IPv4 Address Crunch In 2 Years, IPv6 Not Ready

An anonymous reader writes "We've known for ages that IPv4 was going to run out of addresses — now, it's happening. IPv6 was going to save us — it isn't. The upcoming crisis will hit, perhaps as soon as 2010, but nobody can agree on what to do. The three options are all pretty scary. This article covers the background, and links to a presentation by Randy Bush (PDF) that shows the reality of the problem in stark detail."

Re:FUD

[RobGeek]

Totally Agree! WOLF WOLF WOLF WOLF and now we are supposed to believe that there is really a "shortage" that we are to worry about? Oh heavens! The whole Internet will collapse!

Re:Well duh

[PrescriptionWarning]

That sounds like an "if it ain't broke, don't fix it" argument to me. Which in fine and good for simpler technologies, but can be disastrous for more modern technologies. Just think what would happen if you didn't change your car's oil until the car simply refused to run. What would happen if we all decided not to curb our oil consumption habits until we either ran completely out of oil reserves. You see its the shortsightedness that in the long run costs you WAY MORE than if you simply keep the options in mind and work towards a solution.

So in two years when they can't add any more addresses, the only ones to blame will be those who stuck they feet in the mud and wouldn't budge. Besides, they can always just start taking away all those spam sites that offer no real content and just distribute those to other who actually need them, I'm sure there's at least another 2 years worth of those.

Re:Dupe

[IBBoard]

And we need to retrieve some from the Vatican as well!

Looking at the information here [modernlife...bish.co.uk] then the Vatican has far too many IPs per capita. Ditto for the other tiny nations of Gibralta and Monaco. I'm sure it'll buy us at least a week!

And for anyone geeky enough to care (who isn't geeky enough to have it bookmarked already) here [iana.org] is the assignment list. Each of the companies mentioned owns an entire top level block (e.g. Ford own 19.xxx.xxx.xxx) and some like the Defense Information Systems Agency (whoever they are) own multiple blocks! That's an awful lot of addresses.

Hardware compatibility and updating.

[rfelsburg]

The shift will also depend on hardware vendors making sure that their hardware is completely ipv6 compatible. Even with quite a few vendors saying that their stuff is compatible, I know of a quite a few major bugs still lurking with those same vendors. Not many large companies are going to switch to IPv6 until they need to upgrade hardware, if their existing hardware is only IPv4 compatible.

Re:Dupe

[Anonymous Coward]

And as I said before, the solution is to take back some of those huge class A blocks from companies like HP, Ford and GE, which are not using all the space. That would buy a few years.

I feel a pang of guilt here since we have a full class C block at my business. We use about a dozen static IP's (mail and web servers) but we've had the full class C block for 10 years or more from our original contract with uuNet. I think over the years they just forgot we had the block of addresses, and with the change from uuNet to Worldcom/MCI and now Verizon it's a wonder they can keep anything straight.

Time for the Government(s)?

[grumbel]

One thing is rather clear to me: We won't run out of IPv4 addresses anytime soon, instead the price will increase more and more and thus people will end up behind ISP enforced NATs, because IPs are to expensive for the average consumer. This is after all already the case, at least in part, static IPs are a premium service, not something you get for free from most ISPs.

So how to fix this? How about some good old government regulation? If you want to provide a "Internet service", you have to provide IPv6 or you can't call it "Internet". With a little force it shouldn't take all that long till the switch to IPv6 is done. But unless that happens the rarity of IPv4 addresses will simply be seen as a nice way to make money, instead of a problem that needs to be fixed.

Re:Dupe

[kellyb9]

2^24= 16,777,216 addresses for each of those companies seems excessive. If there was a major crisis, I would wager to bet they would begin leasing out these addresses to private consumers at a premium. Regardless, I've heard so many estimates about when this is going to happen, I find it difficult to believe any of them.

And?

[SmallFurryCreature]

That is one way to do it, keep patching it up and hope it becomes somebodies elses problem.

The problem is simple, the way we want to use the internet means we are getting more and more devices which desire their own internet adress. Some people suggest solutions like NAT but these only have so many uses especially when mobile phones become internet capable. If you want your internet node to be independent then you need an ip adress.

Don't believe me? Fine, give up your internet connection with its own IP and use the NAT solution of your ISP. Good luck running a torrent.

