Follow Slashdot blog updates by subscribing to our blog RSS feed


Forgot your password?
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. ×
The Internet

IPv6 Promotion Effort. 138

rafa writes "The IETF may soon launch an IPv6 promotion effort. The new IPv6 is an improvement over the current standard IPv4, that has a larger amount of addresses available, improves routing and has several other benefits. "
This discussion has been archived. No new comments can be posted.

IPv6 promotion effort.

Comments Filter:

  • It's not FUD because it already happened. When Microsoft needed a TCP/IP stack quickly because Netscape was fucking them on the Internet they took *BSD TCP/IP stack. This probably wasn't just a cut&paste (needed some reenginering) but this wasn't developpng their own stack.

    But this don't seem to happen with IPv6 (they are developping their own stack).

    BTW, I know that linux support IPv6 but how god is the implementation?
  • So any newer tech will not work on their just-previous release of the OS.

    That's less than accurate, and Micros~1 know it. FTP Software announced IPv6 support in their Winsock 2.0 stack for Win95 in 1996 - though I don't remember if they ever delivered one with the promised support.

    (1996 - that's three bleeping years ago. I thought progress was fast in this industry.) :-P

  • DHCP is evil!!! it's such a pain. especially when your network is up but the dhcp server running nt is dead.

    Good thing for my job that we're running DHCP on something other than NT. "What?" you ask... We don't need NT? Nope.

    The sales guys can take notebooks back and forth between offices and all they need to do is plug into the network. They're happy and I don't have to answer questions about how to reconfigure their machines.

    For those who are curious, we run dhcpd on linux, but it's been ported to other unices, i'm pretty sure.

    So, um, get a clue.

    A host is a host from coast to coast...

  • There is a shortcut. All the null sequences can be left out. So your addresses are:

  • Apple got the IPv6 stack as part of its OpenTransport streams architecture from Mentat. It was demoed a few years ago but hasn't yet been released into the official MacOS.
  • by Anonymous Coward
    > The main reason for IPv6 - the potential address
    > shortage

    This is one reason but not the only one. It is a silly reason to be sure and not a good enough one to convert to IPv6. IPv4 wasn't meant to be used the way it is today and has been hacked to make some things work. the IPv6 protocol removes uneeded clutter from the IP header and generally redesignes the way the packet is layed out, etc. The biggest advantage, imho, of IPv6 is that it is more efficient (better routing for example). Of course we can decide this isn't important and just keep throwing bandwidth at the problem which might actually work for awhile but isn't the right answer in the long run.

    I agree though that the biggest problem is the cost. However, it is more than the cost of REPLACING old equipment. Routers are being replaced for other reasons, for Gigabit Ethernet, for example. It isn't a matter of how much is it going to cost to replace my router as much as a question of when I replace my router how much is it going to cost to get an IPv6 capable router rather than one that just does IPv4?

    So the move to IPv6 probably won't happen until the cost of IPv6 capable routers and such (specifically ones that are capable of BOTH IPv6 and IPv4 for the duration of the transition period) become not much more expensive than IPv4 ones. A manufacturer isn't going to drop the price on IPv6/IPv4 capable routers because he isn't producing as many and they cost him more and he therefore makes less profit on them. But if the market demand for them rises then manufactuer's will produce more IPv4/IPv6 capable routers and find ways to produce and sell them at nearer the cost of IPv4 routers. It's a nice little circle. Some people won't bother with IPv6 if it's too expensive but they won't get cheap enough for those people unless enough OTHER people demand them.
  • There already is an allocation system setup. Check with the specs to see the details. Offhand there is the geographic layout that you suggest but the preferred addressing scheme at this point is based on provider. This greatly simplifies the routing since the address acts as a builtin list of directions over the internet's true topology.

    The fields are already of variable bit length.

    The design is intended to allow multiple addresses for each interface that corresponds to a different provider. This way the routers can get the clients to switch providers based on who has the cheapest rates at the time of connection.
  • by drendite ( 3 )
    Cool.. Note AAAA is not a part of the address, it is only the DNS record for ipv6 addresses (like MX is for mailing)
  • I'd like to ask a question I read in an article about IPv6 a while ago. Now that IPv6 uses : to seperate different sections of the address, how do you specify ports? If I use or whatever right now, how do you do that in IPv6? Might be a dumb question but I haven't seen the answer yet.
  • by Anonymous Coward
    All you need to get your block of 2**80 ipv6 addresses is to ask one of the access sites (listed off one of the links on And ignore that "experimental" crap -- if history repeats itself, the "experimental" network will become the real network, despite what anyone says.
    -- Guges --
  • I won't be able to connect my PDP-11 :~(
  • No, he/she (let's not discriminate) are just an uninformed net user that doesn't have any idea how the internet works or why. Oh I pine for the days when you actually had to meet certain criteria to get access to the internet, such as having a real good idea of how it all worked.
  • So let me get this right: You mean I'm going to have to remember sixteen 8-bit numbers now to get to my machine? Dangit! If only there were some way to hash an easy-to-remember name to an IP address . . . now that would be really useful. Oh well.

