back to article Sheffield ISP: You don't need a whole IPv4 address to yourself, right?

Facing the shortage of IPv4 addresses, and glacial adoption of IPv6, UK ISP PlusNet is looking for volunteers among its customers to test out sharing the IPv4 addresses on its network. The technique being tested by PlusNet uses a NAT (Network Address Translator) to share a single internet-facing IP address between multiple …

COMMENTS

This topic is closed for new posts.
  1. Lusty

    Any chance we could use public and private rather than real and fake IP addresses? 192.168.x.x is not a fake address, it's a private address, and 90.52.x.x is no more or less real but is a public address.

    1. diodesign (Written by Reg staff) Silver badge

      Sometimes simplification of tech news goes too far. The language was tidied up earlier today.

      C.

  2. Phil W

    Volunteering for the unknown

    Surely anyone who really understands what this actually means, probably won't be the sort of person willing to volunteer due to have some kind of port forwarding in place for remote desktop or some such thing.

    Also what about sites that restrict file downloads per hour/day based on IP. If the person/people I'm sharing with use it all up I'm screwed.

    I'd much rather have an IPV6 address, than a NAT'd connection with some stranger.

    1. Anonymous Coward
      Anonymous Coward

      Re: Volunteering for the unknown

      I'm wondering that. Why aren't they asking for volunteers for an IP6 experiment? I smell cost-cutting or somesuch.

    2. Anonymous Coward
      Anonymous Coward

      Re: Volunteering for the unknown

      What about various companies that track illegal downloads based on ip data?

      Pretty stuffed now if half a tower block in Yorkshire are using the same ip to download pirate material.

      1. janimal

        re:illegal downloading

        "Pretty stuffed now if half a tower block in Yorkshire are using the same ip to download pirate material."

        ...except they won't be able to accept incoming connections for p2p

    3. Andy Fletcher

      Re: Volunteering for the unknown

      That's the trick though isn't it? No self respecting Reg reader would bother - imagine playing COD on a setup like that. The latency is already terrible enough thanks.

      It'll be punted to people as a budget option, and joe public will see the price first and worry about the technical details when they've already bought it and realised it's crap.

      1. Anonymous Coward
        Anonymous Coward

        Re: Volunteering for the unknown

        You seem to be ignoring that not all people who frequent this place are dumb enough to play silly computer games like COD etc.

        Some of us have a life away from the computer.

        Methinks you might be addicted to IT and should seek some treatment without delay.

    4. Charlie Clark Silver badge

      Re: Volunteering for the unknown

      Makes you wonder doesn't it? Here my ISP already takes care of the IPv4 / IPv6 handling and servers are also increasingly ambidextrous:

      $ host h-online.com

      h-online.com has address 193.99.144.80

      h-online.com has IPv6 address 2a02:2e0:3fe:100::8

      h-online.com mail is handled by 10 relay.heise.de

      Meanwhile in steam-powered old England:

      $ host theregister.co.uk

      theregister.co.uk has address 92.52.96.89

      theregister.co.uk mail is handled by 10 aspmx2.googlemail.com.

      theregister.co.uk mail is handled by 10 aspmx4.googlemail.com.

      theregister.co.uk mail is handled by 1 aspmx.l.google.com.

      theregister.co.uk mail is handled by 10 aspmx5.googlemail.com.

      theregister.co.uk mail is handled by 5 alt2.aspmx.l.google.com.

      theregister.co.uk mail is handled by 10 aspmx3.googlemail.com.

      theregister.co.uk mail is handled by 5 alt1.aspmx.l.google.com.

      Given the apparent cluelessness of ISPs and technical publications is it time we thought of jumping ship?

    5. big_D Silver badge

      Sounds like...

      the old shared party telephone lines... *shudder*

      Here in Germany, since October, I think, Deutsche Telekom has been giving customers of their combined DSL and VOIP (i.e. no ISDN or analogue telephone connection accompanying the DSL) packages IPv6 addresses by default.

      Most other ISPs seem to be going over to dualstack IPv4 and IPv6 connections. At least the most popular modem/routers over here (AVM Fritz!Box devices, which account for over 50% of households) have a good IPv6 firewall, with a proper default configuration.

      1. brooxta
        Go

        Re: Sounds like...

        Now, you see THAT approach almost sounds sensible. Hang on, no, it DOES sound sensible. Unlike the game of IPvX chicken being played on this side of the North Sea.

      2. CynicalOptimist

        Re: Sounds like...

        a 'party line' was used as a plot device in the Doris Day romcom Pillow Talk to connect two strangers who antagonise each other at first but ultimately fall in love. Perhaps a 'party IP' remake is due. Better hurry up though, it will be an anachronistic joke soon (at least outside of the UK).

        1. Anonymous Coward
          Anonymous Coward

          Re: Sounds like...

          (at least outside of the UK)

          Downvoted for the repugnant* "of"

          *no, I didn't misspell redundant... but that too.

    6. Steve Knox
      Happy

      Re: Volunteering for the unknown

      I'm not so sure what the problem is here. For years I had a local network NAT'ed (10.x.x.x)* behind a Wireless Router which itself was NAT'ed (192.168.1.x) behind a VoIP router (192.168.1.x) and I was able to run any p2p client apps (games, bittorrent, chat apps) I wanted, without punching holes, and I did run quite a few.

      * A bit overkill, I grant -- I only used about 10 addresses in 10.1.1.x -- but it's not routed anywhere and 10.1.1.x is much easier to type than 192.168.1.x. So why make things difficult?

      1. Athan
        Thumb Down

        Re: Volunteering for the unknown

        Likely the other end of any connection in your uses cases was not behind any (non-forwarding) NAT. You connect out through NAT, the other end simply replies. But it would have meant any P2P clients couldn't initiate connections to you, you'd have to talk to them first.

        So, no, not working 100% without issues.

      2. Anonymous Coward
        Anonymous Coward

        Re: Volunteering for the unknown

        @Steve Knox

        I liked it when we used 10. addresses for private IP addresses on our network. Not for ease of typing - though it helped - but because it was more obviously not a public IP address and if I saw 192.168. then I knew it was something that had reset itself.

      3. MR J

        Re: Volunteering for the unknown

        So you had triple layer nat, running game servers on the back layer, and NO forwarding/DMZ rules were set up on any of the three layers.

        SURE YA DID.....

    7. streaky
      Pint

      Re: Volunteering for the unknown

      Well, native ipv6 and 6to4 NAT is the best deployable solution in this circumstance. Actually it's what ISPs should have done ~10 years ago.

      All my stuff is pure IPv6 so I'm in a position to laugh in the face of anybody who has this issue.

  3. Anonymous Coward
    Anonymous Coward

    I'm with plusnet and this had better stay as voluntary.

    If they take away my static IP address I shall definitely go elsewhere.

    1. Ben Holmes
      Thumb Up

      As a current PlusNet customer I wholeheartedly concur. A

  4. Phil O'Sophical Silver badge
    Unhappy

    No surprise, I predict that there will be more to come

    I can see this becoming more common. It's a great pity that the IPv6 developers chose a new mechanism that was unable to permit a phased change. I assume that was in part due to the "ivory tower" mentality that pervades academia; only a perfect solution to every problem would be acceptable and anything with extended headers or other compatible hackery would have been beyond the pale, architecturally.

    Also, of course, when IPv6 was being developed the internet was a lot smaller, and the idea of switching it all off one night and restarting the next day with new addresses probably wasn't as unthinkable as it is today.

    I seriously don't see how we can have even a semi-painless move to IPv6 worldwide. Is there a plan? (serious question)

    1. Gerhard Mack
      FAIL

      Re: No surprise, I predict that there will be more to come

      Well you know what they say: "anything is easy if you don't know what you are talking about". IPv4 is a fixed length header field so there was no way to just expand the address length. DNS allows both address formats and IPv6 addressable machines can also have an IPv4 address. The ideal plan was for IPv6 to coexist with IPv4 and then phase out IPv6 when IPv6 was ready. The problem was that everyone waited for everyone else to go first so the ISPs didn't bother because there was no software support and the OS providers didn't bother because there were no uses yet.

      To put it simply: we had over a decade to do this the easy way but everyone waited until the last possible moment and now the transition will be painful. Don't blame the IPv6 designers for stupid people who can't see the benefit of spending money on anything that doesn't bring a result before the next quarter.

      1. Phil O'Sophical Silver badge
        Thumb Down

        Re: No surprise, I predict that there will be more to come

        > Well you know what they say: "anything is easy if you don't know what you are talking about"

        Sorry, my last 20+ years working with comms protocols must have got lost, I suppose.

        I didn't suggest expanding the address length, that is of course fixed, but other protocols have worked around this by adding additional extended headers. It makes temporary co-existence possible. Look at some of the original suggestions in RFC 1287, for example.

        IPv6 went through many proposals, TUBA, SIP etc. The final one chosen was designed to fix all the perceived problems of IPv4, and direct compatibility was not seen as a requirement.

        As for

        > Don't blame the IPv6 designers for stupid people who can't see the benefit of spending money on anything that doesn't bring a result before the next quarter.

        Why not, as I said they were academic purists who had little practical regard for commercial interests. Let's face in, when IPng work started, the World-Wide Web hadn't even been described outside of CERN!

        I'll bet if you made IPv6 vanish, and asked Google to come up with a solution to IPv4 address exhaustion, you have something that was workable in a year. Ugly, but workable.

        Downvote away :)

        1. Anonymous Coward
          Anonymous Coward

          Re: No surprise, I predict that there will be more to come

          Hi Phil,

          What exactly is your suggestion for a better proposal than IPv6? It's not very clear to me what you think should be done as an alternative, apart from "not ipv6". Add additional headers to ipv4? How would that solve the problem?

          Is your solution really "hope Google come up with something if we ask them to (so far they haven't)"?

          Cheers

          1. Phil O'Sophical Silver badge

            Re: No surprise, I predict that there will be more to come

            Oh, it's at least 15 years too late for any other solution now. CG NAT is what we'll have to get used to as a transition measure, probably as part of a two-speed internet where IPv6 also exists, but isn't widely used for a long time.

            It will be very interesting to look back at this in, say, 2020. Barring a killer app that makes IPv6 essential, no matter what the cost, my money is on widespread CG NAT, at least for domestic ISPs. It's a horrible thought, but I can't see a viable alternative. I suppose we might see IPv6 appearing on mobile networks more quickly.

