something every institution, every organisation wants to do so they are going to listen to you. You don't tell them how much they have to spend.
NATHALIE TRENAMAN: I agree, but the software ‑‑ that part, the purchasing part is basically already in RIPE 554, for software are as well. I am more thinking in how can we help the coders there with their software and IPv6. But, I see your point there with the cost issue, with saving money.
SPEAKER: For the coders, I don't know what we can do for them really, we have to educate them. But we have to educate them in networking and then we can educate them in IPv6. This should be a second step. What I recommend is that companies have to build a test lab, anyway, so invite the developers in this test lab, make it feasible for them to do some testing and if someone comes, if a developer comes to you and say, I want to test IPv6 and you have to say, well go away, I have no place where you can test it, then you are lost.
NATHALIE TRENAMAN: Thank you. More thoughts on this? Experiences?
JEN LINKOVA: It's nearly hard to be almost the last topic on that.
(Applause)
So, we have almost no slides for the last part, supposed to be open mic discussion which we have started already. I have just a couple of graphs with 7% of you doesn't like. Probably it's not 7, probably 3. I just like the form of the graph, even if you don't put numbers, it still looks like. And the totally ‑‑ you could not see with country for European countries, right, showing some deployment. Because those three or six or 7% it's because some countries has 20 and some countries has zero, right. So, question ‑‑ actually, three questions:
So in this room, who could not deploy IPv6 now but would like to do this? Come to mike and tell me why. And tell me what could we as community, as Working Group do for you, how could we help?
PETER HESSLER: This was for a previous company that I worked for. We were a content provider and there was two major blockers for doing IPv6 deployments outside of our control. First one was advertising networks, the ones we were using had zero support for IPv6, zero interest in going to IPv6, so if we went to IPv6, we would lose all ad revenue there. The second issue would be the monitoring services, kink come to doesn't support IPv6 and we would lose a lot of our extra monitoring abilities.
JEN LINKOVA: I got a feeling from all this discussion that we have a kind of broken communication chain between network people and developers, it's network people who do care about IPv6 and developers who block the deployment have no slightest idea about this v4 and v6. So I am not sure. It's definitely an education problem, right, do we need presentations or documents? But I, for example, do not have much visibility about developers coming, from anybody could suggest talks, where shall we go, with some education programme or presentations on conferences or I would love to hear some ideas.
BENEDIKT STOCKEBRAND: And again about enterprises. One of the big problems I have seen with various enterprise customers is, that they have some sort of highly specific software they bought somewhere for their particular business. And the more particular the stuff is, the more arrogant and incompetent the developers tend to be. If you are a lawyer specialising in whatever law and the border crossing between Germany and Belgium, for example, there is very likely exactly one software or at most one software that does the job and the developer honour pick that because knew it would never be able to compete with any real developers. So, they chose that. And one of my customers told me of one experience basically, they asked this vendor so what about IPv6. The answer, you don't need that. Well, yes we need to for so‑so, believe me, you don't need that. And that was the only answer they got. And they couldn't possibly switch to any other software because there wasn't any. And that is when things get ugly. So again, education is important for the developers but it's also important to put pressure on them to make them realise you are eventually losing money on that. And that is going to be a real tricky one. I don't have any idea how to do it. I just understand this is getting more and more important every day.
SPEAKER: Sylvia from Switzerland and I have been working on this question for quite some time, and I think it's mainly an awareness problem on the developer side. I am not a developer so for me sometimes it's hard to counter‑argument when I talk to developers because all the developers I tried to wake up in Switzerland they told me, we are fine, we have our frameworks and they take care of the network layer and we don't need to worry about that. I felt this is not true. So I spoke, there is not really too many developers in the world that are aware what have it takes to develop for IPv6. So actually, I am sort of, at the moment, evaluating all my resources and my network and trying to sort of create an interface and I think it would be much needed that we develop some kind of high level at least high level developer guidelines and I think we should work on two tracks and one is how do you develop with current state‑of‑the‑art developer frameworks, what does it take there, do we need to take care of IPv6 and if yes, how? The other is for the enterprise, how do we test all applications which have historically grown and are sometimes written in very old languages, how do we test those and ensure they could work with IPv6 or find out we need to migrate to new applications sometime in the future. So I will update when I have results.
JEN LINKOVA: Thank you.
SPEAKER: NCC S, so first point even if you have the developers who are sensible to the network part, can understand there are cases when upper management says, don't work on this, this is a non‑priority, stop wasting time on this. This is one issue. And about what can we do about it, about previous years we used to have an IPv6 lunch and IPv6 turn up, how about we start organising an IPv4 shut down.
JEN LINKOVA: We have been talking about this. I think some people actually did last year on 6th of June, I can't remember any results but I know that some people were going to turn over v4 at their homes for 24 hours and so on, yes probably need to invite those people and talk to them. I will take this as action item.
SPEAKER: At the IETF we did it in 2008.
JEN LINKOVA: I am signature on v6 only network and I switch to dual stack just once, so for me everything works, right, but apparently I am very special use case. So, we need to reason from normal people. My input is not so valuable.