    (I know the script kiddies are not gonna like this. "Hey! I got wArEz and p0rn at 20.221! What's your site?")
  • Look how long it's taking for Unicode to be adopted... will IPv6 be any faster?

    It'll take decades... if it happens at all :-(

  • by Anonymous Coward
    Great. I'll pick '7'.
  • by Anonymous Coward
    Actually, IPv6 will be adopted fairly fast due to the simple fack that we are about to run out of IP adresses.

    It's also designed to interoperate easily with IPv4 - all IPv4 adresses maps directly to IPv6 adresses, so that it's easy to set up gateways.

    IPv6 will likely first make it's entrance behind firewalls, and in situations where only limited access to the IPv4 "old" internet is required.

    You can use IP-in-IP tunelling to gateway IPv6 content over an IPv4 network, and you can easily use transparent proxying/masquerading with any host that supports it and both IPv4 and IPv6.

    Many backbone routers etc. can also start using IPv6 fairly soon, since it's typically not a requirement for ordinary users outside the backbone provider to access linknets or specific routers within the net.

    Another way to help transition is to use IPv6 with IPv4 compatible adresses whenever all equipment on the LAN supports IPv6 (or enough that you can route the rest via a translating router or machine). It won't give you that many benefits right away, but it means that whenever your acess providers starts supporting IPv6, you can move over seamlessly.

  • We asked about this at a recent conference.

    Microsoft's response was something like "We will support it when there is a standard. As of right now all support for IPv6 from other vendors is still pre-beta, and we don't see any point in wasting our time with that in a shipping product."
  • by Anonymous Coward
    There is a general crackdown on address allocation .. I got my class-C just before our local mom+pop ISP stopped offering them and moved to 16-address subnets. I expect it to get harder before it gets easier, but the adoption of IPv6 should make it easy again.
    There are two sides to the dynamic IP vs static IP issue: the technical side and the psychological side. On the technical side, static IP's are only a logistical nightmare because ISP's route by *contract* (ie, who is paying them how much for how good of a connection -- talking about routing between ISP's here, not between the ISP and the end-user). If routers began routing according to optimal path (or a reasonable facsimile -- eg OSPF) then the complexity can be more completely hidden by automation, the way the Internet was supposed to work. Whether or not routing policies will change depends on whether it becomes more profitable to utilize bandwidth more efficiently than it is to negotiate complicated contracts, something that's in flux right now. On the psychological side, there is a *huge* established network of ISP's who think of the network as organized in a certain way -- static IP's for routing between ISP's, and dynamic IP's for end-users. The assumption that this is "just the way it works"
    has caused a lot of proprietary software and network infrastructure to be developed to *only* support this model. These johnny-come-lateys to the internet (ie, almost everyone) don't look at the new possibilities that the new protocols open for themselves or customers -- they just want to keep doing things like they've always done them. There is tremendous resistance within companies to change a process once it is written down. What this means is that even though IPv6 makes it practical for everyone to have their own block of 254 (ie, 256 sans one for network (0) and one for broadcast (255)) addresses -- maybe one for their watch, one for their walkman, one for their home computer, one for their microwave, etc -- the ISP's won't make this option available because there is a lot of momentum behind the concept of using dynamically allocated addresses, one per customer. A seasoned engineer writes a process to be as modular as possible, and to abstract away the details of the implementation, but most
    internet companies are not run by (or even necessarily hire) competent engineers -- implementation-details can often be found as far and wide as customer billing policies, bug reporting and tracking, et al. For instance, look at Pacific Bell, who is limited by their *policies*, and not by any limitations intrinsic to the technology, to associate DSL lines with POTS voice lines. A friend of mine has been having a hell of a time trying to get them to call him using his voice line, and not the (unused) POTS line they installed with his ASDL line, when providing technical support for that ASDL line. He's told them that there is no phone connected to that line, and to use , but their support process just isn't set up to handle it. It's stupid, it's suboptimal, it should be better, but it's the norm.
    -- Guges --
  • I think a better reason that we'll switch to IPv6 relatively soon is because there's a lot of money to be made during the switch-over. There aren't a lot of economic incentives associated with switching to Unicode...

  • 7 isn't a random number. But 17... that's a good random number...