            So far nobody has answered my question, though. What is the plan for really achieving migration to IPv6, other than waving our hands in the air and saying "well, somebody should make it happen", while downvoting the doom-mongers :) ??

            1. Anonymous Coward
              Anonymous Coward

              Re: No surprise, I predict that there will be more to come

              Actually Phil, it seems like people have answered your question quite a few times already. IPv4 and IPv6 can co-exist, so they will do, till businesses transition themselves over gradually.

              Do you really think there's "A Plan" that every business in the world is going to sign up to? For IPv6, or anything else? Does that exist in any sphere of business anywhere? I can't think of an example.

              Not sure what your point is really, you have no actual alternative proposal, IPv4 and IPv6 can co-exist, be routed between... what is it you actually want to know?

              1. Jon Press

                Re: IPv4 and IPv6 can co-exist

                They can, like telephones can co-exist with the postal service. However if you can only have a phone or a letterbox - but not both - the co-existence is not necessarily helpful.

                And that's unfortunately the territory we're entering - the transition plan has always assumed that the coexistence phase would have pretty much come to an end before the IPv4 space was exhausted.

                1. Roland6 Silver badge

                  Re: IPv4 and IPv6 can co-exist

                  To drive Jon's point home the issue isn't co-existence it is end-to-end interoperability. Try having a telephone conversation with some who can only be reached by letter.

                  IPv4 and IPv6 can co-exist on the same network as they use different network protocol Id's, this was by design. Similarly the V series of modem protocols co-existed on the same network as voice communications, however there was no way for a voice terminal (ie. a phone) to meaningfully communicate with a data modem (beyond dialling the recipients number).

                  The lack of practical real-world interop and/or migration between IPv4 and IPv6 can be blamed directly and wholly upon the team responsible for drawing up IPv6; as they could have drawn upon the vast body of expertise and experience in the OSI and CCITT communities instead of adopting a superior "not invented here" attitude, which has left us with a problem that we have been grappling with ever since.

              2. Phil O'Sophical Silver badge

                Re: No surprise, I predict that there will be more to come

                > IPv4 and IPv6 can co-exist, be routed between... what is it you actually want to know?

                Of course they can be, I've been doing it since before IPv6 was officially published as a protocol.

                The point, as was made in a post just above, is that having a host with IPv6 is pointless if the network equipment between it and it's destination can't handle IPv6, and no-one will upgrade the network equipment until there are enough IPv6 hosts to make it worthwhile. My question is what incentive there is to fix that, and no-one has been able to answer me. What do you think it will take for BT/Plusnet/Sky/Virgin etc. to replace every single customer router with an IPv6-compatible one? We've been waiting for businesses to transition "gradually" for 15 years, and despite the huge explosion in connected devices we're now at the staggering level of 1% IPv6 penetration. Now it looks like they're transitioning instead to the lower-cost and less painful alternative, CG NAT. Yuk.

                There are ways to add new features to protocols such that existing protocol stacks can still process them, while just ignoring the new features. The IPv6 designers chose not to do so, which may have seemed like a good idea architecturally but is now a severe disincentive to upgrade.

                1. Anonymous Coward
                  Anonymous Coward

                  Re: No surprise, I predict that there will be more to come

                  > There are ways to add new features to protocols such that existing protocol stacks can still process them, while just ignoring the new features. The IPv6 designers chose not to do so, which may have seemed like a good idea architecturally but is now a severe disincentive to upgrade.

                  But there is no way to physically extend the 32bit IPv4 address space. The IPv4 header is a fixed length (20 octets). Even if a way was devised you would still have exactly the same problem. Some routers/equipment would support the extension and some wouldn't. Customers routers would still have to be replaced for those supporting the IPv4 "extension" (firmware flashing would be as much of a solution as it is for IPv6).

                2. Ken Hagan Gold badge

                  Re: No surprise, I predict that there will be more to come

                  "There are ways to add new features to protocols such that existing protocol stacks can still process them, while just ignoring the new features. The IPv6 designers chose not to do so, which may have seemed like a good idea architecturally but is now a severe disincentive to upgrade."

                  Back in academia, I think the assumption was that you would upgrade by loading up revised firmware for your router and run a dual stack. Back in 1995 or whenever, that would have had an expected financial cost of nothing and a labour cost of bugger all.

                  Fast forward to 2010 and the world is full of domestic ADSL boxes whose manufacturers refuse either to issue new firmware or to help anyone else write new firmware, because *they* want the cost to be "£50 for a new box thank you very much". (Rather ironic that the sticking point here is *software* that in nearly every case was Linux, gratefully picked up for free by the manufacturer.) Notice that extensions to IPv4 would have suffered the same fate. (One can argue that the ISPs are also part of the problem if they want to sell IPv6 as a chargeable extra.)

                3. PyLETS
                  Gimp

                  IPV6 transition can happen now

                  It's already happened as far as I'm concerned - home network and hosted server and applications I maintain all upgraded to run dual stack. That's even though my best available ISP (Cable offers better speeds than ADSL) is IPV4 only, so I tunnel IPV6 over IPV4 using protocol 41. I had to do something similar to get Internet in the late eighties by tunneling IPV4 over X25. Well, now that FTTC makes better ISPs available in perhaps the next 18 months, my existing one (Virgin Media) is going to have to offer dual stack pretty soon, or I'll be ditching them for one of the ISPs which are offering IPV6 natively. At least Virgin Media don't block protocol 41, like some crummy ISPs do.

                  I suspect most consumer equipment could be firmware upgraded to run dual stack, and ISPs providing tunnel servers more locally could support consumers they can't firmware upgrade without requiring hardware replacement.

            2. Anonymous Coward
              Anonymous Coward

              Re: No surprise, I predict that there will be more to come

              The internet of things is the killer app.

              You can hardly blame the designers, IPv6 was ready to use in 1996 before the internet was very large at all. ISPs could have saved themselves all the transition hassle if they had adopted it sooner.

            3. Decade
              Childcatcher

              Re: No surprise, I predict that there will be more to come

              Well, my plan for achieving migration is saying, "I will make it happen," on my network. If you are a network administrator, now it is YOUR personal duty to enable IPv6 connectivity on your network. IPv4 was deployed by millions of individual decisions to join the Internet. IPv6 will be deployed by the same.

              In my section of the USA, the ISPs are trying to eliminate the home router market. When you get new Internet service from Comcast or AT&T, you get a combination modem and wireless router device, too. The upside is that the routers that they've starting shipping in the last few months support IPv6. This means the homes in the USA are gradually shifting to IPv6, without the consumers having to learn new technology. This is a positive development.

        2. Gerhard Mack

          Re: No surprise, I predict that there will be more to come

          An extended header would have left us in pretty much exactly the position we are in now only with slower header processing (a big deal for routers), no one would have bothered to implement them until the last moment.

          One of the main reasons they chose 128 bits was because the designers couldn't see a way to extend the addressing and they knew the transition would be painful so they decided to make it large enough that there wouldn't need to be another transition in the foreseeable future. That's hardly "ivory tower"

      2. Fatman

        Re: we had over a decade to do this the easy way

        I remember when I first started diving into IPv6, one item I noticed was a specific prefix that would have allowed existing IPv4 addresses some space within the IPv6 universe; and wondered why it was never readily used.

        Then the old standby reminded me - it's called the bottom line. Fuckers too cheap to spend the $$$$.

    2. Skoorb

      Re: No surprise, I predict that there will be more to come

      RIPE NCC in September:

      "On Friday 14 September, 2012, the RIPE NCC, the Regional Internet Registry (RIR) for Europe, the Middle East and parts of Central Asia, distributed the last blocks of IPv4 address space from the available pool...

      "It is now imperative that all [ISPs] deploy IPv6 on their networks to ensure the continuity of their online operations and the future growth of the Internet."

      Somebody didn't read the memo.

      http://www.ripe.net/internet-coordination/news/announcements/ripe-ncc-begins-to-allocate-ipv4-address-space-from-the-last-8

    3. An0n C0w4rd
      Stop

      Re: No surprise, I predict that there will be more to come

      IPv4 and IPv6 can co-exist. There was never any plan for an overnight switch from v4 to v6. IPv6 was designed to run in a dual stack environment where both v4 and v6 addresses were in use. If you were connecting to a system that also had a v4 and a v6 address, then local configuration will determine if you make the connection over v4 or v6, with the default to be to use v6.

      Of course this isn't perfect. People can have a v6 iP that isn't able to connect to your v6 IP, and people don't want to wait 90 seconds for the v6 connection to time out before retrying on v4. This lead companies to develop a standard called "Happy Eyeballs" which try and learn whether to use v4 or v6. Yahoo! also sponsored an extension to BIND to try and help mitigate the split network scenario.

      The real reasons nothing happened until it was too late was lethargy and inertia. Vendors didn't want to spend the (considerable) effort in making their products IPv6 compliant as it wasn't affecting purchasing decisions in the vast majority of cases. Customers weren't asking for V6 because either the people making the decisions didn't understand or because they didn't want to pay more to get v6. End user devices (DSL modelms, etc) didn't support v6 as ISPs didn't offer it, and ISPs didn't offer it because no devices could use it. A set of classic chicken and egg scenarios.

      Yes, IPv6 was largely driven by academia, and that can be witnessed by the original specs for IPv6 autoconfiguration where the end client figured out what the subnet it was on was and then used its Ethernet MAC address as the last 48 bits of the submit and hey presto, you got a unique routable IPv6 IP. Its why IPv6 subnets tend to be /48s - the 48 bits in the MAC. It wasn't until years later that someone pointed out that MACs are globally unique (or are meant to be) so it didn't matter what network you were on, your computer could be uniquely identified through the IP it chose. A new autoconfiguration mechanism was released in the past 2-3 years to address that.

      Even though Cisco has supported IPv6 natively in IOS for years, initial implementations were not carrier grade. IPv4 was handled through very efficient DCEF, which IPv6 packets were process switched, a very expensive process. Only recently has Cisco moved IPv6 into DCEF (or whatever they call it today).

      As for CGN (Carrier Grade NAT, the industry term for what Plusnet is trying), there are a lot of implications and not all of them are well thought out. The issues raised in the article are relevant, and search engines and other such web sites are already concerned about the rise of CGN as it impacts their operations to not only monetise their search results and make them more relevant by using geolocation, it makes defending the sites against attack a lot more difficult as you can't just block the IP and affect a single user any more. The search engines have been in talks with ISPs for years about IPv4 to IPv6 transitions, and the need for CGN as an interim phase. Search engines would much rather we all move to v6, but thats not going to happen any time soon.