We could easily solve the entire problem if we just used NAT for every major ISP. It would free up countless adresses and keep IP4 usuable for decades rather then years.

So who is first? Who is going to give up their IP for their home for the greater good?

Thought as much, absolutly nobody.

It is the problem with humans, we don't want new power installations, we don't want to use less power and we refuse to switch to more economical appliances. Something has to give, but goverment or business is NOT going to do it. Sooner or later it just breaks down (see the LA brownouts) and finally a decission will have to be made.

Same with a solution to IP4 limited adress space. We will keep coming up with patches and ignore the problem until finally it can no longer be ignored and then we will have to really bite down to implement it at great cost and inconvenience when we could have solved it easily right now.

Because lets be honest, it ain't all that much of a problem. In the EU we switched currencies. A hell of a job but because it became accepted that it had to be done, it just happened.

We could easily do a switch to IP6 but only when the majority just accepts that it has to be done, and bites the bullet.

Analog mobile phones no longer work in the US, holland no longer airs analog tv signals, switches happen all the time. It is nothing special, but in each case somebody just had to say "we are switching and if you are not ready, though".

So what if countless devices will no longer work, at a given point you just have to be able to say "upgrade or be left behind" or you will be forced to increasinly bend over backwards to accomadate out of date tech.

Not compatible, not happening

[fuzzy12345]

DJB said it best at http://cr.yp.to/djbdns/ipv6mess.html [cr.yp.to] Why switch from an Internet with a billion people on it to one that has nobody on it that can't be reached by IPv4?

Why should most people (including 'nerds') care?

[apathy maybe]

OK, I'm interested in technology, I know what IPv4 and IPv6 are, I know that there are many more advantages to IPv6 then to IPv4 etc. Yet I'm failing to see why I should care whether IPv4 addresses are running out or not.

But more to the point, what can I (as an individual who isn't part of the technocratic elite) do about it if I did care?

I don't code network stacks, nor kernel drivers, most of my software is written by someone else, and is automatically updated to fix problems and include new features.

I assume that by the time everyone else is using IPv6 I shall be too (simply by virtue of my software being updated).

So, why should I care? And what should I do if I did care?

Re:Dupe

[Spad]

This [xkcd.com] is a much prettier depiction

Re:Dupe

[Anonymous Coward]

heck, I used to know a private company with ~100 employees that had gotten a "Class B" because the owner had known someone on the allocation board "way back when".

The company doesn't exist anymore, but it wouldn't surprise me to know the ex-owner was still hanging on to the Class B (he had mentioned that he'd been offered lots of money for it quite a few times, but preferred to have it for the prestige value).

Class 'C' address space for sale.

[arthurpaliden]

The company died and no longer needs it. Maybe I will put it up on ebay.

Re:Well duh

[A beautiful mind]

While I appreciate the point you're trying to make, but there are quantitative differences between the thinking of a country like Japan and for example the USA. In Japan, they did have the foresight to make their systems IPv6 ready, so maybe just our expectations are too low? I'd rather tell people what to do than to make excuses in the technology/politics field referring to Joe Sixpack who allegedly wouldn't understand or care.

Off topic

[oyenstikker]

That you mentioned India might come up with a solution reminds me of a book I read that discusses in the context of game theory (primarily Prisoner's Dilemma) why people (Indians in particular) make poor decisions as far as society is concerned to maximize personal returns.

"Games Indians Play" by V. Raghunathan
ISBN: 9780670999408

Re:Dupe

[gclef]

I get that info from here [potaroo.net] which is looking at the actual allocation rates from the RIRs.

Progress from the top 100 sites - none!

[GReaper]

One useful site I tend to look at on a regular basis is Lars Eggert's IPv6 Deployment Trends [nokia.net], it uses the Alexa rankings to find the top 100 sites for various countries. You could always argue that these aren't the most visited sites - but it does give you an idea.

The top 100 sites for all these countries comes to a big fat total of 0%. I'm not expecting fast adoption, but it would've been nice to see some progress being made with these sites. Even the two sites which I regularly visit that report about IPv6 stories (Slashdot and Ars Technica) don't even have IPv6 records!

I suppose I'm just as bad as none of my personal sites don't have IPv6 records either, but then again my server host doesn't provide any native addresses yet.

What would you have to do?