  • by Anonymous Coward
    All addresses that start with zeros, then FFFF, and a 32 bit number, in IPv6 is a one-to-one mapping of IPv4 numbers. So you'd start loosing access to lots of stuff when more and more sites move to IPv6 only, but most sites will likely at first move to IPv6 with an IPv4 compatible address.
  • Actually, I believe this is "taken care of" in the IPV6 standard. It's my understanding that there is a "provider" part of the address and a end-user part of the address. There has also been some talk on the 6bone list within the past month or two about how the organization that will be/is handing out addresses is setup and functions.

    As far as routing, that's in the IPV6 "standard." They can't use the same routing protocols as in use today.

    While I'm not sure that my statements are 100% acurate, I do believe that they are somewhat more than yours. You should at least know something about a topic before posting...

    Check out RFC2472, it describes how IPV6 addresses SHOULD be chosen for PPP links. It appears to me that if a node (your home PC) has any EUI-48 or EUI64 address configured on any interface, it should use these addresses in "suggesting" an IPV6 address to the PPP peer in the config request. So, if your PC has an Ethernet card, as mine does, then you should use your Ethernet MAC as part of the IPV6 address (the "interface" part). There is even a method for turning the 48-bit MAC into the 64-bit interface ID.

    All IPV6 related RFC's are available via 6bone []
  • All right, this is the second comment that I've seen that I would personally classify as FUD. I wish I had moderator rights at the time so I could downgrade the comment.

    If you read the RFC's you will see that they have addressed all of the concerns you list above. There are already two other comments listing some specific responses to your concerns, but I humbly suggest you visit the 6bone [] and read the friggin RFC's yourself.

    I thought this site was supposed to be "News for Nerds. Stuff that matters." I also thought the "nerdy" thing to do would be to read up on a topic before making senseless comments. May be that's the "old nerd" way of doing things.

    I wonder how many (what percentage of) people on /. even know what a RFC is anymore...
  • by Anonymous Coward
    I need my ISP's *routers* to support v6; I don't particularly care what their servers can do. And since real ISPs route with dedicated hardware (general-purpose computers have serious trouble with that sort of volume), upgrading isn't easy.
  • Yes, you read that right. MS's research group has been releasing their stack with source code for over a year. The license even allows for redistribution of changes, so I think this fits within ESR's definition of "open source" but IANAL.

    Check out [] for the scoup.

    So I suspect MS won't be needing to copy the *BSD work. If anything, it looks like they're farther along.

  • by Anonymous Coward

    The main reason for IPv6 - the potential address shortage - has been blown out of proportions.

    While it is true that this is not a problem yet, it could become a problem of Y2K proportions if not headed off soon enough.

    Another thing to think about is new types of Internet devices. Today, we're mostly just running IP to desktops. If you want every cell phone in the world to speak IP, however, you've suddenly increased the number of needed IP addresses by a lot. A whole lot. A since cell phones are mobile by nature, you'd really like those addresses to be global, not hiding behind NATs.

    And if you want to make every electronic device in your home Internet connected (even if just to set the time automatically via NTP -- no more flashing VCR clocks) than you want IP security (which comes with IPv6) and global addresses. NATs come with a whole slew of interoperability problems that will bog down development of new IP devices. IPv6 is a much better solution.

    There are already more people on the planet than there are IPv4 addresses. There will never be more than there are IPv6 addresses.

  • Look how long it's taking for Unicode to be adopted... will IPv6 be any faster?

    When people start getting denied access to the internet because there aren't enough ip addresses, I'm sure the collective screaming will be enough to make sure the transition happens relatively quickly.

  • THis is not the cost of IPv6-capable router
    (well, practically all the routers used on
    INTERNET backbone are IPv6 capable) but the
    *cost of converting the INTERNET backbone*
    which is the major stumbling block.
    I couldn't access it of course.

    If more sites like this emerge, demand for ipv6 will increase correspondingly.

    I don't think there's anything interesting there, but I wanna know for sure dammit!
  • So I suspect MS won't be needing to copy the *BSD work. If anything, it looks like they're farther along.

    How so? Have you even looked at the *BSD efforts? There are two major stacks that work quite well with the various BSDs, and in fact the US Military has developed a third "major" stack. And MS has one pre-beta one. Hmm. Which is further along?
  • It's, not (Dunno what is--InterNIC doesn't even seem to have a listing.)

    And as to your comment that the experimental network will become the official one--it can't. Unless we continue using IPv4 forever. The 6bone tunnels through the IPv4 network, overlaying itself on top of it. If IPv4 goes away, then the 6bone does, too. (This also means that some features of IPv6, such as anycast addressing and host mobility, cannot be fully realized until the underlying network is all-IPv6.)

    Furthermore, the 6bone has a temporary TLA assignment. This means that only a fraction (an infinitesimal fraction, though still pretty huge by IPv4 standards) of the total address space is available in the 6bone. Going with the 6bone as the official network would radically reduce the total address space available. It would also enforce the current address allocation mechanism (which is very bad and intended to be temporary) forever.

