Catch up on stories from the past week (and beyond) at the Slashdot story archive


Forgot your password?
The Internet

Asia Running Out Of IP Addresses 732

miladus writes "According to a story at Zdnet, Asian countries are running out of IP addresses. China, for example, was assigned 22 million IP addresses (for a population of 1.3 billion) under IPv4. The US owns 70 percent of current IP addresses. Perhaps IPv6 will solve the problem."
This discussion has been archived. No new comments can be posted.

Asia Running Out Of IP Addresses

Comments Filter:
  • 2 solutions (Score:0, Interesting)

    by IamTheRealMike ( 537420 ) on Wednesday May 28, 2003 @04:07PM (#6060382)
    1) Deploy IPv6

    2) Actually allocate the addresses in a way that has some semblance of fairness to it.

    Of the two, I'm not sure which is easier. Sad really, isn't it?

  • IPv6? (Score:1, Interesting)

    by A Guy From Ottawa ( 599281 ) on Wednesday May 28, 2003 @04:09PM (#6060397)
    The US owns 70 percent of current IP addresses. Perhaps IPv6 will solve the problem.

    Or perhaps the US could solve the problem by not being so damn greedy?

    IP Everywhere... not just the US!

  • Is this... (Score:1, Interesting)

    by SharpFang ( 651121 ) on Wednesday May 28, 2003 @04:09PM (#6060403) Homepage Journal
    Is this a part of the economical battle between Asia and US? Give them less IPs so they can't compete on the market on equal conditions? Usually I'm against theories of conspiracy but in this case I'm willing to make an exception...
  • by Mistah Blue ( 519779 ) on Wednesday May 28, 2003 @04:09PM (#6060407)

    I wonder why they don't use the non-routable address spaces and NAT.

    Let's also remember (since I detected some trolling in the article) that Asia was a backwater for the Internet 20 years ago when address blocks started to be doled out, so naturally the U.S. and to a lesser extent Europe got the bulk of the blocks.

  • by krisp ( 59093 ) * on Wednesday May 28, 2003 @04:09PM (#6060411) Homepage
    Let the other billion or so people NAT the remaining ip addresses! 10.x.x.x adds another 16M, and they can 192.168.x.x behind those :)
  • by illumin8 ( 148082 ) on Wednesday May 28, 2003 @04:11PM (#6060446) Journal
    I work for one of the largest Unix vendors out there (hint, we used to put the . in .bomb).

    Anyway, I can tell you that in one of my many Unix classes when we were learning how to configure IPv6 the instructor mentioned that the reason why IPv6 had been added by default to our new versions of Unix was that we were getting a tremendous amount of pressure from our customers overseas, primarily in Asian markets, who were unable to get IPv4 address blocks from their ISPs, and were therefore deploying IPv6 exclusively.

    I believe currently a lot of Asia is running IPv6 with IPv4 gateways at main NAPs.

    -obdisclaimer, the opinions expressed are not those of my employer.
  • by agentZ ( 210674 ) on Wednesday May 28, 2003 @04:15PM (#6060497)
    And ditto for some class A networks. I know that MIT does a LOT of computer research, but do they really need an entire class A? Did you know that each fraternity at MIT has their own class B? Really! For an example, try looking the hostnames for the routers in some of the frat houses.

    $ host 18.[231-238].0.1

  • by wronskyMan ( 676763 ) on Wednesday May 28, 2003 @04:19PM (#6060541)
    Unless the MPAA (or its Asian equivalent) bribes all the governments to ban NAT boxen (so they can be hotbeds of technological innovation, like the People's Republic of Illinois :-) )
  • good, less spam (Score:2, Interesting)

    by kevincw01 ( 676545 ) on Wednesday May 28, 2003 @04:21PM (#6060583) Homepage
    And when they're forced to move to ipv6, ill be sure to take note of what they get so I can REJECT that in my /etc/mail/access.db too.
  • NAT China (Score:4, Interesting)

    by Nom du Keyboard ( 633989 ) on Wednesday May 28, 2003 @04:26PM (#6060631)
    China, for example, was assigned 22 million IP addresses (for a population of 1.3 billion)

    Given that China has already firewalled the whole country, why don't they just NAT the whole country as well. Then, with a little cleverness, they can have the whole address space available to them alone.