      CGN also has implications for non HTTP traffic such as VOIP, as SIP really REALLY doesn't like NAT. That is one issue that I don't think is easily solved in CGN deployments without sniffing and rewriting the SIP control packets, which would be a non-trivial exercise with high traffic deployments.

      1. Jamie Jones Silver badge

        Re: No surprise, I predict that there will be more to come

        An0n, I agree with your post apart from this bit:

        "search engines and other such web sites are already concerned about the rise of CGN as it impacts their operations to not only monetise their search results and make them more relevant by using geolocation, it makes defending the sites against attack a lot more difficult as you can't just block the IP and affect a single user any more"

        This has been an issue for decades, since DHCP largely took over from static IP's for home users.

        And as a CGN will be allocated to an ISP, geolocation won't be affected. Even if the ISP has customers (and hence peers) in other countries, they will be using a different CGN for that country, else all their traffic will be routed back to one country and out again (and indeed, if there is an ISP that works like this already, then geoloation would already be similarly affected)

        Nope, search engines and advertisers have long gone for cookies etc. not specific IP address

        1. Alan Brown Silver badge

          Re: No surprise, I predict that there will be more to come

          "This has been an issue for decades, since DHCP largely took over from static IP's for home users."

          ISP-level HTTP transparent proxies have been around a long time too and that doesn't seem to have affected them much (hint: That's why cookies are used)

      2. Mr Flibble
        Boffin

        Re: No surprise, I predict that there will be more to come

        Subnetting using /48 isn't to do with MAC addresses. Subnetting using /64, however, is. And I recall reading something saying that the preferred allocation is now /56. (I have a /48, but I've yet to need more than a /60. Well, two /64s, really. I could get away with one but I need a publicly-routeable address on the external network interface and I don't have any sort of IPv6 address translation going on, not that anybody should need that.)

        Why not /80? But then you couldn't have local addresses which don't contain something which could be a MAC address…

      3. Vic

        Re: No surprise, I predict that there will be more to come

        > SIP really REALLY doesn't like NAT.

        SIP is fine with NAT. All my phones are behind at least one NAT router.

        The trick is to set up the STUN service properly - that means the external IP address can be embedded in the SIP packet.

        Oh - and turn off all those "helpful" SIP ALG implementations in routers. They're universally crap...

        Vic.

    4. Anonymous Coward
      Megaphone

      @Phil

      "It's a great pity that the IPv6 developers chose a new mechanism that was unable to permit a phased change."

      And why wouldn't it? As other already mentioned; DNS can use both A and AAAA for the same site; thus leaving it up to the user to, well, use either one.

      If any the only thing I think you can blame the "IPv6 lobby" for is that they've been playing "Cry Wolf" for too long. You know; predict the end of the Internet due to running out on IPv4 addresses after which nothing happened. And not once, not twice but at least four times in a row. That was a very good way to lose a lot of credibility really fast.

      They should never have made it as dramatic as they did, then people may have taken them a lot more serious than now. Good luck presenting IPv6 implementation plans to the upper brass now: "But haven't we heard those doom stories for the last 10 years now? And everything just kept running, so why should we bother with all this when everything works just fine?".

      1. Alan Brown Silver badge

        Re: @Phil

        "But haven't we heard those doom stories for the last 10 years now? And everything just kept running, so why should we bother with all this when everything works just fine?".

        Step one was dynamic IP allocations - that staved off things for a few years

        Step two was NAT - that staved off things for a few more years.

        Step three was allocating all the reserved ranges - that staved off things for a few more years.

        Not bad for a protocol which was a kludgy hack only intended for general use for 5 years or so (and that's why 4 billion addresses was "enough") until IPX was ready.

        It's a pity that IPX turned out to be more of a clusterfuck than IPv4

      2. Vic

        Re: @Phil

        > the only thing I think you can blame the "IPv6 lobby" for

        I'm blaming them for giving us non-memorable IP addresses.

        IPv4 is *nearly enough*. 32 bits is a bit less than 1 address per human - particularly when you take out the reserved portions of the address space.

        So what we need is a little bigger - 34, 35 bits, something like that. That would give us an address each, and still be feasible to hold an address in your head. But we like multiples of 8, so a 40-bit or 48-bit addressing scheme would have been wonderful - 40 bits is still 256 instances of the current IPv4 Internet, and that is a lot.

        128 bits is unwieldy in extremis - it's far, far bigger than we're ever going to need. We have MAUs of /64 - why would anyone need a /64 as the *minimum* allocation? We're never going to have 2^64 humans, and I can't imagine any one of them needing 2^64 addresses. I personally have a /56 and a /48, and I have a sum total of 3 machines on those 10^23 addresses...

        Vic.

        1. Ken Hagan Gold badge

          Re: non-memorable addresses

          IPv4 gave us non-memorable addresses. A memorable one uses words and has variable length segments that not only uniquely identify the addressee but also tell you how to get there. As proof of this, I'd wager that the ordinary end-user can probably quote you quite a few DNS names but no IP addresses.

          Funnily enough, the reason the ivory tower academics jumped to 128 bits for IPv6 was the realisation that a longer binary address could be used the same way. I don't know how successful this is in practice, but the extra bits were supposed to be carved up so as to make it really easy to implement routing algorithms.

          1. Vic

            Re: non-memorable addresses

            > IPv4 gave us non-memorable addresses

            Nonsense. I spend every day surrounded by people who readily recite IPv4 addresses from memory.

            The same simply cannot be said for IPv6.

            > I don't know how successful this is in practice

            It isn't...

            > the extra bits were supposed to be carved up so as to make it really easy to implement routing algorithms

            Having a MAU of /64 pretty much obviates that.

            Vic.

  5. Mr Tumnus

    Network Ranges

    Hi Bill, it's not strictly true to say 192.168 addresses are "fake" or "not valid", they are valid ip addresses, it's just they're reserved as private ip ranges, for anyone who wants to use on their private networks. Therefore they're not routable on the internet.

    They're private network ranges, rather than public network ranges, and are defined in RFC1918. But in every other sense, they're "valid" and "real" ip addresses. So one company in Texas might use the same private IP range behind their internet routers as another company in Liverpool, eg, 192.168.10.0/24.

    The reserved ranges are;

    10.0.0.0 - 10.255.255.255 (10/8 prefix)

    172.16.0.0 - 172.31.255.255 (172.16/12 prefix)

    192.168.0.0 - 192.168.255.255 (192.168/16 prefix)

    You could use routable addresses as your local network range, but it would mean the routable addresses weren't reachable, because your clients would expect that range to be local, not out on the internet (they'd arp for those ips, rather than routing traffic to their local gateway)

    Eg, if you used 173.194.78.0/24 as your local network range, you might have trouble getting to some of Google's services, because that range is used for www.google.com (among other things. Although Google also have www.google.com on other ranges too)

  6. M7S
    Meh

    "The days when everyone ran their own servers are long gone and almost everything is available in the (better secured) cloud these days"

    I'd like to run my own mail server, and intend in future to have things like the ability to check home cctv, respond to callers at the entryphone (fitted with IP camera) etc from my smartphone. Surely with all these future connected households we hear about, where one can start the bath running whilst still travelling home valid connections will be a requirement?

    1. Arrrggghh-otron

      "The days when everyone ran their own servers are long gone and almost everything is available in the (better secured) cloud these days"

      I still do, on a plusnet static address. I like knowing who is looking after my email... particularly after the email storage cock up at plusnet a good few years back.

      Maybe this will spur plusnet to get on with their public IPv6 deployment - it has been trailed on and off for the past few years, so this seems as good a reason as any to roll it out.

    2. AndrueC Silver badge
      Meh

      > "The days when everyone ran their own servers are long gone and almost everything is available in the (better secured) cloud these days"

      I still do and after ten years I prefer to think of 'a machine I own locked in my spare bedroom' as more secure than something hosted by a large corporation that sees me (and possibly my personal data) as a cash cow or an ISP that makes no money off the service. I'm not sure about the uptime. I've lost mine for a day four times in the last 12 years (always during the week and always after I've left for work). How would a cloud service compare with that?

      Anyway I think this service has merit for the average user. More support for IPv6 would be better though. My ISP supports it - IDNet - but sadly my nine month old NetGear WNR1000v3 doesn't. Or at least not the UK firmware. Apparently there's a firmware version somewhere that does give or a take a few bugs.

      1. John Sager

        I guess Plusnet would at least have considered using NAT64, except that most wireless routers don't support it, and probably don't even have a firmware upgrade to do so. NAT64 still has the NAT-specific drawbacks that v4 NAT has, but at least it would make the home network IPv6 by default. Perhaps as home wireless routers become capable of IPv6 they could start to run NAT64 in parallel.

    3. Jason Bloomberg Silver badge

      @M7S - Connected households

      Surely with all these future connected households we hear about, where one can start the bath running whilst still travelling home valid connections will be a requirement?

      You can have the home-side systems push status to the cloud, ask if it's meant to do something and do it, so your direct interaction is with the cloud not your home, your home is always polling and making outward connections rather than accepting incoming connections.

      There are disadvantages to polling but it's likely to work for most homes for most things which don't need instantaneous interaction. Response times can be reduced at the cost of more frequent polling and greater bandwidth use and it can be dynamically adjusted.

      Most residential customers probably don't get a static IPv4 anyway so that model id going to be adopted to allow an internet of things to work before everyone is on IPv6. That seems to be what Electric Imp is doing.

    4. Ken Hagan Gold badge

      that "better secured cloud" in more detail...

      Perhaps the killer app for IPv6 is an appliance (quite possibly built into the domestic router) that lets you run your own cloud. You do the backups. You determine the privacy policy. No-one turns round a couple of years later and says "all your stuff belongs to us next month unless you find (and migrate to) a new provider".

      It needn't be expensive. You need a disc large to store your crap on, something large enough to backup to, and a router capable of taking the "load" of your immediate family and friends all rushing for the baby video at once. An off-the shelf router with a couple of SD cards would probably be sufficient for most households. The only problem is installing, configuring and maintaining the software stack.

    5. Roland6 Silver badge

      I'd like to run my own mail server,

      >You are in the minority of domestic broadband users

      > and intend in future to have things like the ability to check home cctv, respond to callers at the entryphone (fitted with IP camera) etc from my smartphone. Surely with all these future connected households we hear about, where one can start the bath running whilst still travelling home valid connections will be a requirement?