    In sum, you're wrong. And anyone who deploys production services on the 6bone risks getting bitten.
  • Oh please. I've had more "per capita" trouble with my previous Linux based ISP ( than I have with Hooked -> Wenet -> GST.

    DSP was very unorganized and just offered wretched performance. Hooked incidentally ran BSDi modem and smtp+pop3 servers and now has a FreeBSD news server.
  • And most of those fscking loosers run Linux. What's your point?
  • pay a major pron site (persian kitty or playboy or something) to move to an IPv6-only network. i guarantee that ISPs will be forced by user demand to switch over asap.

    never forget the power of the masturbation superhighway.
  • They've been doing this since the later DOS versions...

    At this time, Microsoft Research has no plans to support this experimental stack on Windows 95 or Windows 98.

    So any newer tech will not work on their just-previous release of the OS. Unless a 3rd party makes it work. In which case you've 'invalidated the warantee' or whatever, and if you're a clueless newbie (you know who you are :P ) who needs to call MS for support you're basically fucked.

    The site doesn't look MSish which is good.. but they handle themselves just the same, wether they release a buggy proprietary code or not.

    you know what they say, if it looks like a duck, quacks like a duck...
  • No way! I was gonna pick 7!
  • Thanks. Any info about why they decided to use ':' instead of the trusty '.'?
  • You can't invalidate the warentee when it comes
    to Windows.

    It explicitally dosn't have one.
  • The IPv6 protocol allows headers to be hooked on after the main header that can hold other data. That means then an old SunOS box with an normal IPv4 address communcates with an IPv4 machine an
    gateway embeds the IPv4 header inside an IPv6 packet thats sent over the IPv6 network, then it
    arrives at the network its translated back into an
    IPv4 packet thats sent to the taget host with only IPv4 stack. this way IPv4-to-IPv4 communcation over an IPv6 network works fine. I guess is that IPv4 and IPv6 will co-exist in parallel for many years to come.
  • Read the Fu**ing Comment. DUH

    P.S. Its accually Request For Comment, I'm just being silly.
  • Part of the design of IPv6 was "avoid lots of little class C adverts in the backbone routers". Each router only needs to hold local information, instead of the 40,000+ IPv4 entries in the top level routers. If you change provider you also change your IPv6 addresses, v6 has built in support for easy renumbering.
  • Us techies can only take on so many things in mass. Right now we are still working on tricking the world into hiring us for making sure their pentium III's are Y2K compatible. Once we've gotten all the money we can outta that (sometime in mid January) we'll suddenly announce that by some date in 2001 we will run outta IP numbers, and that the world's computers will suddendly not be able to connect to the internet due to the problems of dynamic IP which are a rare resource like oil. But we can fix their problems with a genius idea called IPv6 which will give them 19^23 (or something like that) possible numbers. Everyone will be stressing up untill everything is declaired cleared, and till the prospeced date even. Then we will find some other "Tragedy"
    (note: most of this was sarcasm)
  • You idiot.

    How many Cisco 7500's run on Linux?

  • But i think if they upgraded to IPv6 they should make it's structure compared to IPv4 similar to Unicode's structure compared to ASCII. It would make things alot simpler, make it backwards compatible with the 32bit addresses but then let you also use 128bit addresses. So each domain of the address would go from 0 to 65000 or so. which means all current adressing using IPv4 would be compatible with the IPv6 addressing. It should be as simple as possible but not simpler, everyone should have the ability to use the internet, not just those who have the money to buy new hardware and software.
  • A few weeks back I was thinking about the ultimate way to get people to switch to IPv6, and then it hit me...

    Setup a giant free porn site with only IN AAAA DNS records.


  • IPv6 has a variety of other mechanisms for securing corporate hosts more flexible and less intrusive than NAT.
    Employ me! Unix,Linux,crypto/security,Perl,C/C++,distance work. Edinburgh UK.
  • But with IP6 there will be no need to use DHCP or other dynamic address schemes. Dial-in users could have static IP addresses, which would make life a lot easier for most people. It would make it easier for the dial-in user, and make it much easier to trace abuse etc as the IP address would identify the customer/user rather than an ISP port so there would be no need to correlate the IP address with logs to identify who was using the address at a particular time.
  • Back when IPv6 was first being discussed, I'm sure I recall seeing a fair amount about guaranteed QoS -- i.e. an ISP could sell a service with guaranteed n bandwidth and m latency between two addresses, for a price; and if you couldn't afford the price, you could accept a lower quality service.

    This sounds great -- I'd love to be able to pay a tenth of the price for a nice, slow-but-permanent, link for my email, while the people who want 128K for their Quake matches pay their own way -- but I don't see any mention of QoS on the website.