JAN ZORZ: So let me share with you the situation in my country. We have constantly been very good on the content side of the IPv6 and that happens because when the first turned on the IPv6 support, we shouted very loudly to other content providers, look, they enabled IPv6 and put some competitive pressure on them so everybody else enabled IPv6 like national TV and search engine, everybody. And the same is happening with ISPs. They all deployed IPv6, they all tested it, they are running the pilots with friendly user tests so they are completely ready. But they are not enabling IPv6 by default to international customers, just because others are not, and if we can make one deploy IPv6 and enable it by default, everybody else will go, of course, we need to shout very loud about this, that this guy enabled IPv6 by default. But what I see is the lack of competitive pressure between the ISPs, and if we can achieve that and make one in the country, make it work, and shout very loudly about this, I think we can achieve some progress.
JEN LINKOVA: Thank you. Any ISPs would like to comment on this?
SPEAKER: That is what I was mentioning as well ‑‑ IPv6 deployment and why is a big provider like ‑‑ we don't deploy it couldn't deploy it and the communication on the internet is not clearly and not loudly and I hope the access provider would come together and make it work and share their information, how will they deploy it then and maybe more visible in the media. I think that would be a wise thing to do.
JEN LINKOVA: So, if I got would like to see some more success stories and deployment experience and recommendations.
SPEAKER: Yes, for example, how does XS4ALL do it, how do they deploy their network?
JEN LINKOVA: So, as I am responsible for agenda for next meeting, so if anyone in this room would like to share the success, or even unsuccess story, come to me, come to us, we can get you a slot.
MARCO HOGEWONING: Wearing my hat when I used to work for XS4ALL, I did present how it deployed v6 it was four Amsterdam meetings ago.
SPEAKER: Yes, XS4ALL. We have shared this many, many times, so I mean, it's all in the archives, I have done presentations also for several network operating groups and stuff, so it is all there.
JEN LINKOVA: People have very short memories, apparently people don't want to use any search engines and search for presentations, apparently they prefer us to invite presenters and ask ‑‑
SPEAKER: I will happily do it again, it's not a problem, I can show you where we are now, if you have a problem ‑‑ give me a slot.
JEN LINKOVA: Any other comments? I have a question. So, we have this lovely IPv6‑only still experimental environment here. I have been building some IPv6‑only test‑beds to find out what exactly rely on v4. And I realise that sometimes, network people probably would like to play with IPv6, IPv6 only, they are responsible for routers but then they have to look after DNS. DNS 64 if talking about ‑‑ it's completely different, you have to set up some serve, you most likely, don't even probably allow to touch server side, so Jan mentioned he has DNS 6to4 which is publically available and ‑‑ 64. I would encourage to you play with it and if you think you need ‑‑ we as a community more of this stuff here, I would like to hear. I ‑‑ how many people would play with V‑only environment if they don't need to think about a DNS part of it. And as they have to care only about routing and address translation. You can take it on the mailing list, it's just a question for you to think about.
SPEAKER: Andrei SIS net from the other side, I take care of the servers and I am not allowed to touch the routers, so I can have the view from the other side. And I actually started, for instance, by deploying IPv6‑only, let's say management networks in our data centre because usually you have your management network closed behind firewalls and most people use something like VPN, so in that case, you already have IPv6 in your VPN so what is the issue, why would you use some R C 1918 addresses if you can use some part of public IPv6 address and put it behind firewall so you can precisely measure your access policies. I do it with our ‑ cluster and unfortunately I ran into many many issues with software that (many) supports IPv6 but usually only in a dual stack and only in about 90 or ‑‑ 80 or 90 percent of its functional but there is still tiny part of functionality that doesn't support IPv6‑only properly. So even if I plan this management system for ‑‑ from the scratch with IPv6‑only, I came to a point where I had to deploy some RFC 1918 addresses to make this system work. Which was quite sad for me.
JEN LINKOVA: From my experience, it's actually quite attractive idea to start deploying IPv6 only management path because it doesn't affect our users, it's something inside your network you don't care. Apparently it's the hardest part of the network to make v6 only because management software are normally ‑‑ no idea about v6, however, I believe if you start early and you are not making urgent project, right, you do what you could and you slowly fix what you have to fix. So I think it's actually very good idea to to start playing on this path because it's unlikely that you break something badly. It's something which is visible only to you. Quiet. Apparently I am standing between you and workshop. So if there is no other comments, are we might call it a day. Thank you very much, if you have any again comments, suggestions, since you would like to hear the next meeting, it doesn't mean ‑‑ I would like to hear what topics we should talk about, please let us know, to Working Group Chairs on the list. Thank you. And before we finish any other topics people would like to discuss, like any other business? Jan is looking at me like ‑‑
JAN ZORZ: Would you like to send me to the mailing list a URL for the next ‑‑
JEN LINKOVA: It would be great, yes, thank you very much, Jan. So thank you very much and see you in November.
(Applause)
LIVE CAPTIONING BY AOIFE DOWNES RPR
DOYLE COURT REPORTERS LTD, DUBLIN IRELAND.
WWW.DCR.IE