[mitchplanck]

At my company we have two main Cisco routers. One is about 7 years old and the other about 3 years old. The older one used to be able to handle full BGP routes but as the routing table grew and Cisco IOS bloat happened it's 128MB of RAM could no longer hold all that. I've had to trim it to connected routes and I can't update the IOS as all the current ones use too much RAM and wouldn't even work with what I've got it doing. So forget doing IPv6 on that one.

The other router isn't doing BGP and could probably handle IPv6. The problem then becomes all the machines on our network. Lots of legacy systems. If they can't handle IPv6 then we either have to replace them or have an IPv4/IPv6 gateway - another machine probably since I don't think the newer router could handle this.

The next issue then becomes our upstream providers. Neither of them are Tier-1 providers and neither offer IPv6 addresses yet.

Then there's the issue of network admins knowing how to use IPv6 addresses. I've been doing a bit of reading about them but until I start actually working with the systems it won't really sink in. I know my colleagues here haven't been attempting to learn anything about this and it will probably fall to me to educate them on this.

I'm not looking forward to any of this...

Re:Well duh

[Gorbag]

You see its the shortsightedness that in the long run costs you WAY MORE than if you simply keep the options in mind and work towards a solution.

Actually it's an insurance problem. There are an infinite number of possible future disasters, and we'd all be broke in the stone age if we tried to address all of them. Like lazy evaluation, sometimes putting off actually solving the problem makes a lot of sense because the problem may never even materialize, or by the time it does, there are better and cheaper ways to fix it.

Climate instability (nee 'global warming') may be a case in point. It's not clear that CO2 is the cause, and even if it is part of the problem, sequestering it is getting cheaper (certainly a lot cheaper than having everyone stop driving or using electricity). There are other "problems" that seem to be more excuses to spend massive amounts of money relative to the actual risk (anything from worry about near earth collisions, to the "health care crisis")

Adam Smith's invisible hand will take care of many things. While I certainly am not arguing against research, I don't think rushing to implement half baked solutions is ever a good idea (though it seems to be the only way things get done in Congress).

This too, shall pass.

Re:FUD

[Brian Gordon]

NAT will solve the problems, but why live with that when we can actually come up with a viable solution- IPv6? It will be expensive to implement because, like always, past engineers haven't planned for their 1970s technologies to ever go out of date, and whiny slashdotters will finally have to upgrade their windows boxes to Vista because XP has 1990s networking support (read that pdf if you don't believe me). But we'll end up with a significantly better Internet than if we just keep expanding NATs around more and more IP addresses to free up address space.. the way we're going, eventually (and keep in mind that "eventually" in computing usually turns out to be in less than a decade) you're going to have to be a multibillion-dollar conglomorate representing thousands of web hosting companies just to bid for a single 5-address block of address space... though the way inflation's going, little billy and his friends might be able to pool their allowance and come up with that kind of money :) But can you imagine how horrifying the architecture of the internet will be if the solution is NAT, NAT, NAT? Development in router design is already unable to keep up with traffic growth. How are you going to pay for a $100 million server farm just to manage the American Eastern Seaboard NAT, and can you imagine what the latency would be to go through a 10 terabyte NAT table? Might as well upgrade to IPv6, save yourself the trouble of trying to stay v4.

Re:Off topic

[mikael]

I've read some of the reviews for that book. The story about everyone in a street ending up using water amplifiers (pressure boosters) to guarantee that they get their fair share of water is funny. Some things don't seem to different from other parts of the world.

Dumping garbage in the street - that happens elsewhere whenever the authorities impose apparently madhatter legislation; Example, a country in Europe creates a whole nation-wide network of recycling centers to reduce the amount of waste going into landfill - Totally sensible. Anyone could enter, and recycle their old boxes, cartons, polystyrene boxes, lawnmowers, furniture, whatever. Then the authorities decide that too many people are making too many journeys, so they decide that each family can only get a ticket to allow them to recycle once every two months. So now, everyone drives around looking for somewhere to dump their recyclables, even filling in the communal rubbish bins of neighbouring villages. Others simply burn it instead.

Oddly, Google will Fund Switch to IPV6

[RonBurk]

The untrue, but unchangeable, folklore of Google Adsensers (people who try to make a living via free search engine traffic to web pages that display Google ads) is that it's crucial for your Google rankings that your website be hosted on a server with a "static IP" (I don't know why people can't say "IP address" anymore in that community). These are the folks that will pay more, and more, and more for the privilege of having their own IP addresses as scarcity increases. Thus, Google money will ultimately and indirectly fund the switch to IPV6, as ISPs serving the hordes of must-have-my-own-static-address Adsensers will be able to afford conversion.