  • by patniemeyer ( 444913 ) * <> on Wednesday May 28, 2003 @04:27PM (#6060645) Homepage
    I did the following fun calculations once for a book I was working on (let me know if they're wrong):

    There are about six billion people on earth and each person's body consists of about 100 trillion cells. With 128 bit addressing each individual cell in every human being could have 100 trillion addresses. I believe that is on par with 1 address per molecule.

    To put it another way we cannot, with current technology, use all of these addresses in any physical way. We can't even count them (literally). Suppose you have a machine that can do a trillion operations per second; then suppose that you have a billion such machines connected via the Internet and we ask each one to simply start counting through part of the address space. I believe it will take about 3 billion years for them to finish.

    Pat Niemeyer
    Author of Learning Java, O'Reilly & Associates and the BeanShell Java scripting language.
  • Re:Is this... (Score:5, Interesting)

    by HiThere ( 15173 ) * <{ten.knilhtrae} {ta} {nsxihselrahc}> on Wednesday May 28, 2003 @04:28PM (#6060649)
    No. It's that when they first started handing out TCP addresses it didn't ever seem possible that everyone would want, not just one, but several. So they handed them out in big blocks to make administration easier.

    The people who were in at the start all ended up with huge domains that they didn't expect to fill, but then they didn't expect that the address range would "ever" fill up. So why be picky.

    Countries weren't really thought of during the first round of allocations. Or even companies. Or most government departments. Except for a few who were a part of the process. The second round, all those were assigned "fair" chunks. But they didn't think of ISPs, or such. That was the third round, which added in ISPs and a few involved techie users (who now wanted an address at home that didn't depend on where they worked).

    I don't know which round of assignments we are now. Must be around the sixth or seventh. (A round comes to an end when people figure out that they are running out of addresses, so they revamp the rules of how they are allocated.) Somewhere in there DHCP and bootp started being used so that people didn't get "permanent" addresses anymore.
  • by moosesocks ( 264553 ) on Wednesday May 28, 2003 @04:33PM (#6060706) Homepage
    I'm not sure if IPv6 will solve anything. It seems to me that the whole concept of "IP addressing" seems quite archaic. The international telephone system doesn't need to 'dish out' phone numbers between countries - each country has its own country code, and everything else is handled wihin the country.

    Hell. The whole concept of the 'internet' by means of Tcp/IP is becoming quite dated. Why can't we combine the domain naming system with the IP system. What I would propose is to give each computer on a given domain an alphanumeric name (can contain any type of characters, and is decided by the owner of the domain - basically the same as todays concept of a 'hostname'. The domains, in turn, are managed by an independent organization in each country, followed by a country code. For example, a sample address would be
    Joe Smith@Earthlink@USA (users within the USA can leave the @USA blank)

    this eliminates the need for a domain naming system. takes a lot of power away from ICANN, would help to solve cybersquatting, and provides an infinite number of computer addresses (at no point should the 'name' need to be translated into a numeric address.

    Computers behind a router should be able to have their own address as well (multiple servers on one address without the mess of port forwarding! With many home users now running their own web/music servers, this could be a godsend. For example:

    Anybody should be able to get their domain, but those who do not have their own should simply share one with their ISP.

    Unix geeks will probably balk at my radical ideas. but it needs to be done. the numbered IP system was concieved when the only computers on the 'net were run by the people who wrote the protocols,. Nowindays, computers are used by everybody (and their grandmothers!). and it made sense too, as bandwidth was very limited, and the programmers never intended for so many computers to be on the net, and cut corners to gain a small speed advantage (a few bits per packet - which was a lot back then. now, it's nothing). IPv6 simply continued to use (longer) archaic addresses - the problem still exists; we need another layer for domain names, and it's confusing as hell to non-geeky types)

    I know my ideas seem radical, and will probably never be accepted... but I certainly would hope that we fix some of this. IPv6 isn't a solution - it's avoiding the problem.

    (yes, this was somewhat inspired by Apple's rendevous, which addresses many of my concerns, but is by no means acceptable for a worldwide scale. On a side note, I believe that in order for rendevous to succeed, Apple needs to open it up, and allow M$ and Linux to interoperate with it.)
  • by swb ( 14022 ) on Wednesday May 28, 2003 @04:34PM (#6060710)
    I know for a fact this is true.