      Simple you just have an agent out in the cloud to which both your home and remote systems connect; this is the way LogMeIn and other similar services work. This also means that your home can use a dynamic IP address, who's actual value at any particular time isn't something you need to worry about.

  7. Thomas Whipp

    Privacy issue

    From a "Police" perspective this would probably require PlusNet to retain a lot more logging information. At present for general network access all they would need to do is log which IP is assigned, under a NAT arrangement they would need to log individual mappings as just knowing someones private IP wouldnt be any use when asked who connected to dodgy site X.

    I know there has been a significant uplift in logging over the last few years - but this does feel like it would be another step up.

    1. Anonymous Coward
      Anonymous Coward

      Re: Privacy issue

      total nonsense

    2. Anonymous Coward
      Anonymous Coward

      Re: Privacy issue

      The destination IP logging etc would be done at the point where an end user's traffic still had a unique IP address - probably one of the "local" ones. That could either be a dynamic lease or a "permanent" reserved one. The final PlusNet NAT would then multiplex several users onto an external IP address - wasn't that called PAT? There's no limit to how many times a connection gets NAT/PAT manipulations. All that matters is there is no ambiguity of "local" IP address routing in the various stages.

      In fact I presumed that's what ISPs already did for "economy" users - it is certainly what large organisation's intranets do.

    3. Anonymous Coward
      Anonymous Coward

      Re: Privacy issue

      Logging requirements on the ISP side aren't as bad as you think, the ISP could always NAT a certain user to a certain IP and port range, or at least allocate a block of ports on-demand and log the block allocation rather than individual connection mappings. This also reduces accuracy requirements on timestamps which is also pretty much necessary otherwise it is going to be difficult to correlate logs between different organisations; not everyone has accurate timekeeping.

      Logging requirements on the server side are likely to be more troublesome. Logs for *many* service daemons (think: typical web server, etc) only record source IP addresses, not source port numbers, so correctly reporting abuse to ISPs is going to be difficult. In some cases it is going to be easier for server operators to run IPv6 than change logging to also record port numbers...

  8. Crisp

    When the government can't track down an individual twitter user

    You can bet that we'll be moving to IPv6 pretty sharpish.

    1. Phil O'Sophical Silver badge

      Re: When the government can't track down an individual twitter user

      > You can bet that we'll be moving to IPv6 pretty sharpish.

      The fun thing there is that with IPv6 you can probably get a new address before every tweet. Good luck tracking that!

      1. Anonymous Coward
        Anonymous Coward

        Re: When the government can't track down an individual twitter user

        "The fun thing there is that with IPv6 you can probably get a new address before every tweet."

        The intention with IPv6 is that every device gets its own unique, permanent IPv6 address. IIRC it is usually a compound of the ethernet MAC address and a user's prefix. Mobile traffic is probably also carried in IPv6 tunnels that allow location changes without the device's basic IPv6 address changing.

        1. Phil O'Sophical Silver badge

          Re: When the government can't track down an individual twitter user

          Well, of course that's the intention, but that was also the intention with IPv4 and it didn't take long for anonymizing services to popup. IPv6 will make that even easier, you'll be looking for a sand grain on a beach instead of a needle in a haystack!

        2. Vic

          Re: When the government can't track down an individual twitter user

          > IIRC it is usually a compound of the ethernet MAC address and a user's prefix

          No, that's just the link-local stuff, and is just a recommendation, not a requirement.

          Externally-visible IPv6 addresses can be whatever you like within your allocation - and for servers, using the MAC address would be an extraordinarily bad thing to do[1].

          Vic.

          [1] It makes replacing a failed network interface rather tricky, for example...

  9. El Presidente
    WTF?

    The days when everyone ran their own servers are long gone

    Oh? Really?

    I have a static IP and, soon, will have 10 or 12Mb/s upstream.

    I'm planning to save myself a lot of money I now pay to hosting companies by running my own .. servers.

    1. Anonymous Coward
      Anonymous Coward

      Re: The days when everyone ran their own servers are long gone

      Look at your electricity bill and how long servers last in a home environment - you won't be saving money. Hosting in a datacenter has numerous advantages AND works out cheaper.

      1. El Presidente
        Facepalm

        Re: The days when everyone ran their own servers are long gone

        You've absolutely no idea how much I spend per annum on various web and mail hosting services so how you can tell me I won't save any money will have to remain a mystery to man and science, I suppose.

        Believe me, I've done the figures, I'm already paying for the fat pipe, I'll save enough money to have paid for the hardware by month 10 and be well in profit by month 12. Plus, my skill set will have improved.

        It's not all abut the money, is it?

        1. Grogan Silver badge

          Re: The days when everyone ran their own servers are long gone

          Except that your links to the internet will be poor compared to a datacenter, that has routers from the major carriers right on their doorstep.

          You, on the other hand, with a residential or business connection intended for client use, will be behind much more infrastructure and have considerably more latency for clients connecting to you. Some clients may not be able to reach you at all.

          That's the reason why people have servers in professional datacenters. Running it from home, it will be hit or miss for clients, and you'll easily saturate that pipe you think is so fat. ISPs don't give you unmitigated 24/7 full duplex connections at your stated speeds. If you're just goofing around that's fine, but for any serious hosting with any traffic it's not viable.

          You will see, as many others have.

          1. NullReference Exception
            Thumb Up

            Re: The days when everyone ran their own servers are long gone

            Who says it has to be for paying clients?

            Disks are cheap. RAM is cheap. Broadband is something you're paying for anyway, and even here in the United States of Verizon you can get halfway decent uplink speeds if you live in the right place. A DynDNS account costs five bucks a year. Buy a low-spec Dell PowerEdge or build the equivalent from parts from newegg, stuff it full of the aforementioned cheap disks and RAM, install your favorite VM solution and go to town. Run your own cloud backup for your family and friends. Run your own Exchange server and sync your phones without having to sell your soul to Google. Run BES, if you're a masochist. Run FreePBX or Elastix to get unified communications, also without having to sell your soul to Google. Do other stuff that you could never afford to pay for if you had to do it through a third party service provider. Then turn around and use everything you just learned in your day job.

            Take away the routable IP address and the Internet becomes a lot less fun, and a lot more like cable TV with five trillion channels and nothing on.

            1. Grogan Silver badge

              Re: The days when everyone ran their own servers are long gone

              If this is in reply to me, I didn't mean paying clients. I meant it in context of "client and server". A client would simply be a visitor's web browser etc. Of course you can have a few clients connect for whatever reason. (e.g. filesharing)

              Try running a site like TheReg or even a forum with, say, even just a few hundred regular users on a home internet connection. You'll lose most of your membership because they will be frustrated.

              Also, if running your own mail services on an ISP connection, few will accept mail from you. You will have problems, regardless of having your MX records and everything in order.

              It's not that you can't do a lot of stuff yourself (like you describe) or that you can't host anything, I was just reacting to the notion that you don't need a datacenter when you have high speed internet at home. It's kind of deceiving if you don't know any better, one would think "12 mbit upstream" would be dandy but in reality other factors make it undesirable.

              It's certainly worth the $150 or so a month it costs me to have a server in a datacenter. I'd spend thousands a month to lease lines, and pay an upstream ISP to have viable hosting from home. (That's the point I'm stuck on)

              I guess I am way out of context because I am absolutely not saying I would want or even tolerate NAT from an ISP. I sometimes open ports for services when I have a specific reason to (e.g. I could pop open an ftp server right now if I wanted someone to get something from me directly). If faced with NAT, I would cancel the service so fast, and so nastily, that they would need counselling.

              NAT is for ME to do, with my one public facing IP address.

      2. DJ Particle

        Re: The days when everyone ran their own servers are long gone

        Even if the server is a low-power Mac Mini that only moves files?

      3. Vic

        Re: The days when everyone ran their own servers are long gone

        > how long servers last in a home environment

        Mine have lasted many years so far..

        > you won't be saving money.

        That depends on what you're running.

        > Hosting in a datacenter has numerous advantages

        And several disadvantages. The "best" way to host your stuff depends on what stuff you want to host...

        Vic.

  10. /dev/null
    WTF?

    "odd versions are experimental"

    Huh? There was no IPv1,2 or 3 - IPv4 was numbered to match the corresponding TCP protocol version (RFC793). TCP actually predated IP, (see RFC675) and hence was in its fourth version at the time.

    And IPv5 didn't really exist either - version 5 was used to distinguish IEN-119 ST stream protocol packets from IP packets. ST was not intended as a replacement for IPv4.

    I think someone might be getting mixed up with the old Linux kernel version numbering scheme...?

    1. koolholio
      Thumb Up

      Re: "odd versions are experimental"

      ipv4

      xxx.xxx.xxx.xxx

      ipv6

      xxxx:xxx:*null*:xxxx:xx:xxxx

      UDP im sure was before the TCP handshake too! Along with IPX

    2. bigdish
      Alert

      Re: "odd versions are experimental"

      According to IANA, who dish out IP version numbers...

      IPv5 was used for the ST Datagram mode protocol (which apparently went on to feed into the development of MPLS)

      IPv6 was originally allocated to an experimental version known as SIPP (Simple Internet Protocol Plus)

      IPv7 was assigned to the experimental TP/IX: The Next Internet, although the same version number was also used for a proposed version called CATNIP (Common Architecture for the Internet)

      IPv8 was assigned to the experimental PIP (The P Internet Protocol)

      IPv9 was assigned to the experimental TUBA (TCP/UDP with Bigger Addresses)

      So, all but two version numbers have so far been experimenatl (or just skipped) and the next version of IP will have to be IPv10...probably...

  11. Phil W

    They're doing this backward...

    I think the sensible thing with this really would be to make it an opt out scheme, maybe even charge a small fee for static IPs.

    If necessary I'd pay say an extra £1-2 per month for a static IP, where as Joe Public who doesn't really know what a static IP is could get NAT'd and not notice or care.

    It also seems likely they'd free up a lot more IP addresses this way.

    1. Anonymous Coward
      Anonymous Coward

      Re: They're doing this backward...

      doesn't matter at some stage there won;t be enough.

      1. Phil W

        Re: They're doing this backward...

        Indeed, it is a delaying action at best. But that's hardly a compelling argument against doing it.

        To argue otherwise is to also argue against fuel efficient cars because the oil will run out eventually anyway.

    2. Shonko Kid
      Holmes

      Re: They're doing this backward...

