Some ISPs are already supporting IPv6. Admittedly, only a handful of residential ISPs are there yet.

Comcast is doing trials now and will probably be adding IPv6 for most customers by the end of the year.

If your ISP isn't doing IPv6 yet, it's time to start asking them about it.

RIRs will be out of IPv4 before the end of the year. That means ISPs that want to keep adding 30-50,000 customers per day to the internet are going to have to do something different from what is being done today. IPv6 is the solution to that problem and it will roll forward rather quickly after IPv4 runs out.

You can plan for it now, be proactive about preparing, and be ahead of your service provider and others, or, you can stand on the tracks waiting until you hear the train coming around the curve. I guess which one you choose depends on how fast you can run and how confident you are in your hearing.

Um, that depends.

If your ISP or the ISP for the company you work for operates with your attitude, it may be well past the point when you loose connectivity to things you consider important when you stop ignoring IPv6.

I would say that instead, you should be contacting said ISPs and making sure they will be bringing IPv6 to you sooner rather than later.

Otherwise, by the time you stop ignoring IPv6, the modern IPv6 internet may be already ignoring your legacy IPv4 environment.

First, it's not really IPv6 brokenness so much as it is an issue with hosts that think they have IPv6 connectivity, but, really don't.

Second, in most cases, affected users will see long page load times, not complete inability to access the site.

The 0.05% number is probably pretty accurate. Several sites have used embedded tests to measure this and come to the same number. However, the good news is that a year ago, this was 0.1% and it is continuing to trend downward.

With IANA running out of IPv4 this month, it's not surprising that Yahoo is moving forward. It's disturbing that so many others appear not to be.

I did not get my addresses from a tunnel broker. I got mine from ARIN. I can use them outside of a tunnel, just as my IPv4 addresses. However, there is not a cost-effective non-tunnel way to get BGP based IP transit services into a house, so, my IPv4 and IPv6 BGP sessions are handled over GRE tunnels.

If you sign up for IPv6 service with an ISP you should be able to get at least a /48 from the ISP. If you are multihomed, you should easily be able to get at least a /48 from your local Regional Internet Registry.

There are many options for obtaining IPv6 addresses today. There will be more in the near future.

I disagree. I was easily able to obtain 1,208,925,819,614,629,174,706,176 of them and so can anyone else from various internet tunnelbrokers* or from several other sources.

*Full disclosure: I work for the company that provides the largest tunnel broker, but since it is a free service and I have not named the particular tunnelbroker, hopefully this is not regarded as a promotional message. It is intended as informative only.

Nope... I deliberately chose one that was even beyond Peta because the scale is really much much larger. Peta wouldn't cut it. Exa is probably even short, but, I don't know what the term is for exa *10^11.

IPv4 = 4.2*10^9
IPv6 = 3.4*10^38 (or almost IPv4 * 10^29)
Exa = 10^18

So, IPv6 = IPv4 * Exa * 10 ^ 11.

Bottom line, the claim "640k ought to be enough" was based on 10x standard memory of the day.

IPv6 ought to be enough is based on 8.5 * 10^28 times the IPv4 address space, so an accurate memory comparison would be the claim:

85 brontobytes ought to be enough memory for everyone

Do you know any one who is likely to even be able to conceive of 85 brontobytes in the lifetime of anyone now living, let alone actually procure it or address it in a system?

Didn't think so.

I'll try...

I have no idea of any meaningful measurement of Library of Congress for comparison, sorry.

It takes 39 digits to define the number of addresses in IPv6. Only 10 digits to define the number of addresses in IPv4.

If you treat each address as a unit of mass and consider IPv4 to have mass equivalent of 7 liters of water, then, IPv6 would have mass equivalent roughly to Earth. (The whole earth, including all the oceans, lakes, land masses, people, buildings, etc.)

In IPv4, there are more than 1.5 people alive today for every address.

In IPv6, there are 50,041,524,547,196,832,862,260,971,681 addresses for each person alive today.

Or, perhaps consider the following:

The US public debt is 13,848,000.000,000. If IP addresses were pennies, we would need 3,462 IPv4 internets to pay it off. The IPv6 address space, converted to pennies, OTOH, would pay the public debt more than 24,572,672,365,752,344,270,896,491 times.
(If anyone wants to send me even a single IPv6 /64 network worth of pennies, please
email me for contact information.) ;-)

Hope that helps.

There is a difference here. IPv6 would be the equivalent of IBM saying something more like:

640 exabytes ought to be enough for anyone.