    One company that I've worked with uses a routable /16 (same size as a class B) externally and a routable /16 internally and NATs between the two of them.

    What's super annoying is that we have some permanent connectivity to them and they give out different IPs depending on the source of the DNS query. We're not fully integrated with them, so it makes for loads of fun trying to do resolution correctly.

    I think it's a waste of addresses. Give back the public-facing /16 they use, use the private /16 globally and use some other NAT scheme or other RFC1918 addressing internally where needed.

    But even RFC1918 is a problem as everybody thinks that is "theirs" and then you do NAT-NAT, which breaks most troubleshooting tools.
  • Re:2 solutions (Score:2, Interesting)

    by 10Ghz ( 453478 ) on Wednesday May 28, 2003 @04:35PM (#6060731)
    But if you do it based on number of systems potentially needing an IP then the US will still be high up on the list... probably #1. Certainly not 70% of the IPs, but far more than the population would otherwise indicate.

    How many people there are in the USA? Well, I'll make an generous guess and say 300 million. That's about 5% of the global population. Let's be generous and give USA 7 times as many IP-addresses as there are people in the US. That would mean that US would have 5% x 7 = 35% of the available IP-addresses. Seems fair to me. So why do they need 70% of the addresses? It seems that Europe has ALOT less IP-adds than USA does (15%? are the shares of different regions available anywhere?), even though there are more people here and were are equal in technology (in fact, more advanced in mobile tech)
  • IPv6 + NATPT (Score:5, Interesting)

    by nsayer ( 86181 ) <nsayer @ k f u . c om> on Wednesday May 28, 2003 @04:36PM (#6060738) Homepage
    The migration path, in general, is to use DNS proxies and NATPT to make the transition appear to IPv6 users to be instantaneous.

    I did this a while ago at my house. My network actually had no IPv4 on it at all for a few weeks. I stopped because a couple of applications didn't support IPv6 and because the KAME NATPT I grafted into my FreeBSD source tree broke. I did it sort of as a proof of concept, and it succeeded sufficiently for me to propose that IPv6-only ISPs could easily use the technique.

    You first set up a DNS proxy. totd (Trick or Treat Daemon) is a good one. Its job is to turn requests for AAAA records into requests for AAAA or A records, and to translate A record replies into AAAA records with a special prefix tacked on to the high bits. This will make it look as though the whole IPv4 Internet is hidden inside of a special /96 prefix.

    Coincidently, you route that /96 prefix into a NATPT. IPv6 packets go in, IPv4 packets come out and are sent to the IPv4 Internet as if they had gone through a NAT.

    Having done this, all of the ISPs customers would see a complete IPv6-only Inernet, but they could still interact with legacy (IPv4) sites as if they were IPv6. As more and more ISPs convert over, the IPv4 network will simply shrink slowly until it's gone, but in the meantime remain as accessable as it currently is.

    With such a transition plan in place, the more people who move to IPv6, the emptier the IPv4 Internet experience becomes (however, folks trapped with IPv4 only providers could use techniques like 6to4 to escape the legacy network), which in turn becomes the driving force for transition.

    So, Enough stories are turning up... When is /. going to support IPv6?

  • by nurb432 ( 527695 ) on Wednesday May 28, 2003 @04:40PM (#6060774) Homepage Journal
    I dont see IP6 happening anytime soon, perhaps if they enforce NAT connections for everyone they can extend the lifetime a bit...

    True it sucks to be stuck behind firewalls but its better then nothing..
  • by PD ( 9577 ) * <> on Wednesday May 28, 2003 @04:43PM (#6060816) Homepage Journal
    So you're saying that it is incorrect of me to number my home network that uses NAT? I use,, and so on for my network. I picked it just because was a little harder to type. Also, from the RFC: "If a suitable subnetting scheme can be designed and is supported by the equipment concerned, it is advisable to use the 24-bit block."
  • by Anonymous Coward on Wednesday May 28, 2003 @04:44PM (#6060830)
    On the other hand blindingly portscanning IP ranges is infeasible. Can you imagine scanning a /64? That's like 18,446,744,073,709,551,616 IP. If you could scan at a rate of 1 Million IP per second it would take over 584,942 years. And with the minimum packet size of 576 bytes it would take a 9,2 Gbps of bandwidth just to ping 1 Million IPs in a second.
  • by xchino ( 591175 ) on Wednesday May 28, 2003 @04:49PM (#6060877)
    No ISP worth their salt would. I work for an ISP, and I can confirm that it does cost us money to give you an IP, so it's going to cost you money too. When IPv6 is implemented it won't cost us anything, so it won't cost you anything. I've seen both our cost and our customers cost for IP addresses/ranges so much that it amazes me IPv6 isn't being implemented by every ISP already. It's just the chicken and the egg problem.
  • by spaceyhackerlady ( 462530 ) on Wednesday May 28, 2003 @04:55PM (#6060935)
    There are about six billion people on earth and each person's body consists of about 100 trillion cells. With 128 bit addressing each individual cell in every human being could have 100 trillion addresses. I believe that is on par with 1 address per molecule.