    Did that stuff get left by the wayside?
  • Actually, I think that you can only use the :: notation to specify a field like so: 0000:0000:0000. I'm not sure if this is still the way that it is, my text on the matter being from 1996. I believe IPv4 notation will still be allowed for some time while the transition is going on.
  • Why should the interchange points be involved in IPv4/IPv6 issues? At least the Stockholm D-GIX is only a link-level interchange.
  • In the past proposal address was variable-length, but this was rejected in standardization process because it gets harder to code.
  • QoS still needs big discussion until put into real wide area use. So, IPv6 reserves necessary header field for future use by some-QoS-protocol-deployed. This is the best thing IPv6 guys can do at this moment, I believe.
  • by rafa ( 491 )
    I think that there's an experimental IPv6 network called the 6-bone. It's used to test IPv6 in a real-world situation, and get any glitches out before wide-scale adoption.
    I don't know how open it is. it might be interesting to find out - since Linux already has support.
    IPv6 should be very easy to use for new users since the specification includes autoconfiguration.
    It also includes end-to end encryption, and flow labelling.

  • We can be free at last of the scourge of IP NAT!


  • There's a HOWTO at


  • by drendite ( 3 )
    Here's some real IPv6 addresses and how they look: AAAA 3FFE:B00:C18:1:0:0:0:10 AAAA 2010:836B:4179:0:0:0:836B:4179 AAAA 0:0:0:0:0:0:836B:4179 AAAA 3FFE:1200:2001:1:8000:0:0:1

  • by rafa ( 491 )
    More info at
    That page is supposedly also assessible via IPv6...
  • I don't understand what all the fuss is about. I've read up on this a bit from the IPv6 website. In reading , I realized that IPv6 has already been implemented.

    The preferred form is x:x:x:x:x:x:x:x, where the 'x's are the
    hexadecimal values of the eight 16-bit pieces of the address.



    I thought that FEDC:BA98:7654:3210:FEDC:BA98:7654:3210 looked remarkably familiar, and sure enough, I was right! It was my W98 CD Key!
  • Posted by FascDot Killed My Previous Use:

    Linux has been ready for IPv6 for quite a while. Why are we waiting?

    Here's how:
    1) Ask your ISP if they use Linux (or other IPv6-aware OS) on their servers. If not, find a new ISP.
    2) Tell them to get on the 6bone.
    3) Get on the 6bone yourself.
    4) Ride the wave of the future.

    Put Hemos through English 101!
  • Posted by 2B||!2B:

    I just worry that if the whole address space isn't properly allocated then we could end up with a true nightmare. I place 90% odds that it won't be done in a very practical manner initially (which makes it even harder to fix later).

    It would be cool if IPv6 were modified in a DNS kind of way. Instead of 128-bit numbers, I would much prefer the option of something like:

    [country].[region].[industry].[specialty].[compa ny].[division].[department].[machine/user]

    where the subdivisions aren't necessarily 8-bit, but are instead scaled to need, and there aren't necessarily 8 of them. Each blank could be filled by either a name or a number, as appropriate. This could also allow more than one path to the same site when appropriate. It's potentially long, but could be significantly abbreviated in a local context (which IPv6 is going to do anyway). I've always thought normal DNS naming is rather worthless, anyway. (something being under .COM or .ORG tells me absolutely nothing)

  • This looks like MAC Addressess on steroids. Looks like I'll be setting up DNS for all my IP's now, as it might be impossible to remember them otherwise.
  • Wellfleet - er - Bay - er - Nortel has IPV6 capabilities on thier routers. Are you telling me that cisco does not support IPV6 in their latest IOS?
  • I've always thought normal DNS naming is rather worthless, anyway. (something being under .COM or .ORG tells me absolutely nothing)

    That is not the fault of DNS, but rather of Internic and the way they chose to deal with .com/.org/.net.

    I am well aware that at a certain point, they got "too many" applications for domain names to be able to handle this, but applying a sensible level of bureaucracy (and educating people) could have helped.

    Myself, I've even helped to add to the confusion by letting a few Norwegian companies get through with registering under .org and .net, even though they don't belong there. So I'm definitely to blame, too.

    BTW, how were you planning to guarantee that some dweeb doesn't register his company under the wrong country/region? "Hey, look, we're a multi-national company, and..."
  • Well, perhaps it's about time noticing that "the INTERNET backbone" (as you call it) is continuously being replaced.

    BTW, what do you mean by that? What else on the backbone but the routers do you think need converting that will make it so much more expensive ("major stumbling block")?
  • Er, lots of places loose power at the same time on a regular basis, or so I would think.

    Millville is having thunderstorms and the high was supposed to be 97 today. I'd say it's probably a combination of those thunderstorms and brownout because of too many people cranking thier AC's.

    Cape May is similar, with thunderstorms and a high of 95.

