Less Than 1 Year Until The Internet Runs Out of Addresses | Hacker News Reader

Less Than 1 Year Until The Internet Runs Out of Addresses

readwriteweb.com

23 pointsssclafani24 comments

The world isn't going to end when we "run out of IPv4 addresses", because it can't: a huge portion of all the world's networking software can't deal with IPv6 addresses and can't be forklifted out.

The train has already left the station on NAT; end-users (and components in the "internet of things") expect transient addresses, and large-scale apps are designed so that rendezvous points are the only things that need fixed addressing.

Meanwhile, we'll eventually hit a point of scarcity with IPv4 addresses where allocations will stop being political and become market-driven instead.

Finally, all talk about a shangri-la of direct v6 addressing ignores the fact that no matter how quickly ISPs adopt the protocols, and even after the majority of Internet software that end-users demand are ported to handle v6 addressing, BGP routing still isn't going to scale to individual direct addresses. What's the current situation on prefix filtering at the major ISPs? How easy is it to get a /24 (even) portable?

Here's a link to think about:

http://www.circleid.com/posts/20090609_verizon_mandates_ipv6...

"The device shall support IPv6. The device may support IPv4"

This is requirements for the LTE devices from Verizon. Of course all the devices being linux-based, I expect dualstack for quite some time.

As for provider-independent addresses, I understand you should be able to get /48: http://www.ripe.net/ripe/docs/ripe-483.html

As for the Locator-ID split problem (which what the BGP scaling problem is about), there are a number of experiments going on at the moment. LISP (http://tools.ietf.org/html/draft-farinacci-lisp) is one of them. There was a few more but with less memorable names (edit: memorable for me. because I did hear more about LISP)

Of course the world is not going to end. It will just start developing an ever increasing slice of IPv6-only, whereas IPv4 growth is bounded.

Google, Facebook, YouTube, Netflix are already IPv6-enabled. This is "The Internet" for your average mom and pop consumer type. And maintaining the NATs with the few million hosts behind is a bit of work. So I would not be surprised if a cheaper "IPv6-only internet" options might emerge.

Full disclosure: my day job is at cisco. And untangling the creative NAT configurations and bugs caused by them has paid my bills for the past 10 years. Everything posted here is my personal opinion and not of my employer.

Or,

"The device SHALL support IPv6. The device MAY support the Internet."

Pure grandstanding.

The real problem isn't address exhaustion, it's managing the migration from IPv4 to IPv6 and running mixed networks in the interim. I deal with 3 ISPs on a regular basis. Of which only 1 (Comcast, surprisingly enough) has their act together and can provide you with IPv6 connectivity. Meanwhile out of 5 datacenters I regularly deal with there is not a single one that can provide a direct IPv6 routing to the rack. So for the moment, at least for the networks I deal with it's IPv4 and tunneled IPv6 and given the nature of the business, IPv6 is really not on my priority list, and won't be until it actually saves people enough money to cover the switching costs.

There are many corporations hording Class B blocks. Lots of them. Time to use them or lose them.

This is also a Windows problem believe it or not.

Class E block has 268 million addresses but no existing version of Windows will see/talk to them.

The E block is really 'reserved' and 'meant for test purposes'.

If Windows can't talk to those IP addresses they're actually within spec. Though it is a useless waste of a lot of bits.

Not only that, but Windows before Vista won't work with IPs ending in .0 or .255

http://support.microsoft.com/default.aspx?scid=kb;en-us;2815...

I figure one change that will happen at the beginning of this "crisis" will be that ISPs will start NATing traffic, as tptacek said.

A lot of ISPs already have policies for not having daemons on ports, so they could use that as justification for not having your own IP completely accessible from every port (which I don't believe would be possible without explicit port forwarding from the ISP itself).

What would this mean for end users' internet? Not much, but IPv4 IPs would be shared by multiple users (hopefully not many at first). A lot of websites have session policies and all sorts of mess based on the remote address, so it will be interesting to see how the aging webapps deal with the change.

I figure servers will get first priority at v4 addresses. Yeah, there's useless crap going on like IRC vhosts, but I figure once the market closes in on the addresses people will likely sell them (hell, I know people squatting on address blocks until the 'apocalypse', thinking they'll score a huge profit!).

Fun times.

But wouldn't the end users be behind a double NAT in this scenario? (ISP's nat, end user's router's NAT.) Lots of voip protocols break when they try to deal with double NAT, and I'm sure there are other that will break too.

Even single NAT is not trivial when a few hundred thousands of active users are behind it, just because of the sheer scale.

For a major part of the users "The Internet" == "The Web", and applications like Skype use any available transport including avian carriers, this limits the impact.

But does having NAT increase the code complexity and increase the maintenance cost ? It does.

So, I would treat it as a short-term high-interest loan. If you do not have other choice - sure, go for it. But if you can avoid it - you will avoid a couple of gray hairs too.

This is just until the last /8 is assigned to an RIR, right? There will still be plenty of IPs available in many locations throughout the world... it'll just be North America that will have a hard time of it.

Why North America? For the most part North America has mature deployment of Internet services. What is really driving IP address exhaustion right now is Asia where millions are coming online for the first time. They are the ones that will have a hard time when unused IPv4 addresses become scarce.

I would think that in places without heavily established infrastructure, it would be much easier to deploy straight to IPv6.

Perhaps the only reason we're running out of IPv4 addresses is because you can still get large blocks of IPs for questionable reasons like "IRC vhosts" and "proxies"...

http://article.gmane.org/gmane.org.arin.ppml/8304

Plus, both internet connections at my workplace gave us 6 IPs without us even asking for them.

Plus, you can't get smaller blocks allocated even w hen you need them.... I have a legitimate need for an independent /24 that I can BGP route around to a few datacenters. ARIN says I can only apply for a /22 at the smallest. People are forced to apply for more address space than they really need in order to get any at all.

As for hte reasons "WHY" - it was never supposed to be a judgement as to relative value - only that they were actually needed and going to be used. IP addresses were never meant to have value in the first place.

Fortunately that's very close to changing. The ARIN Advisory Council recommended adoption a month ago and now it is just awaiting board review.

https://www.arin.net/policy/proposals/2010_2.html

Not to make light of the situation, but I've been hearing this same thing for the past 6 years. It gets a little old and diminishes the message.

* We already have our /48 bought & deployed in production.

Make me wonder why our hosting provider keeps handing over 6 IP addresses for free each time we order a server (we usually don't need them).

It could be that they have a large enough allocation that they can keep doing this for some time to come. However, if they do start running up against their limits, then they are going to have to figure out how to reallocate and possibly even claw back those giveaways. I don't envy them, because even a low monthly charge per IP is going to cause bad feelings amongst customers who formerly got addresses for free.

What is the company, and are you otherwise happy with their service?

Softlayer, and yeah, they're great (at least on their dedicated hardware, not sure about their cloud services).

The internet will not run out of IP addresses, a market will develop for internet addresses and as price increases those that own IP blocks will sell them to those that need them.

There might be a business in there setting up a clearing house for IP addresses.

Four syllables: IPv6.

I heard that a year ago.

Reply on news.ycombinator.com