    A necessary number: number of IPV6 addresses is 2**128 = 3.4E38.

    Hmmm...lessee now, 6E9 people, 1E14 cells per person, that makes 6E23 cells. That's about 5E14 IPV6 addresses (five hundred trillion) per cell.

    Per molecule? Let's assume an average person's mass is 60 kg, and that the average molecular weight of the human body is 25 (we are mostly water). That makes (60 * 1000) / 25 * 6.02E23 = 1.4E27 molecules per person. Total Earth population is then 6E9 * 1.4E27 = 8.4E36 molecules. Actually about 40 addresses per molecule.

    My other favourite number is how many IPV6 addresses each square micron of the Earth's surface could have:

    Earth's surface area in square microns = 4 pi (6378 * 1000 * 1000000) ** 2 = 5.1E26

    3.4E38 / 5.1E26 = 6.6E11

    A big number!


  • by hesiod ( 111176 ) on Wednesday May 28, 2003 @04:59PM (#6060975)
    > How soon, who knows...but saying that it won't happen is like saying no one will ever need more then 640k of memory..

    Considering the scale of this issue, it seems more like a homo erectus saying "No one need fire. Too hot and not portable, like Linux." Well, except for the Linux thing.

    But seriously, I think the planet itself would be long gone before that many IP addresses was even close to being used. Until, of course, nanobots start self-replicating and join the Internet Continuum & start taking IPs (those dirty bastards).
  • Re:IPv6? (Score:4, Interesting)

    by York the Mysterious ( 556824 ) on Wednesday May 28, 2003 @05:01PM (#6060995) Homepage
    My old school district had a neat NAT setup. Every server in the district had the same NATed IP, but if you made a request for the DNS address of a server on a specific allowed port it wold forward it to the internal IP. Very smart NAT. It also makes a lot of port scanners that require IPs worthless. request on port 80 go to request on 25 goes to request on port 80 goes to
    It was probably loads of fun to manually set this up, but it works
  • by SuperKendall ( 25149 ) on Wednesday May 28, 2003 @05:14PM (#6061120)
    considering the size of our galaxy.
  • Re:Are you sure? (Score:4, Interesting)

    by swillden ( 191260 ) * <> on Wednesday May 28, 2003 @05:22PM (#6061199) Homepage Journal

    Hum... till ppl start assigning one ip address per fridge, cooker, toaster, air cond., etc...

    Umm... no.

    2^128 addresses means that every one of the 7 billion people on the planet can have 48,611,766,702,991,209,066,196,372,490 addresses of their very own.

    That's a lot of appliances.

  • by Skapare ( 16644 ) on Wednesday May 28, 2003 @05:28PM (#6061256) Homepage

    IPv6 is fundamentally broken. The routing system for it does not scale to the same level the address space does. There are enough addresses for everyone to have their own portable /64 assignment (if not larger), but IPv6 can't handle the routing. The routing technology was not improved to scale up, even though it could have been done (although I don't know if it can be done with the way IPv6 was designed). But that's not a valid excuse for not having scalable routing as the IP layer structure could have been designed to allow for it. Wedging another layer in below IP for IPv6 might also work, but I think we would be better off waiting for a clean re-design, perhaps to be called IPv7 (and pushing them to hurry up with it).

    If you don't believe me, just post a call for portable address assignments in IPv6 for everyone. You're get plenty of responses saying that the routing can't handle it. And that is the problem.