      You seem to have missed the bit where Plusnet (or anyone else) doesn't know what effect it will have on various applications that Joe Q Public might be using. Sure, they could opt everyone in, but then have 90% of their customers complaining that X no longer works for them, and switching ISP. Not a great move.

      1. Phil W

        Re: They're doing this backward...

        Well, obviously don't opt EVERYONE in at once.

        Do it the virgin media way (one thing they do quite well) and test new network config/firmware updates on a small test group and increase the test group size until you're confident it works ok. Then opt everyone in.

    3. ZimboKraut
      Facepalm

      Re: They're doing this backward...

      Weeeeelllll.....

      There is an issue with NAting and joe public...

      As some people may know., NATing can cause problems particular when playing online games.

      It just depends how well the online games handle packet loss.

      Particularly UDP does not overly like NATing.

      As most games use UDP for the transport this can cause problems for gamers who may not know or understand too much about TCP/IP.

      Also when you go into double NATing, (which you would have automatically when following the PlusNet proposal, you would have the PlusNet NAT as well as the NAT on the DSL-router), you can encounter even more problems, as there are plenty of applications, that totally dislike double and more NATing, which in turn would cause problems for the enduser and then again more support requirements for the clientservices.

      etc, etc....

      IPv6 - yes

      IPv4 + NAT = Chaos....

  12. Anonymous Coward
    Anonymous Coward

    Cloudy with a chance of unreachability

    "The days when everyone ran their own servers are long gone "

    I disagree with that. More people probably run home servers than ever before given the availability of plug and play NAS boxes. Some of those may even sync to cloudy services. Carrier-grade NAT may break a lot of this, which is presumably why PlusNet wants to test CGN. I wish them the best of luck. Section 5.2 of RFC6598 has some of the things to expect.

  13. Anonymous Coward
    Anonymous Coward

    for V4 this is going to afftect _every_ ISP and at some stage in the future everyone will need to use CGNAT of some description.

  14. banjomike
    WTF?

    (better secured) cloud ??

    If I had a pound coin for every time the Reg has been forced to run a story about a chunk of the cloud either going "titsup" or otherwise crapping out, I would have ... several coins. Better secured cloud ?? Not sure about that.

  15. Anonymous Coward
    Anonymous Coward

    The Department for Work and Pensions

    are currently hoarding a /8 so take that back and we've got another half a million IPs to play with.

    1. Nigel Titley

      Re: The Department for Work and Pensions

      That gives us a couple of months at the current run rate.

      IPv6 will be here at some point, how quickly depends on how many people are willing to put up with workarounds like CGN and for how long. The folks who are currently riding on the IPv4 transfer market reckon they have about 3 - 4 years to make their killing and then IPv6 will out number IPv4 and the internet will flip to IPv6. I think they are probably right.

      Personally, these days I don't buy service from someone who can't offer me IPv6. I reckon if they are too stupid or too mean to manage it then I don't really want to be on their network.

      1. Anonymous Coward
        Anonymous Coward

        Re: The Department for Work and Pensions

        Back in 2006, the IPv6 folks were predicting that 50% of internet traffic would be IPv6 by 2010. Here we are in 2013 and IPv6 hasn't even got to 1% yet.

        People will *have* to put up with CG NAT, like it or not. Aren't there already some S. American countries where the only service offered by any ISP is CGN?

    2. This post has been deleted by its author

    3. Skoorb

      Re: The Department for Work and Pensions

      Sorry, nope, that's the entire Government Secure Intranet, currently 80% of addresses in that range are in use, the remaining space is earmarked for the new Public Services Network:

      http://www.whatdotheyknow.com/request/internet_protocol_ipv4_address_a

      http://www.whatdotheyknow.com/request/internet_protocol_ipv4_address_a_2

      1. John Robson Silver badge

        Re: The Department for Work and Pensions

        So it probably doesn't need to be in public IP space...

      2. Anonymous Coward
        Anonymous Coward

        Re: The Department for Work and Pensions

        In use but completely wasted in a private network that could be behind NAT like everyone else. Also if government was facing the same problem of exhaustion as everyone else they might see the need to do something about it!

      3. Anonymous Coward
        Anonymous Coward

        Re: The Department for Work and Pensions

        UK gov have 12 million publically accessible hosts? I doubt it

    4. Anonymous Coward
      Anonymous Coward

      Re: The Department for Work and Pensions

      In reply to AC@13:30

      If they're hoarding a /8 they're sitting on >16 million addresses, not half a million.

      Though, reading the links supplied by Skoorb@13:49 there are two relevant /8 blocks. The MoD's usage of their /8 was above 60% in Dec 2011 and the DWP's usage of their /8 was around 80% at approximately the same date. Summing the addresses in both blocks that's still >9 million spare. But even if it was possible to extricate the addresses from the (inevitably inefficient) network assignments it would still only delay the IPv4 address exhaustion by a few months maximum.

      The answer is to shift to IPv6 post haste, not faff around scrounging IPv4 scraps.

  16. bcollie

    Phased Change

    What ISPs need todo is start shipping new customers routers that are configured to use IPv6 to the ISP, but are running IPv4 NAT for the customers devices.

    For 99% of customers this would be fine.

    For the 1% needed static addresses, leave them as IPv4 all the way through.

  17. Anonymous Coward
    Anonymous Coward

    Confession of ignorance...

    By default plusnet offers dynamic IPs, but you can opt to have a static IP for a £5 fee unless you're on one of their budget packages.

    So, I'm assuming that the dynamic IP is still a 'genuine' one and not natted, just allocated from a pool. The change they're planning to make would be to share these addresses between multiple users with NAT. I must admit I'd already assumed that the ISPs were using nat for their dynamic IPs, hence my confusion..

    Have I got that right?

    Anon because I can't bear the shame...

    1. Jamie Jones Silver badge

      Re: Confession of ignorance...

      Yes, the dynamic IP address is genuine, and routable. They are not using NAT for these. These IP addresses are shared in the sense that when you disconnect, the IP address you had goes back into the pool, and so is then available for use by another customer. However, whilst you are connected, that address is assigned to you, and only you.

      NAT would involve one of the IP addresses being used by more than one customer at once, in that the customers will get a network-unique 'private' address, which NAT's out to the public internet via one IP address - in other words, different customers will be seen to use the same "public" IP address at the same time,

      If you currently have more than one computer at home connected to the internet at the same time, then this will be down to your NAT router. Your dynamically assigned (but unique) IP address is visible to the outside world, whilst your home systems get a seperate home-unique 'private' address.

      TL; DR : Basically, the sort of NAT setup that most people have at home to get more than one computer online at the same time will be extended to the ISP level, where the ISP will have groups of public IP's NATted to a group of simultaneous customers.

      You'll then be running NAT through NAT to reach the internet, but as long as the 'private' ip address range used by the isp don't clash with the private IP range used at home, then this will still work.

      ( I'd guess that plusnet would start using some of the 172.16.x.x - 172.31.x.x private address range, or something obscure in 10.x.x.x because most home setups tend to default to using 192.168.0.x)

      1. paulcupis

        Re: Confession of ignorance...

        Or they could use 100.64.0.0/10, which is set aside for this sort of use. See also RFC6598.

  18. nigel 15
    Megaphone

    you'll get my Ip out of my cold dead hands

    banned from twitter for repeatedly calling piers morgan i giant bell end?

    better than sharing with a couple of users would be more on the mobile model. once you share an IP it doesn't matter with how many people. it may as well be loads, the whole network.

  19. brooxta
    FAIL

    I've got a bad feeling about this...

    Plusnet, BT, Sky, TalkTalk and all the other IPv4-only ISPs need to wake up and smell the coffee. Their management need to understand that there's a market opportunity for their smaller IPv6-also rivals here. All it takes is a year or so more of these kinds of bodges on a creaking IPv4 infrastructure and then along comes a killer app that needs direct contact to a home network resource (I'm thinking something along the lines of ifttt.com or that fork from CES (a killer fork app ... Ouch)) and bosh there goes a sizeable chunk of your customer base. The more the established players prevaricate and procrastinate the more catastrophically vulnerable their market position becomes.

    Come on UK ISPs. Grow up and deal with a C21 internet.

    1. Arrrggghh-otron

      Re: I've got a bad feeling about this...

      Plusnet have IPv6 capability, they have had it for a long time, but it just seems to be perpetually stuck in trails.

      They announced last year that they are stopping the trails due to core infrastructure upgrade. Still no word on when new trails will start up again.

      http://community.plus.net/forum/index.php?PHPSESSID=54e736f8618511fd8c3a5f976d9b0311&topic=106125.0

      1. Jellied Eel Silver badge
        Pint

        Re: I've got a bad feeling about this...

        "They announced last year that they are stopping the trails due to core infrastructure upgrade. Still no word on when new trails will start up again."

        The IETF doesn't understand networks, which is why calls from netheads to de-fund the ITU get amusing. The ITU develop global standards, which is why the Internet works. Buy an STM-64 or an OTU-2 pretty much anywhere in the world and you'll get the same thing, and how it works is pretty tightly defined. Netheads however like Ethernet, because Ethernet is perceived as cheap. Which it can be, if you know what you're buying.

        If you're buying it for xDSL backhaul, knowing about MTU size is important if your IP traffic is turned into PPOE and carried over an L2TP tunnel over an Ethernet. Which is a bunch of extra header bytes that may not fit in the 'standard' MTU of an Ethernet link. So you get fragmentation, which can be a bad thing. You may get it earlier if your Ethernet is EoMPLS because you need bytes for the MPLS lables as well. That can be.. challenging with v4 networks, especially if they're expecting to be able to send a 1500 byte frame and set the DF bits. Add IPv6 into the mix and your overhead bloats given the address inflation.

        But this is OK, because you are not allowed to fragment IPv6 packets at the router. If they're fragmented, they just get dropped and you may or may not become aware of this depending on how well (or badly) PMTUD has been implemented. The IETF workaround is to specify a minimum MTU of 1280 bytes, which may just work on Ethernet links that don't support jumbos, baby jumbos or pink elephants. 1280 is of course less than you get with IPv4, so goodput on large packet transfers will have to drop to ensure delivery. It's less efficient (by a lot with small packets), but that's progress for you.

  20. jake Silver badge

    Marketing? I've heard of that ...

    "The days when everyone ran their own servers are long gone"

    If by "most people", you mean "most people who use TCP/IP in some way, even if they don't even know what TCP/IP actually is", this is very true. Most people have no idea how TehIntraWebTubes work. Note the "most".