The best thing that can be done to accelerate this process is to perpetuate the myth that it's crucial for your search engine rankings to host your website on a server with its own static IP address.

Re:Is this REALLY a problem?

[Bryansix]

But what if I want to run TWO web servers? What then? I can't port forward port 80 to two places. IPv6 is the real answer and the telecoms and ISP's need to get their heads out of their asses and support it already. The DNS records already exist.

Re:Three Things for Widespread IPV6 Acceptance:

[anticypher]

That is true, the AEBS only does 6to4 tunneling, but that tunneling works with both Hurricane Electric and Sixxs static service. In fact, it works pretty well for home use, and if you've got Macs behind it, they pick up their IPv6 address quite nicely and it all works pretty transparently. I'd recommend it as a good (but expensive) way for geeks to get up and running on v6 with a minimum of hassle.

I've tried making some of my AEBSes work on a native dual-stacked network connection, with no luck. It doesn't listen to Router Advertisements, DHCPv6 service, or anything I can detect. You can manually set a local node address, but it doesn't seem to route or bridge at that point. Apple's forums have been less than enlightening, and I've never heard back from their developer tech support on the issue. There firewall is very buggy, it seems to be just a simple two line IPFW entry to block incoming connections and keep state on outgoing. Any kind of P2P activity causes the firewall to fail badly.

A Chinese company last year gave me a DSL router that speaks IPv6. It is some kind of OEM version of a popular Belkin model, but with a Chinese only firmware installed. They claimed it was the most widespread model inside of China, where many ISPs can only hand out IPv6, and there is a NAT-PT+totd translation service somewhere within the ISP. I played around with it for the few days I had, and couldn't figure out how to make it work for what I expected. Some of the configuration pages looked identical to Belkin, but in Chinese and with some obvious IPv6 entries on some pages. It certainly worked as an IPv6 only DSL modem, and dual-stack v4/v6 just like a Belkin, but I never got it to work with a NAT-PT gateway.

There was a muttered admission that by having a lot of IPv6 only services that aren't announced outside of China it makes it a lot easier to do the great firewall of china function. There is apparently a government funded push toward IPv6, but none of it is announced externally because of firewall issues.

the AC

6 to 4

[PeterJFraser]

I do not know why since every IP4 address has exactly 1 IP6 address, the backbones could be made to run IP6, and at the edges, there would be a transparent 6 to 4 and 4 to 6 for those set of addresses. Big companies who converted to IP6 would directly continue to use their IP4 address in its IP6 format so IP4 users could communicate with them. Associated with each IP4 address is 2^16 IP6 sub addresses with the sub address 0 being the natural mapping for IP6 to IP4. If users were initially restricted only to the IP4 sub set of IP6, it still allocates each IP4 address 2^16 new IP6 address, so there would be no shortage for users with at least 1 IP address. An IP6 user (using this subset) setting up an connection would attempt to use IP6, but if the connection failed then the router would NAT the none zero sub address to IP4 subset address and try again.

Re:What would you have to do?

[jguthrie]

You know, when a megabyte was a lot of RAM, this whole "routing table is getting too big" argument carried some weight. The size of the routing table is an incredibly stupid thing to worry about, now. Perhaps it isn't because people don't do routing the way that they should.

With BGP, you're not going to route anything smaller than a /24, so your entire routing table can be an array with 2^24 entries in it. Those entries are going to be (for the most part) outbound queues, one per interface, so all you need in most cases is a single 8-bit number with the queue number in it. I can buy 16 MBytes of RAM with pocket change, nowadays.

To route a packet, you simply shift the destination address right eight bits, look up the queue number, and put the packet in that queue. The total elapsed time for that operation is easily measured in nanoseconds. Some queues might do further routing (you might have a queue to route local packets, for example,) but you wouldn't see a lot of those on any router that needs a full picture of the Internet.

Now, building that array is a lot more work, but it's not that much more work and, besides, it's the handling of the incoming packets that is time-critical. Processing of BGP (or RIP or OSPF or whatever) can take a lot more time and still be plenty fast enough to handle the changes as they happen.