  • Re:As I see it... (Score:3, Interesting)

    by amorsen ( 7485 ) <> on Wednesday May 28, 2003 @05:53PM (#6061440)
    Why don't you switch? What is stopping you?

    And before you turn that one back on me: I am already dual-stack with NAT'd IPv4 and real IPv6 addresses for the hosts. So I am not holding things back. I love autoconfiguration by the way. No configuration on the hosts at all. IPv6 is so simple and easy compared to IPv4.

  • by Skapare ( 16644 ) on Wednesday May 28, 2003 @06:11PM (#6061589) Homepage

    The legacy class A assignments just became /8 assignments. Not all of them have been chopped up (other than inside the companies with those assignments). Maybe those companies should be the first to go with IPv6.

  • by Fzz ( 153115 ) on Wednesday May 28, 2003 @06:15PM (#6061630)
    Actually it went international long before IP was even in use. University College London joined the ARPAnet back in 1973 []. TCP and IP were only standardized in 1978.
  • by bob ( 73 ) on Wednesday May 28, 2003 @06:21PM (#6061694) Homepage
    At my suggestion, a few years ago my employer tried to give back a class B because we didn't really need it, asking only for a handful of class C numbers in return. Turned out to be harder than you might think, and it never happened. Now, since we never got the class C nets either, parts of the class B are in use and it would be a huge PITA to rip it out, so most of it's pretty much lost address space. So don't put all the blame on the holders of those nets -- a lot of the problem stems from mis-managment of the resource.
  • by Anonymous Coward on Wednesday May 28, 2003 @06:33PM (#6061812)
    They are definitely MIT fraternities. The ones off campus have a T1-equivalent DSL line that's paid for by MIT, complete with its own class-C address space. I don't believe each fraternity has the ability to use all of the addresses ( were only allowed), but the remainder are earmarked for use by the fraternity. It's inefficient, but MIT owns the whole class-A space and there's no need to get more IP addresses. Any odd allocations of IP addresses is out of politics or logistical convenience, not technical requirements. When the ISP I was working for found out about the number of IP addresses that we had available at our disposal at the fraternity, they nearly shat a brick. They were jealous, seeing that IP addresses from upstream providers were very valuable ...
  • strict requirements (Score:2, Interesting)

    by Tancred ( 3904 ) on Wednesday May 28, 2003 @06:43PM (#6061878)
    If any of you have been finding it a bit difficult to get space from ARIN, consider yourselves lucky. RIPE's tougher to get space from. And APNIC and JPNIC sometimes seem to guard their remaining IP space insanely. I hear about registries asking for receipts for all the supposed servers you've got installed as well as why you need so many. Sometimes they'll tour the facility to be sure you really need that large a netblock. This one's through the grapevine, but it's a usually reliable grapevine - one company was turned down and told to buy a certain type of hardware that would handle higher loads with fewer IP addresses.

    It's a tough question - should early adopters with huge blocks of space be required to renumber (a very painful process) even though they're generally the ones that got the whole thing moving in the first place? Those early class A blocks don't even require annual payments. Go to an ARIN meeting sometime to hear some lively debate on that subject.

    IPv6 is going to have to be pushed outside the U.S. None of the big backbones are rolling out anything substantial anytime soon. Nobody in the U.S. is feeling enough pain from IPv4 to need to do it. Besides, we'd really need some v6-optimized hardware too to get it going natively everywhere.
  • by mnmn ( 145599 ) on Wednesday May 28, 2003 @07:14PM (#6062147) Homepage

    Some time ago, PakNet was the biggest ISP in Pakistan serving hundereds of thousands under ONE ip address... interestingly using Linux kernel version 1.3.x. I also remember every user had a shell account from which we could cat the /etc/passwd, which was not surprisingly humungous. For a while, BrainNet and PakNet were the only ISPs in Pakistan, and later ISPs could only connect to Paknet, and their single monolithic IP address. I remembed always being banned from IRC servers which were blocking users by their IP addresses. Talk about one huge NAT and this is the biggest Muslim country in the world.

    And on this side, here in Toronto, Bell assigns a subnet of 8 IPs to every customer, including ones who need just one. 3 of those IPs are gateway, broadcast and 00 host, which leaves 5 IPs. two of them are assigned to the on-site router and off-site routers which are connected via DSL. Its one of the best examples of IP address waste, while the Chinese crave a personal, their very own IP address!