    "and almost everything is available in the (better secured) cloud these days"

    Uh ... no. Total fucking bullshit. Unless you're part of the "most", of course. So it must be true.

  21. koolholio
    FAIL

    It would prevent services that use UPNP or prefers/demands strictly open NAT based connections

    .e.g. xbox, PS, mobile wifi, android

    oh and would decrease security for VPN links!

    *applauds some idiots idea* I take my hat off to you, but you didnt quite think this one through properly!?

  22. troldman
    Thumb Up

    Good on them

    I think it's commendable that they're dipping their toe in to this area and feeling around for what the best way forward is. Just because other ISPs aren't doing this, doesn't mean they're not considering more draconian options when the time comes. They're doing a trial and appear to be being transparent about it...

    Re: Phased Change...isn't that really the same as what Plus is proposing? Where the NAT is done is a technical choice, but NAT is NAT is NAT.

    1. bcollie

      Re: Good on them

      No they are proposing NATing people using IPv4 inside their network. Customers would still be ipv4 to the internet. I was proposing they give almost everyone Ipv6 on their network and the internet at large.

      1. troldman
        Meh

        Re: Good on them

        So if they give everyone an IPv6 address only, how would these users access hosts that only have an IPv4 address without some kind of NAT, somewhere?

  23. Irongut

    Another reason to avoid PlusNet. As if I needed more.

    1. Lee Dowling Silver badge

      PlusNet used to be amazing. I had them for decades and they were fabulous and the ultimate test "knowledge of the first guy to answer the phone" was passed flawlessly (changed my ADSL interleaving settings to alleviate latency in interactive connections within, about, 1 minute).

      Hell, they even took over the company hosting my domain names, and I'd again looked long and hard for a good company there and ended up with a fabulous one that I was happy for PlusNet to take over because they were similarly fabulous.

      Then they got taken over by BT. Since then it's been downhill I think. My brother has been fighting for three months with the domain-name host that is now owned by them because all of a sudden tons of things just stopped working properly, after literally 15 years of perfect operation. The ADSL side drops in rating every time I read an ADSLGuide review. And the technical side is now abysmal if people I've recommended to them are telling me the truth (and I have no reason to doubt them).

      Now they've "run out" of IPv4 addresses (telling me that BT don't have enough to go around? Honestly?), but can't be bothered to run a proper IPv6 trial. How about "If you let us issue you with only IPv6 addresses, we'll give you 50% off?" - an INCENTIVE to the technically literate on both fronts, and a way to free up IPv4 addresses for the technically-illiterate who have no idea what that service is or what it means to sign up for it. And last time I recommended someone, they were told you couldn't sign up over the phone, and given that the person in question had no Internet, they just used someone else.

      No, basically, BT have killed PlusNet. Hell, I had more IPv6 connectivity through PlusNet several years ago than they even offer today. It's ridiculous.

      I wouldn't sign up for it. I'd actually take it as a sign to move on to another provider.

      On an pseudo-related note, my external server host (not PlusNet related) is still offering 5 IPv4 IP's (no reason or signing things required) with every virtual server they sell, from £9.99 a month. Can't be that much of a shortage of them. Hell, if it came to it, I'd rather pay the £9.99 extra and VPN all my stuff through a real external IP.

      But really, the fix here is to offer IPv6 instead. But no, they don't even publish AAAA records for their main domain so that people can even GET to their website using it, let alone use it as part of one of their products.

  24. Anonymous Coward
    Anonymous Coward

    Still got my static IP, and a server running at home. As well as having it host my dev kit, I run automatic encrypted delta backups of every other device I have directly to it over a key-locked SSH connection.

    What was this "more secure" cloud bullshit, again?

  25. koolholio
    Stop

    6-to-4

    Simple solution is there, its been made...

    This solves the issue with NAT'ing which is Directly unneccessary and hardly secured. DNAT versus SNAT

    Now would someone please masquerade (oops, serenade) the 6-to-4 ... http://en.wikipedia.org/wiki/6_to_4

    then they could assign us all ipv6 addresses without needing to affect anybody?

    1. Anonymous Coward
      Anonymous Coward

      Re: 6-to-4

      6-to-4 is a way of linking two IPv6 networks over an IPv4 network, so that v6 islands can talk to each other. It doesn't let v4 and v6 hosts interwork.

  26. xanthora

    NAT is now DHCP?

    The NAT allocates fake IP addresses to devices on the network, often starting 192.168.x.x as those numbers are reserved for this purpose, and then maps outgoing connections so that returned data (websites and such) is sent to the requesting computer.

    Actually, DHCP will assign the INTERNAL (not fake) address, or you could set it manually.

    In its simplest form, NAT basically receives traffic from the public IP, and then distributes that traffic to the intended internal address based on which protocol and TCP port address the data packet has in its header.

  27. Peter Gathercole Silver badge

    Does the author know anything about TCP/IP?

    "NAT makes it impossible for anyone on the internet to establish a connection to a computer behind it"

    Not true. You just have to include port information in the address, and set up an inbound port redirect on the device doing the NATing. So outside, you advertise, say, port 2080 for your web server, and have the NAT device redirect inbound packets received on the 'RED' side port 2080 to port 80 on the private address of the device on your 'GREEN' or 'ORANGE' network. All of the devices that I have used that provide NAT have this functionality, so I'm sure that an ISP could deploy it.

    In case anybody does not understand, a valid URL can include a port number, so you can have a URL like www.mywebsite.co.uk:2080/home.html

    It works, but there are caveats, particularly on URLs that refer to other pages on the same site. But it works very well indeed for single port services such as SMTP as long as it is known to use a non-standard port.

    IIRC, DNS has support for providing port information as well as IP addresses for name lookups, it's just not used.

    1. Evan Essence
      Thumb Down

      Re: Does the author know anything about TCP/IP?

      How's that going to work with Plusnet's proposed double NATing?

      1. Peter Gathercole Silver badge

        Re: Does the author know anything about TCP/IP?

        If Plusnet give a fixed IP and port number(s), then it is still possible to do port forwarding even in a double NAT environment. You just have port forwarding on both NAT devices.

        I would be quite happy to be given a range of ports (say 16) for input services on a fixed IP address, as long as I knew what the external port range was, and what ports each would map to when presented to the local NAT device. This would be preferable to me than having all the ports available on an indeterminate IP address, and having to use a dynamic DNS solution to find my servers on the Internet.

        A more complex setup, but I'm fairly certain that the people who want it are the ones most likely to understand how to set their side up.

        Alternatively, you could run your ADSL/cable router in bridge mode, and have them map directly to your servers (only having ISP run single NAT in this case), but that is not a configuration I would want as the ISP would then have sight of your private network unless you put another firewall in.

        1. Evan Essence

          Re: Does the author know anything about TCP/IP?

          You just have port forwarding on both NAT devices.

          The chances of Plusnet co-operating with customers in this way are non-existent.

  28. TheBully
    Meh

    I dont really understand

    IPv6 the addresses are too long and have hexadecimal I can work with IPv4 addressses and have certain ones commited to memory dns servers etc. I remember studying how to work out subnet masks etc for IPv4 and that seemed complicated for my little mind how are you supposed to do all that with those addresses. Too complicated might as well just jack it all in and become a dustman.

    1. Charlie Clark Silver badge
      Coat

      Re: I dont really understand

      Is yours the donkey jacket? Close the door on the way out, ta.

    2. Peter Gathercole Silver badge

      Re: I dont really understand

      IPv4 or IPv6 addressing is largely irrelevant to most internet users. DNS and stateless address autoconfiguration or DHCPv6 takes the pain out of knowing IP addresses.

      Let me ask you. Do you know, off the top of your head, any IP addresses of servers on the Internet?

      And do you care what the address that systems have on your private network?

      For most home users, the answer to both of these is no, in which case, apart from the pain of switching your router and systems over to only use IPv6, the change will be almost entirely unnoticed.

      Of course, some of us (and I am in this category), do care, and I am dreading the switch, because I want fixed addresses in my network for certain systems (no uPNP for me, no sir). I have to do some learning to find out what I need to do to, and I'm not looking forward to that.

      1. Sandtitz Silver badge

        What to do with the waste?

        I know several IP addresses off the top of my head:

        208.67.222.222

        208.67.220.220

        8.8.8.8

        8.8.4.4

        several DNS servers belonging to different ISP's.

        I'm all for change to IPv6 but I haven't bothered with the IPv6 equivalents of OpenDNS and Google DNS...

      2. Jamie Jones Silver badge
        Thumb Up

        Re: I dont really understand

        "Of course, some of us (and I am in this category), do care, and I am dreading the switch, because I want fixed addresses in my network for certain systems (no uPNP for me, no sir). I have to do some learning to find out what I need to do to, and I'm not looking forward to that."

        Peter, you'll do fine.

        There are special addresses dedicated to local networks, and special addresses dedicated to same-network configurations, and other weird and wonderful allocations (ip4 in ip6 etc.) but if you ignore all that, the configuration of ip6 is pretty much the same as ip4. You assign a subnet. A router has an ip on that subnet, as do all other hosts, which use the ip6 address of the router as their default gateway.

        The ip6 address format can look a bit intimidating, but it's just a way of representing 128 bits, in the same way the ip4 a.b.c.d format is just a way of representing 32 bits.

      3. TheBully
        Meh

        Re: I dont really understand

        >>Let me ask you. Do you know, off the top of your head, any IP addresses of servers on the Internet?<<

        Pretty much just public IP addresses of my networks and dns server addresses for configuring forwarders and troubleshooting connection issues etc. I have not got my head around the ipv6 what address ranges to use for internal hosts etc try to avoid it all together if I can all I know is I once installed SBS 2007 and thought it would be a good idea to disable it which made the machine run like treacle until I put the tick back in. I have not disabled it on anything else since except for dialup vpn settings. Its all a bit here be dragons.

  29. Anonymous Coward
    Anonymous Coward

    National only IP ranges

    How about putting whole ranges aside to be 'national only' - rather than international? The example above about 51.x.x.x being the GSi is a good instance of a network that nobody outside of the uk would need to access, so why not let every country have their own 51 networks that are not internationally routable.

    This could be extended such that certain services are on 'national only' ranges - and will save having to implement geoblocking tricks. For security you could choose to be a user on an 'national only' network, etc

    This could help.

    1. Evan Essence
      Thumb Down

      Re: National only IP ranges

      Great way to break the Internet. China and Iran will be very interested in your proposal.