    Fullshear is having thunderstorms all the way through Thursday... (ever hear of lightning striking a power distribution field?).

    Riyadh/Khaled has a high of 109 - holly crap! but "plain old" Riyadh is only 32 - must be celcius... Besides, I don't know enough about their power systems to trust it anyway.

    Get a life, and a clue. If you took five minutes to check out the weather [] at these locations you can see that there is a pretty good expectation that they would loose power at this time.

    Now you go to bed with no cookie and no fireworks (Since 75% of your sites were in the USA I figured you're in the USA and would have been able to watch them if you behaved properly).
  • so it ain't goin away any time soon
  • If you write domain name in URL, you do not need to say "ipv6". For example, has both A and AAAA records and you can just specify [], and you'll connect to either of them (just like when a web server have multiple IPv4 addresses).

    Numeric IPv6 address is little trickier because they have colon inside. for this there are several draft submitted in IETF.
  • ::ffff: type of syntax is just for use inside IPv6-capable node, so that programs that runs on IPv6 socket can manipulate IPv4 connection on the wire.

    Transition issues are discussed in IETF ngtrans working group, so it may be useful if you check IETF drafts named "draft-ngtrans-*".
  • Maybe the 6bone is working ok, but a bunch of the proposed standards already seem kind of dated, and some parts of the proposal seem frankly dumb. Tying IPv6 addresses to link-layer addresses (and using half the 128 bits for it!) is dumb - it makes it very difficult to switch hardware around without having to wait for DNS changes to propagate, and there are plenty of other reasons to object to it too. The proposed reduction in routing tables is, I believe, mostly a sham. There may be a factor of 2 in there, but with the complexity of today's internet topology, I find it unlikely they'll cut down the size of the backbone routing tables by very much. And really, handling more routes is just a matter of more memory, and memory's cheap. The current full BGP internet routing tables contain somewhere around 100,000 address ranges - this fits comfortably in 64-128 MB of memory (even with multiple entries for each set of addresses). The handling of multi-homed sites under v6 is pretty much left up in the air - there is no good solution to real multi-homing (ie. a single site with multiple ISP's) other than an entry for the site in almost every relevant routing table - and the RFC's acknowledge this won't help much if we have more than a couple of hundred thousand multi-homed sites on the internet. And some of the proposed topology diagrams seem laughably antiquated.

    Somebody really needs to study how the various parts of the proposals really match the stated goals of IPv6 given the current (and evolving) structure of the internet - it looks like the proposals are very much based on the internet of 5-10 years ago, and a lot has changed.

    Here's an alternate proposal - simply prepend 96 bits to IPv4 addresses, all zero for current IPv4's, and then sell the new address space to ISP's and exchanges on a per address basis. The proposed IPv6 protocols could still to be used. And we could switch to fancy hexadecimal notation. I bet this would serve the original goals almost as well with a lot less disruption.
  • > Thanks. Any info about why they decided to use ':' instead of the trusty '.'?

    Because the similarity to MAC addresses is no accident: The idea is that you get your own network instead of IP address, and individual addresses are as easy as tacking on the MAC address to that. It's already unique at that point even without the network part (theoretically, MAC collisions can and do happen occasionally) but the rest of the address is a routing hint, you don't want to keep a flat ARP cache for the whole damn internet after all.

    Also because now that they can contain letters (A-F anyhow) it would be harder to distinguish them from names, which will probably still use dots at least as long as we continue to use DNS (I really don't see anyone pronouncing an X.500 address on the radio).
  • I imagine TCP, which is also crufty and largely ineffective at stopping sequence number attacks and the like, is also being upgraded? Is it going to be a higher-level protocol, an extension header on IPv6 packets, or are they just going to have plain old TCP wrapping IPv6 packets instead?
  • It's FUD because it can't happen. Have you tried to cut'n'paste the tcp/ip stack from say FreeBSD to Linux. Good luck. Their effort is better spent on developing their own stack.

    Besides, if they really wanted to, they could cut'n'paste from Linux. W2K is closed source, so you'd never be able to tell (and don't start screaming about tcp/ip fingerprinting, there's a million reasons why a system might have the same fingerprint).
  • I love overkill.
    I had heard a little bit about the new standard. I thought they were just going to add another 8 bit address to the end.
    I thought "Sure, and 20 years from now, when people are hooking up their lazy-boy chair and their furnance and their microwave to the internet, we'll run out of IP addresses."
    Now, I can't even imagine what it would take to use up all these IP addresses.
    By the way, hex is annoying. Fast for hardware, but annoying.

    Erik Z
  • As it would be much easier to trace abuse it would also be much easier to maintain abuse. It's much harder to attack someone when their address changes daily. I would think that only people that would/could use static ip's should have them, I wouldn't want my grandpa to have a static ip (his parinoia is bad enough anyway.)