    Theoretically all of the more than 4 billon IP addresses can be used, and it is VERY unlikely that the whole worlds population would be online. But the imbalance remains with the US holdin on to all the Oil and IP addresses. At least we can do something about one of them.
  • by frovingslosh ( 582462 ) on Wednesday May 28, 2003 @08:13PM (#6062734)
    IPv6 will not run out of addresses - it will use 128-bit address space.

    By my limited understanding of IPv6, this statement is rather false and misleading. Is the address space 128 bits? Yes, somewhat. But does that give a good account of the number of addresses available, NO. IPv6 has several different types of addresses, and the total number of actual addresses is far smaller that 2^128 would indicate.

  • by GlassHeart ( 579618 ) on Wednesday May 28, 2003 @09:17PM (#6063227) Journal
    Given that there are probably no more than, say, 10m PCs in all of China, please tell me why they need more than 2x as many IPs.

    Try not to say "nonsense" and "inescapable logic" right before you start guessing.

    This article [] states that PC sales exceeded 10.1 million units in 2002 alone. Assuming that people keep their PCs for 3 years (which is not unreasonable for a poorer country where a PC is a major investment), we should be talking about a population of over 20 million PCs. Even that conservative estimate is already twice your guess. In fact, if you believe this article [], China overtook Japan as the second biggest PC market in the world last year.

    Prove it. I think you mis-googled.

    The CIA World Factbook China page [], under "Communications", says "Internet Users: 45.8 million (2002)".

    In any event though, even if they have 50m internet users, it doesn't mean there is a problem.

    The trouble with Slashdot, and in particular with folks of "inescapable logic", is that you don't actually read. Where did I ever say there was a problem? I was answering somebody's question as to how many people in China can read or write, or have ever seen a computer, relative to the US. Later, I was correcting your apparent mental block with the low percentages of users from China.

  • by stesch ( 12896 ) on Wednesday May 28, 2003 @09:56PM (#6063506) Homepage
    A necessary number: number of IPV6 addresses is 2**128 = 3.4E38.

    You don't have 2^32 addresses with IPv4. And you don't have 2^128 with IPv6.

  • by GlassHeart ( 579618 ) on Thursday May 29, 2003 @12:59PM (#6068606) Journal
    That stat came from China's state-run news agency which has been documented to inflate figures hundreds of times and even out right lie. Please get another source.

    Sure. This [] quotes the IDC as expecting "China's PC sales to nearly double in a few years, from 11.3 million in 2002 to 21.1 million in 2006." Note that IDC's estimates are even higher than Xinhua's.

    Furthermore, do not confuse current market share of NEW computers with the installed base of PCs as a whole.

    Who's showing signs of confusion? I estimated conservatively (assuming people keep computers for 3 years), that there are 20 million PCs in use in China, based on sales figures in 2002. I further quoted that China now has the second largest PC market, which is not the same as installed base.

    It is quite possible for China to have much millions more NEW sales than Japan because of their economic growth and still have fewer installed computers at the end of the year or even 5 years.

    That's actually less likely. Poor countries are likely to hang on to PCs longer than rich countries. I wouldn't be surprised if there were a good number of 5-year old computers in use in China.

    When, at last count, less than 1% of households have PCs and few people are likely to be able to afford or use multiple computers; it's basic math and a tiny amount of extropolation.

    Your 1% figure is simply inaccurate. The 10.1 or 11.3 million PCs sold in 2002 already account for the 1%, and that's assuming nobody in this third world country throw away their computer after one year.

    However, your meaning came across quite clearly on my end because of your insistance that the apparent disparity needs to be justified somehow.

    Try to understand that some people don't give a damn one way or the other, except that people are arguing the right topics (in this case, actual users versus percentage of population), and are using the right numbers to back up their arguments.

  • by mfrank ( 649656 ) on Thursday May 29, 2003 @02:48PM (#6069493)
    Man, even with a Dyson sphere IPv6 is overkill. You still get over 10^15 (a million billion) IP addresses per square meter.

    Now, if you put a Dyson sphere around every sun in the galaxy, then you'll get down to a few thousand IP addresses per square meter. *Then* we need to think about going to IPv8.

"If you lived today as if it were your last, you'd buy up a box of rockets and fire them all off, wouldn't you?" -- Garrison Keillor