  30. Anonymous Coward
    Anonymous Coward

    fc00::2

    Just give us IP6, the addresses are totally shorter.

    1. Jamie Jones Silver badge
      Thumb Up

      Re: fc00::2

      indeed!

      as is ::1 as opposed to 127.1 :-)

  31. Hayden Clark Silver badge
    Boffin

    VPN and VOIP

    Both VPN and voip sessions require that a NAT router do some fairly sophisticated connection tracking on connectionless protocols. Voip is easier, as at least it uses UDP, which has port numbers as part of the protocol header, which means that the NAT process is free to tweak the source and destination addresses and port numbers to achieve a reasonably reliable pseudo-connection.

    VPN, particularly IPSEC and PPTP are very hard to run over NAT, particularly if there are multiple VPN clients in the local LAN. The protocol (GRE) does not have port numbers, and the payload data is opaque. This means that the NAT router needs to make some guesses to route the packets correctly, and so bad NAT algorithms are bound in a CDNAT situation to cause VPN connection issues.

  32. Rob F

    This is how I see it going

    The final block will finally be sliced up and exhausted and then we will be on to the exciting game of highest bidder wins. Like carbon credits, the price will go up as the availability decreases and especially ISP's and hosting companies will have to acquire ranges any way they can. I have already been involved in a project where a company had an entire migration to a managed datacentre for their web-servers (some major clients) which was pretty involved, for the princely sum of a /20 range that they owned.

    I also know Universities that have so many public IPs that they use public ranges internally so don't be surprised if they suddenly make a quick buck by selling some of those ranges.

    The tipping point is when the price of these ranges changing hands becomes more expensive than just upgrading the infrastructure/using 6to4/etc. The question in my mind is just how quickly these prices go up. If it is reasonably gradual, then the majority will have time to get themselves sorted. If if goes into the stratosphere very quickly, then get out the popcorn because it is going to get interesting.

  33. badger31

    Can they still call this The Internet?

    It sounds to me like they are providing access to the World Wide Web, and little else. ISPs calling this service 'The Internet' would be like calling a broadband connection with a download cap 'unlimited', and they would never get away with that. Oh, wait ... they did, and they probably will.

    1. Peter Gathercole Silver badge

      Re: Can they still call this The Internet?

      Why? For normal users who do not provide internet visible services, but only use client services, the change will be almost completely invisible. Outbound connection requests will still be given ephemeral port numbers, just like they are at the moment, and these will be recorded by the NAT server to allow packets to be routed back correctly.

      In fact, if you have a cable or ADSL router/modem, you are almost certainly running NAT already.

      It is only if you offer inbound services to your network that you are likely to notice anything at all, and if you are, you probably already know how to get around any problems. And it's not like they are not telling you what is happening.

    2. Alan Brown Silver badge

      Re: Can they still call this The Internet?

      Any ISP which signed the "code of practice" recently touted as the alternative to Ofcom stepping in (Which is all the biggies and phone companues except O2 and Tmobile) has legal restrictions on what they can call "Internet"

      Blocking GRE/VOIP or being in a walled garden would stop them being called "internet services" (which is why O2 and Tmobile won't sign the agreement - they both block VOIP)

  34. mark l 2 Silver badge

    If this service were offered at a discounted rate to the standard public IP service and explained the limitations of the stuff that won't work i can see a good percentage of their customers that just do web browsing, iplayer, youbube and facebook being happy with that. And if they do want to run something that requires a public IP address then they can pay the extra to upgrade to the next service.

    I seem to remember way back in the early 2000s that NTL did a similar thing on their low end 128meg cable broadband service where you got a NATed IP address but then they stopped and started issuing public IP addresses.

  35. t0m5k1
    WTF?

    WTF

    I thought +net we more clued up than this!!!

    When I get back to UK the only ISP I will consider will be andrews & arnold

  36. Jusme
    Flame

    Another small step...

    Another small step towards the internet becoming a pay-TV service. NAT'd connections are great for consumers, not so great for creators. No, I don't count uploading your life to Facebook as being creative.

    Now watch these nice adverts then you can see some cat pictures.

  37. Z80

    I don't know if it's a historical thing due to being with PlusNet née Force9 for years and years but in the member centre I'm seeing 4 public IPv4 addresses for my account. They're described as:

    Gateway Address

    address for router

    spare address

    Broadcast Address

    Why would they have done this?

    1. Jamie Jones Silver badge
      WTF?

      @Z80

      It sounds likes you have been/were assigned a /30 network rather than a single IP address....

      That would be a waste, because only one of the addreses is available to use (the other three taken up by net address/gateway address/broadcast address)

    2. xanthora

      Is this a business account?

      They have assigned you a network range (sometimes called a block), albeit the smallest range you can get.

      Gateway address is actually called the Network address - This cannot be used by a computer

      Address for router is as it says, the Public IP assigned to your router (this will be static)

      A spare address, simply because you cannot have a range lower than 4 addresses.

      Broadcast address - once again this cannot be used by a computer.

      If this isn't a business account, I would ring them up and ask them to remove the range, and simply provide you with a static address if you need one, or dynamic if you don't need it. This would free up a further 3 IP addresses for them to assign later (they would probably thank you aswell :D )

  38. JohnG

    IPv6: not enough incentive to move

    There is just not incentive to move to IPv6. Companies which already have sufficient IPv4 allocation (i.e. most companies) don't need to move to IPV6, they also know that a migration would cost them money and that it would carry risks that some of all of their services might not be universally available following a migration to IPv6. The ISPs won't move as they don't want to be in a situation where some of their customers have trouble accessing specific services or using particular applications under IPv6. That some other startup company is having trouble getting online due to a lack of IPv4 address space is not going to encourage any company to take the plunge.

    The only way to get people and companies to move is to have some substantial incentives - some things that are available under IPv6 but not under IPv4. I know there are some usenet servers offering free access to binaries newsgroups but that is not enough. If governments offered a limited period tax break for companies or individuals demonstrating that they had completed migrations to IPv6, that might generate some interest.

    1. brooxta
      Alert

      Re: IPv6: not enough incentive to move

      The internet tends to move fast in bursts and break old stuff. It's a disruptive technology. That's what has made it so phenomenally popular. The only way to survive in the presence of a disruptive technology is to change and adapt and keep up. Or else you risk getting broken by the next stage in its development.

      The incentive to move is history. Keep up or get broken by the next thing. Unless ISPs have IPv6 ready, debugged and waiting for the press of (a big red) button then some day soon their CEOs are going to wake up and discover they're 18 months behind curve and the masses are plunging head first into some IPv6-only thing which is exploding like Farcebook did. And the masses won't be able to do it on their networks. So they'll go somewhere else.

      Consumer devices, OSes, etc. are all IPv6 ready. They can all access IPv6 resources. The only things that are not ready are the mainstream ISPs. They're sitting on a ticking time bomb. There's your incentive. Perhaps they need to do some risk analysis on their business models?

      1. Ken Hagan Gold badge

        Re: IPv6: not enough incentive to move

        "Consumer devices, OSes, etc. are all IPv6 ready."

        All of them? Games consoles, ADSL routers, TVs and PVRs, particularly any of those that are more than six months old? On the other hand, most of them can work behind a NAT (whether it be domestic or ISP) without noticing. The device that really needs to be IPv6-ready and typically isn't (in the UK at least) is the ADSL router.

        "They can all access IPv6 resources."

        Yes, but (as noted earlier in this forum) the majority of online resources are only available through IPv4 addresses, so even where the customer and the ISP have both got their act together, you still end up using a IPv4 connection for most things.

        1. brooxta
          Facepalm

          Re: IPv6: not enough incentive to move

          In reply to Ken Hagan (15/1/13 17:52)

          Ah ... good point. I may have indulged a little too heavily in hyperbole there.

          The point I was trying to make is that smartphones, tablets, and the OSes that run on desktop systems are all IPv6-capable. They can all, given the right kind of network connection, access IPv6 resources.

          The routers (or whatever connects the devices to the wider internet (e.g. mobile 3G/LTE/whatever)) will be supplied by the ISP. And if the ISP is selling its service as proper internet (ie. IPv6 capable) then the router will be too.

          As for games consoles and TVs and PVRs, that's not really relevant to the point as, provided the connection is dual stack, they will not be affected.

          It seems at this stage the consensus is to move from IPv4-only to dual stack IPv4 & IPv6, and then more gradually still to IPv6-only. The problem that many El Reg readers appear to have is that mainstream UK ISPs are moving to dual stack at the speed of a very slow snail. And my point was that that puts them in a vulnerable position.

          While I apologise for the gratuitous hyperbole, I think my point still stands.

    2. Alan Brown Silver badge

      Re: IPv6: not enough incentive to move

      "There is just not incentive to move to IPv6. Companies which already have sufficient IPv4 allocation (i.e. most companies) don't need to move to IPV6, they also know that a migration would cost them money and that it would carry risks that some of all of their services might not be universally available following a migration to IPv6"

      IMO: If there are a number of IPv6-only services and IPv4-only ISPs can't reach them then they won't be able to advertise themselves as "Internet" providers.

      I raised this with Ofcom late last year. They agree in principle and are keeping an eye on the situation - the question is at which point the number of IPv6-only hosts becomes enough to warrant mandating that ISPs provide IPv6.

  39. Thomas 4

    Bad news for gamers

    If you play any kind of P2P multiplayer game, this will not be welcome news. I had any number of NAT failures when trying to play Company of Heroes and god help you if you want to play something like Civilization.

  40. Dave Bell

    ISPs get toenail clippers from Dr. Gatling

    When last I checked...

    1: My ISP had no plans for the transition. Given the likely life for their network hardware, that seems foolish.

    2: Nothing on sale in the fabled "High Street" was marked as having IPv6 compatibility. I was having a look at ADSL boxes at the time, with the aim of upgrading my WiFi.

    3: Windows XP has an IPv6 stack included.

    4: My NAS drive won't work with Windows 7 but it does have a USB connection. It isn't so big anyway, but what is it about my set-up that has my hardware working well for so many years more than anyone expects? Maybe it's the Spontoonie Gods who are the real ones.

    1. Ken Hagan Gold badge

      Re: ISPs get toenail clippers from Dr. Gatling

      On (2), I can't claim to have performed any market research myself but Andrews and Arnold are offering a Technicolor 582 router as part of the IPv6-capable package. They searched long and hard to find it. Chances are its the only reasonably priced ADSL box they could find.