    Net. Admin.
  • Posted by !ErrorBookmarkNotDefined:

    Must be because it's good news about MS.

    Half the so-called 'funny' comments get a 2.

    Computers are useless. They can only give answers.
  • The transition to IPv6 will not happen over-night. Backbone routers will continue to run IPv4 in parallel, and networks will slowly migrate.
  • I don't share your optimism on this subject. Just because we have trillions of addresses available on the net dosen't mean that our providers will magically give us IP's to use for our own personal use. Even if we go IPv6 I expect the cable providers and ISP's to continue their one IP address scheme. It would be *NICE* if we could all get subnets (even if we registered them on our own through ARIN), but the Internet's routing architecture dosen't work quite like DNS, where you are free to point it wherever. I'd love to ditch NAT in favor of real IP's, but I'm not so sure we can do that. This will make more IP's easily available for providers and other users that want/need more IP's, but not for the home user using a DHCP or otherwise dynamically assigned IP address to connect to the net.

    I also don't think we'll all suddenly see free static IP's, since this is just simply put a management nightmare, unless the routing system the Internet uses is given a major overhaul at the same time IPv6 goes into production.
  • If only there were some way to hash an easy-to-remember name to an IP

    Everyone should have DNS anyway. Hooray for -DPARANOID
  • M$ is relatively 'open' in their effort to make NT and W2K IPv6-ready. Checkout the website at [].
  • by Anonymous Coward
    Actually IPv6 are 128bits but instead of the normal 4 8-bit decimals people usually write them as eight 16-bit hexadecimals (rather than keeping the 8-bit decimal notation which would require 16 dotted decimals (one more than you listed actually ;P)). The 6Bone currently includes the first 24bits of the IPv4 address as well part (or all?) of the Mac address. This is just for testing though. IPv6 is classless and has certain bits reserved for stuff like multicasting too (reserved based on the high order bits). Unless this book is already out of date (a certain possibility) all the global unicast address ("normal" ip addresses you use on the internet) are the ones with the first three high order bits as 001. That'll leave us with enough ip addresses to last awhile ;) Also, IPv6 will have much better and faster routing.

    On a side note I think it's an interesting point that they should be ready to go with IPv6, there isn't a reason not to push ahead with it, except perhaps that our favorite dominant desktop OS is probably about the only OS that doesn't support IPv6 yet (well maybe Macs too, I dunno).

    Seriously though, I will be interested to see how the transition goes. We are talking about a lot of money to replace all those existing IPv4 only routers and stuff. Methinks one of the reasons stuff like ADSL aren't being throw into the widespread public very quickly is that the phone companies might be waiting for IPv6 (that and for the overall bandwidth of the internet to be able to support it, which IPv6 should help there too).
  • Decades? I doubt it. We'll run out of IPv4 addresses before that...
  • i would like to see large deployment as it contains not just bigger addresing space but guaranteed capacity and (what is the best) security enhancements and cryptography!

    what just can't be forgotten are IPv4-to-IPv6-migration-HOWTOs!

  • by Anonymous Coward
    Unlike IPv4 which was never meant to be used as widespread as it is today (and has required numerous hacks to make it to continue to work) IPv6 is redesigned very well. This isn't just about getting more address space. This is about sitting down, finding out all the flaws of IPv4 and fixing them, as well as designing it to be able to handle changes in the future. You're example of how to do it may be easy for humans to understand ip address but otherwise its not very efficient or practical. IPv6 uses the high order bits of the address to determine it's use (global unicast address, multicast address, local address, etc, etc). In the case of global unicast addresses the first three high order bits are 001 and then from there (real creative names) there is the top-level aggregation identifier (13 bits), next-level aggregation identifiern (32 bits), the site-level aggregation identifier(16 bits) and the interface ID (Mac address?) (64 bits). The 6bone is layed out a little differently and is really IPv6 wrapped in IPv4 (so they can test over the internet even though few of the routers understand IPv6 yet) though I am not sure of all the details other than the fact it contains part of the IPv4 address and the Mac address right now. The IPv6 packet is also redesigned. Overall IPv6 basically takes care of all the problems that exist in IPv4, including but certainly not limited to address space. Personally I can't wait for it to be used widespread. Someone go kick those morons over in redmond and tell them to hurry up, the rest of the world is waiting. Or better yet let's jump to IPv6 now and screw all the windows users... *grin*
  • % /usr/games/number 340282366920938463463374607431768211456
    three hundred forty undecillion.
    two hundred eighty-two decillion.
    three hundred sixty-six nonillion.
    nine hundred twenty octillion.
    nine hundred thirty-eight septillion.
    four hundred sixty-three sextillion.
    four hundred sixty-three quintillion.
    three hundred seventy-four quadrillion.
    six hundred seven trillion.
    four hundred thirty-one billion.
    seven hundred sixty-eight million.
    two hundred eleven thousand.
    four hundred fifty-six.