(note by exabyte I mean 1000 terabytes, not Exabyte the brand name of many 8mm digital video tape drives).

340*10^36 (the IPv6 address space) is more than 10^26 times the current demand for addresses.
Compare to 640k which was roughly 10^1 times the standard memory size for machines of the day.

In fact, today, I doubt you can identify many (any?) machines with more than a terabyte of RAM.
In fact, it's rare to find more than 128GB of RAM capacity in most machines. (64GB is roughly
100,000 times the original 640KB number, so 128GB would be 2*10^5 times 640KB).

To put the comparison in some perspectives you might be able to wrap your head around...

If you were to allocate an almond M&M for every 256 IPv4 addresses, the resulting amount
of almond M&Ms laid out in a 1-M&M thick layer would cover only 70 yards of an american
regulation football field (NFL, not FIFA). (16.7 million M&Ms, 1 for each IPv4 /24 prefix)

Contrast that with the number of IPv6 /64 prefixes (a bit more than 18 quintillion) which
would provide enough M&Ms to fill all of the great lakes.

Where each /24 can accommodate a single router and up to 253 other hosts, each
IPv6 /64 can accommodate more hosts than you could ever physically put on any
conceivable scale of network gear (18 quintillion+ hosts).

There will not be a likely shortage of IPv6 addresses in any of our lifetimes.

Based on the current trends and even the most liberal theories of allocation being even proposed, in about 50 years, we MAY have allocated as much as 0.5% (yes, 1 half of one percent) of the IPv6 address space.

Variable length schemes are impractical at backbone forwarding rates. Hardware to support a variable length scheme would be incredibly expensive.

While the transition to IPv6 does involve some effort, it is not nearly as bad as many have claimed it will be. My employer operates a three-continent
dual-stack backbone. I run fully dual-stack at home (except for an amplifier, some TiVO boxes and an old terminal server which are not IPv6
capable).

Transitioning my network took a total of less than 12 man hours with an elapsed time of approximately 7 days.

Turning IPv6 off now just means you have to turn it on in a few months. I would rather avoid making two changes to all the hosts in my network.

The simpler solution for the basic network where IPv6 doesn't matter yet is to make sure your router null routes the IPv6 default (::/0), get valid IPv6 addresses, build a basic subnetting plan and put it in place on your routers with appropriate RAs. That will prevent hosts from trying to build 6to4 or Teredo or ISATAP or other bizarre kinds of autotunnels and give quick negative responses to attempts to reach IPv6 hosts resulting in timely fallback to IPv4. Simple, efficient, and, when you do actually need IPv6 connectivity, you just need to change the configuration on your routers (which you'd have to do anyway).

As to when to deploy IPv6, if you're running a network full of end users behind an IPv4 NAT using RFC-1918 space, then, no, you probably don't need to convert that network over right away, but, you will want to deal with all those systems that are now shipping with IPv6 on by default as I have described above or you will see user complaints as a result of their attempts to reach a growing mass of IPv6 content.

However, if you have any public facing content or services (as most businesses do at this point), then, you're going to want to make sure that those are reachable via IPv6 as well as IPv4 as soon as possible. Certainly within the next 12 months or so.

The people depending on the current address calculators and an 18-month clock to RIR runout after IANA exhaustion are in for some rude awakenings.

First, the clocks are wrong. They don't seem to correctly account for current utilization rates, nor do they account for the fact that 5 of the 14 remaining
IPv4 /8s are reserved to be distributed one to each regional internet registry when the others are exhausted.

At the beginning of 2010, there were 21 IPv4 /8s in the IANA IPv4 free pool and 5 in the reserved section described in the last paragraph. Today, less than 9 months later, there are only 9 /8s remaining in the IANA free pool and 5 still in the reserved section. (Note that the clocks all talk about 14/224 which is 9+5).
At the current rate of consumption, we're not talking about 285 days to IANA runout, we're talking January or February of 2011. Feb. 28, 2011 is 194 days from now in my current timezone (Thursday, 20 September, 2010). (Notice the 91 day (or more) error in the countdown clock).

Additionally, once IANA runs out of IPv4, the RIRs aren't going to simply coast for 18 moths. APNIC, RIPE, and ARIN will likely be in a race to see who runs out first. I think the smart money is on APNIC. However, whichever one runs out first, you can bet that the multinationals (i.e. the largest consumers of IPv4 addresses) in any one of those three regions will start pulling space from the other regions too. As a result, whichever one runs out first will accelerate the other two rather abruptly. I predict that the first RIR will run out on a timeframe more like 6 months after IANA exhaustion rather than 18.