      On (3), yes even XP has an IPv6 stack, but the end-user has to do "difficult scary stuff" to actually switch it on so I imagine it is "off" on 99% of XP systems. I wonder if Microsoft would consider a Windows Update or "FixIT" to change that default, or would that count as "maintaining XP" and therefore violate someone's religion.

  41. Number6

    Already Here

    It is possible to be in the UK and use an ISP with native IPv6, I've got it here. It's also possible to set up your home system to use a tunnel broker and have IPv6 that way. It wasn't that hard to set up a Linux router with a tunnel, although I appreciate it's not quite plug-and-play.

    Two things are needed:

    1. Router/modem manufacturers to include ipv6 in their products by default so that if connected to an ipv6-enabled ISP, it'll just work as easily as ipv4. I can see that it might be necessary to include an entry field for an ipv6 prefix address, but that's no worse than having to enter the DSL modem details. Most modern PCs (Windows, Linux, Mac) will trivially fire up ipv6 with minimal prodding, and even Android phones will use it if on a suitable network.

    2. ISPs to bite the bullet and offer ipv6 as a default option so that it will just work, and most people won't even know it's happening.

  42. Anonymous Coward
    Anonymous Coward

    Software compatibility?

    It is one thing for networks to move to IPv6. However users' routers, clients, and servers also need to support IPv6.

    On top of those compatible platforms there are then applications which might be IPv4 address aware - and now need to handle IPv6 format.

    1. Alan Brown Silver badge

      Re: Software compatibility?

      "On top of those compatible platforms there are then applications which might be IPv4 address aware - and now need to handle IPv6 format."

      This is a real problem, but thankfully most enduser apps will just work happily no matter which IP stack they're talking to.

      A lot of P-t-P apps are going to break though. The entire Ed2k/Kad system for starters.

  43. Anonymous Coward
    Thumb Down

    And another ISP that doesn't get it

    If they had just spend the time and effort of setting up CGN in setting up a dual stack system, they would be ready for the future. The more ISPs that do that, the more people who will add IPv6 to their servers.

    If I look at my company's accesslogs, I can see that about 1 in every 200 requests comes over IPv6. Thankfully, some more ISPs have announced IPv6 availability for new connections recently, so this should rise soon.

  44. Anonymous Coward
    Anonymous Coward

    Maybe

    "The solution is for everyone to switch to IP version 6"

    What if it isn't?

  45. Anonymous Coward
    Anonymous Coward

    Yes, I'll have that!

    When you get caught sharing files on P2P then the "enforcer" using the IP address as proof won't be know who was actually sharing the file?

    Mobile phone networks already do this anyway, it's not a new concept.

    1. Alan Brown Silver badge

      PtP requires holes punched in NAT gateways on at least one end.

      If you're behind a double NAT and you're "Sharing" files on PtP then you may as well not be sharing at all.

  46. Craig Vaughton
    WTF?

    Sheffield ISP

    You mean that wholly owned subsidiary of BT?

  47. Alan Brown Silver badge

    "NO fucking way!"

    I've been behind double-NAT setups in the far east. They're a disaster.

    Just get the fuck on with rolling out IPv6.

  48. Alasdairrr
    FAIL

    Plusnet are INSANE

    This is absolutely crazy!

    They could at this point so easily adopt “DS-Lite” – it’s identical to what they’re doing (carrier-grade NAT of IPv4 space) except also dishing out native IPv6 on top.

    It’s the perfect time for them to start the transition by rolling out a dual-stack network, instead they’re being complete idiots and very bad net citizens.

    They’re going to get very bad press/PR over this, and I hope, a customer revolt.

  49. Anonymous Coward
    Linux

    PlusNet and IPv6

    Two years ago I asked my ISP (PlusNet) what their plans were for IPv6. I was told "we have no plans for IPv6" so I switched to AAISP.

    I now have a single IPv4 address for WAN and a /29 for LAN and a /48 for IPv6 which I call off in /64 chunks for each subnet I need (!)

    The real issues I see remaining with IPv6 are multi link routing without some form of NAT (my office has six ADSL lines) and the mayhem caused by a switch of ISP without your own PI and BGP (prohibitively expensive for home and small business)

    1. andretomt
      Boffin

      Re: PlusNet and IPv6

      If your ISP is willing to be a sponsoring LIR, getting your own AS number (and a PI - Provider Independent block) for multihoming it is not that expensive anymore. That is - you dont have to pay RIPE for the pleasure like you used to. However the ISP is likely to require you to be on a business plan - it is extra management and paper work after all.

      Meaning, PI is likely to be much more common in IPv6, than it was with IPv4.

      There are other options too, like prefix translation (a form of completely stateless, and thus fast and scalable NAT), just announcing a prefix from each ISP to the local network, or if the network is sufficiently simple, using IPv6's rapid renumbering features.

  50. -tim
    FAIL

    Wrong shortage

    There is no IPv4 address shortage, there is an IPv4 route shortage which has resulted in a vast minority of IPv4 addresses not being useful for anyone.

    The slash notation was originally a plan to steal bits from the source and destination port addresses so that a 1.2.3.4/34 would extend the host address by taking two bits from the ports much like NAT and PAT do today. The problem with that approach is no one came up with a clean notation system or rules about exactly where the bits came from.

    I was a strong supporter of not consolidating routes in 1991 (since it was a temporary crutch) but Cisco routers at the time just couldn't cope with many routes and claimed it was impossible but AT&T delivered a router that could treat the entire address space as Class-C in 1993 and cope with 16 million routes.

    The artificial route shortage has another drawback in that dual homed systems are only for the large players. If IANA was doing their job, I should be able to go to them and say "I want a /28 that works with both Telstra and Optus" or "I need a /32 to work with these two ADSL providers" and they should be able to allocate me a block that both ISPs have agreed to broadcast and would properly route between them dealing with normal failure modes. If IANA had a clue, they would force all remaining addresses to be allocated this way and ensure that ISPs would never get another address as long as they wouldn't do the right thing and work with their competition to ensure that the end customers of the net got what they needed to provide the redundancy that is now essential for many small businesses.

  51. andretomt
    Thumb Up

    Quad NAT

    Its not really double NAT when everyone have to implement it. When both ends of a "peer to peer" connection is behind a carrier NAT, it becomes quad NAT. Thats right, *four* levels of NAT. For games, voip, Skype and anything peer to peer-ish this is going to suck. Only option will be latency inducing detours through centralized choke points (maybe not even on same continent, I see this already) every time NAT port mappings fails, which will be pretty much always. Quality will suffer, and running the services becomes more expensive as more infrastructure is needed to support bouncing all that extra traffic, perhaps even prohibitive for new players.

    Skype for example, will often have a much better experience when it is able to connect the participants directly together, today. With at most one layer of NAT at each end, it often succeeds too. When it fails to it bounces everything off Microsoft.

    This might be a good thing though. Soon the IPv4 experience might be so miserable that "usable internet" becomes the IPv6 killer application. I hope.

    1. Anonymous Coward
      Anonymous Coward

      Re: Quad NAT

      The number of levels of nat are irrelevant when they are in the same direction. The double nat problem is really that it is occuring at both ends (implying both directions)

  52. Anonymous Coward
    Anonymous Coward

    "Mobile operators have been doing this for years: millions of customers on a single network typically share three or four IP addresses between them,"

    You can't have millions of customers sharing three to four IP's. A single public IP can only handle 65535 connections. You will typically have multiple connections, browsers can use up to 8 or more at times. You can do thousands of users, not millions.

    You can do port forwarding even if the NAT is owned by the ISP. it comes down to *IF* the ISP wants to support it. They would also need to make it some the common use ports can't be used; like 80, 443, etc. They could give the customers X number of port forwards, so if a customer wanted to run a web server on for 6432 they could. The person trying to access it would need to make sure they used that port.

    The proper term is CGN; Carrier Grade NAT.

    1. andretomt

      Mobile ISP's often do this already yes, and its one of the reasons mobile internet is so fricking painful. At least the ones I know of have it opt-out, if I need a public IPv4 I just set another APN on my device, which makes it a bit more bearable for me personally. But that is not really enough for the health of the internet as a whole. When a significant portion of the internets users are crippled, the usefulness of the intertubes suffers in some way or another even for those with proper connectivity.

      Anon is right that you could portforward at the ISP in some cases, you just need a new home gateway/router/cpe to support UPNP IGD/NAT-PMP to PCP proxying ;-) And you will still need some 3rd party negotiation of the ports if its going to be useful for other than techies (on the other hand, not as infra/latency/bw critical as a full detour). At least these new devices will also support IPv6!

      FWIW, my cable operator have provided me with proper native ipv6 for almost a year at home, as many routed subnets as I need, at work there is ipv6 to the workstation, mom has native at her place, dad has at work (home still a few months out). Cable, FTTH and DSL techs covered. Really UK ISP's need to get their shit together and just deploy IPv6. Its not that hard and others are doing it AT SCALE already.

  53. Chris Beach

    IPv6 Plan

    If this was announced alongside their IPv6 plan, then that would have been a good thing. "look, we're investing in a proper solution, but in the meantime can you help us out buy..."

    As it is it looks like they don't want, or can't make the switch to IPv6, and are happy to provide a sub-par service.

  54. Chris Collins 1
    WTF?

    we need to do dual stack instead

    This seems a cheap way out for isp's, seems this country yet again always doing the cheapest option. AAISP might get a lot of new business in the upcoming yers because of this as their own plans seem better. The only real way forward is implementing dual stack 'asap' and then eventually over time ipv6 will replace ipv4. Thumbs up to germany who seem to be doing it the proper way. Also to blame router firmware is nonsense, the isp's in this country are the main customers of the router vendor's so have a big say in firmware direction.

  55. sabba
    Facepalm

    Shared IP addresses...

    ...imagine having Gary Glitter and Jimmy Saville (ok, I know he's dead) as your neighbours and you're all on Plusnet. "No, officer. I've no idea who downloaded the picture of that scantily clad eight year old".

    1. Christian Berger

      That can be solved simply

      NA(P)T requires the router to have tables tracking the connections. So before you know it, data retention will be extended to individual TCP connections (and UDP "connections" such routers think in).

      So what will be logged is no longer "Person X had IP-address Y", but "Person X had connections to the following servers with the following ports". It's trivial to abuse that kind of data for repressive purposes.

  56. Christian Berger

    The only reason I still have IPv4...

    ... is TheRegister. Seriously most of the world already has IPv6.

This topic is closed for new posts.