    It sounds like a child's song. :)
  • by Anonymous Coward
    This is NOT an IETF sponsored effort. The
    IETF does not encourage closed groups with
    member-only web sites. The IPv6 Forum seems
    like it's being started by the same fanatics
    that were in the ATM Forum when ATM was at
    it's deathbed.

    The IETF believes in open specifications and
    open web sites unlike these "Forums" that
    seem more interested in making a quick buck.

    Ignore such doomed efforts, and support the
    real IETF IPv6 task force!
  • Maybe the 6bone is working ok, but a bunch of the proposed standards already seem kind of dated, and some parts of the proposal seem frankly dumb. Tying IPv6 addresses to link-layer addresses (and using half the 128 bits for it!) is dumb - it makes it very difficult to switch hardware around without having to wait for DNS changes to propagate, and there are plenty of other reasons to object to it too.

    IMHO, it's not necessary to use the MAC-adress. Just as long as the machines can find each other through Neighbourhoud Discovery, everything's fine. In some cases it's not even possible to use the Interface Identifier because it doesn't exist: PPP-connections, ...

    RFC 2373 "IP Version 6 Addressing Architecture" cleary states:

    An unicast address or a set of unicast addresses
    may be assigned to multiple physical interfaces
    if the implementation treats the multiple
    physical interfaces as one interface when
    presenting it to the internet layer.

    It's quite annoying when you can't get on the net because the DHCP-server crashed.
  • Your grandfather runs an OS that comes with a telnet daemon? Don't answer that - you'll probably lie anyway. The point is that the other poster's grandfather more than likely doesn't. He represents the "average joe" who, right now, uses Windows. Most of these people don't even know what an IP address is, so why would an ISP want the added hassle of keeping track of which one belongs to what customer? Regardless of how many Addresses we have available, dynamic addressing is still more viable than static for most providers.
  • by jonathanclark ( 29656 ) on Sunday July 04, 1999 @08:35AM (#1819193) Homepage
    This is not what is going to happen, but with 128 bit addresses you could just pick a number at random and use it. The odds of 2 people chosing the same number is low enough to be acceptable.

    The actual probability is approx:

    1- exp(-n*n/m)

    where n is the number of addresses selected
    and m is the total addresses available.

    If 4 billion people select addresses at random then the odds of a collision are:
    1 in 415,828,534,307,635,078.

    To put this number in perspective, your odds of guessing the right number to a 56bit DES encryption key on your first try is 1 is much better:
    1 in 72,057,594,037,927,936

    We won't be running out of IP addresses anytime soon. 2^128 is not big enough that we could assign each atom on the planet is own IP address (this number is ~2^170), but we certainly could assign each atom that could possibly be seen it's own IP (i.e. by excluding those in the earth's core). Considering that you need a fair number of atoms to store just to store a 128bit number, I think we are safe until space travel explodes.
  • Well, half of the point of IPv6 (and it's expanded address range) is to SIMPLIFY routing. If you would need a routing table carrying all IPv6 addresses in the world, you'd have big trouble. (The other half of the advantages is autosetup, of course... The last 64 bits are usually identical to your Ethernet MAC address.)

    Choosing your own random IPv4 address today won't work. Think routing...

    /* Steinar */
  • There is very little motivation amongst the
    major ISPs to migrate to IPv6. The main reason
    for IPv6 - the potential address shortage -
    has been blown out of proportions. More efficient
    address space allocation, as well as extensive
    use of private addresses and NAT make
    'the day when we all run out of address' further
    and further off. There are no other solid reasons
    to move to IPv6. The only other reason I can
    think of - very elegant scheme of assigning
    addresses to the host - is no longer relevant
    due to the wide speard of DHCP.

    Granted, the large address space is indead much
    better than the combination of private address and
    NAT, as well as the IPV6 scenario for allocating
    addresses to the host is better than DHCP -
    but not at the extent to justify the massive
    conversion of INTERNET backbone to V6.

    What I've heard from someone who's been
    participating in IPv6 committee (can't give his
    name but there is a hint: he wrote several key
    RFCs 'on how to write RFCs' and also teaches at
    Harvard) that the whole IPv6 was nothing but
    the big P.R.: IP was still evolving, its enemies
    has been trying to speard the FUD, and 'runing
    out of addresses' scare was part of that FUD.
    Consequently, IPv6 specs was mostly response to
    that FUD. Of course now IETF feels kind of
    funny: they've spent so much time and efforts
    and seems like noone cares.

  • Are U sure we're gonna use IP in the outter space ? :o) It's time to switch to IPX .

"I have not the slightest confidence in 'spiritual manifestations.'" -- Robert G. Ingersoll