It's less clear what will happen with space in the AfriNIC and LACNIC regions due to unique circumstances.

IPv6 is no longer an option, it is a requirement. Time to stop with the FUD and misinformation and start facing the cold hard facts staring us in the face.

Yes, the earliest predictions of runout turned out to be wrong (only because NAT was developed _AFTER_ those predictions were issued, btw).

However, the predictions today are mostly wrong too, but, not in the direction you want and certainly not as far off.

Failure to deploy IPv6 at least to your public content and services within the next 12 months will place you at a competitive disadvantage against other companies that do. That disadvantage will only increase with time. It is also critical to deploy IPv6 capabilities to your support staff and your IT
departments so that they can become familiar with it and learn to troubleshoot it and assist customers with IPv6 related problems.

About the author of this comment: Owen DeLong is an IPv6 Evangelist at Hurricane Electric and an elected member of the ARIN Advisory Council. His statement here is not an official statement of ARIN or the Advisory council, but, his participation on that council keeps him keenly aware of the state of the IPv4 free pool and the issues surrounding IPv6 transition and IPv4 exhaustion.

Network engineers said (briefly) 12 years ago that "The internet is in danger of running out of space very quickly if nothing changes".

Then NAT came along and, for better or worse, it was widely deployed. Better in that it dramatically extended the useful life of IPv4 through dramatic address conservation. Worse in that we now have an entire generation of network engineers that not only fail to understand the advantages of a peer to peer end-to-end networking model, but, somehow have managed to turn terms like peer to peer into dirty words in many circles and create the illusion that because NAT depends on a stateful inspection firewall underneath, NAT somehow offers security.

The reality is that security comes from the stateful inspection which can exist with or without the header-mangling (aka NAT).

Now, you're starting to see network engineers and address policy people (I fit into both camps) say this again. Why? Well, because at the beginning of 2010, there were 21 /8s available for allocation to Regional Internet Registries in the IANA free pool and 5 in a reserve to be given 1 to each RIR when the free pool is exhausted. Today, less than 9 months later, there are 9 /8s remaining in that free pool. At the current rate of consumption, IANA will run out of addresses most likely in January or February of 2011. APNIC, RIPE, or ARIN will then enter a race to see which one runs out first, likely within 3-6 months. Once one of them runs out, the other two won't be far behind, in part because it is very likely that many of the address consumers in any one of those three regions will be able to acquire space from the other two (lots of multinational companies have presences in all three regions).

IPv4 exhaustion is real. Is it time to move your entire IT network inside your corporation from 10.0.0.0/8 over to nice shiny public IPv6 addresses tomorrow? Probably not. However, it is definitely past time to add IPv6 capabilities to your public-facing content and services and to make sure that your support department and your IT staff has at least access to IPv6 networking and some IPv6 labs where they can start learning and debugging stuff.

If you fail to do so within the next 12 months, it is very likely that your company will be at an ever increasing competitive disadvantage vs. companies that did or will by that time.

Remember, if you start turning IPv6 off on machines today, you'r just going to have to turn it back on later. A much better solution is to find ways to turn IPv6 support on in your LANs in such a way that it properly sandboxes the IPv6 and lets the hosts know that they don't have global IPv6 connectivity. This isn't particularly hard to do. If your routers issue RAs on the local network and then provide ICMP unreachable messages back for unroutable packets, the fact that most of your hosts have IPv6 addresses and can't reach the IPv6 network will be almost completely transparent to them.

About the author of this comment: Owen DeLong is an IPv6 Evangelist at Hurricane Electric and an elected member of the ARIN Advisory Council. Owen is not speaking on behalf of ARIN or the ARIN AC in this post, but, being on the AC, he is very aware of the statistics surrounding IPv4 address utilization.

If they have made false and defamatory statements about your product to your customers, then, it seems to me that would likely come under the jurisdiction of the "libel" laws. Check with your attorney (I am not an attorney). If you have had customers trying to return products to you or can otherwise document actual damages, you may well be entitled to recover those from the company. Their failure to cooperate with your very reasonable request for a retraction on their web site could also open the door to punitive damages.

At the very least, if any of these apply, your attorney might be able to draft a nice letter convincing them that a retraction is far cheaper than litigation.

Vote with your $$ and your feet. If a manufacturer isn't giving you the support you need, and you have a significant business history with them, then, call up your sales rep. and let him know about your next big order and explain why he's not getting it.