http://en.wikipedia.org/wiki/IPv4_address_exhaustion

It has been brought to my attention from my Comp Sci friend that IP Addresses are soon running out!
In fact, his prof has calculated that they would run out at around Feb 18th!

There is a new system, but it hasn't been thoroughly tested. I don't know what or where it is. I am no computer expert.

So maybe one of you can describe this better?

Ever wonder what Teredo is on Windows 7?

Win 7 computers are already ready to accept the IPv6 crossover, they just need to command.

*sigh*

IANA controls the IP addressing scheme. There are 5 regional registrars under them. IANA allocates addresses in blocks of /8 (16.7 million) to the registrars as they request them. Registrars in turn on-sell these addresses to ISPs. ISPs use them for their customers, both residential and business - a static IP address is one that you get to keep permanently, while most residential connections get automatically assigned IP addresses and they can change over time. You need a globally routable IP address to allow bi-directional communication on the internet - that is for packet replies to actually reach your computer. You need static IP addresses to host thing like webservers or other types of servers on the internet.

IPv4 has 4.3 billion addresses in total, but a lot of these have been fenced off for specific purposes: 127/8 is loopback, 10.0.0.0/8, 172.16-31.0.0/16, 192.168.0.0/24 are private, 224.0.0.0/24 to 240.0.0.0/24 are multicast, 241-254.0.0.0 'reserved for future use'. The remaining IP addresses are available to be allocated to ISPs in /8 blocks.

What has actually just happened (your friend's professor is out of the loop) is that APNIC, one of the regional registrars, has requested the last 2 available /8 blocks from IANA, which has triggered IANA's final allocation policy of distributing the remaining 5 /8's, 1 each to all of the regional registrars.

This means each registrar has at least one full /8 block (16.7 million addresses) that they can sell to ISPs as they see fit. APNIC estimates that it's current policy of allocations will continue for the next 3-6 months, and after that they'll have to change.

IANA has run out of addresses that it can allocate to the RIRs (unless some companies relinquish their address blocks back), but the RIR's themselves still have a lot of IP addresses they can give out to their supplicant ISPs. The ISPs in turn will each have their own stockpile of IP addresses that they can use. ISPs will probably also be able to reclaim some inefficiently used IP addresses if they need to - there hasn't been any point doing it up till now because they could always just ask for more from their regional registrar, but now that the RIR's are running out, ISPs will have to make sure they're using the addresses they have as efficiently as possible.

What does this all mean? Probably not much will change for the next year, as RIR's and ISPs will have IPs still left to allocate. But towards the end of the year, or next year, some ISPs may run out, at which point they may start charging more money to their customers for IP addresses, and in particular may start taking static IP addresses back from residential customers.


The solution, which has been standardised for over 10 years now, is IPv6, which has addresses that are 16 bytes long instead of 4 like IPv4, meaning there are 34,028,236,692,093,846,346,337,460,743,177,380,000,000 addresses available - considerably more than the 4,294,967,296 available in IPv4.

The problem with IPv6 is that it is not backwards compatible with IPv4 - all of your (networking) software and hardware will need to be upgraded or replaced so that it can handle IPv6. Vista and Windows 7 handle IPv6 out of the box, and Windows XP has supported it since service pack 2 I believe - Apple and Linux also support it. Even though IPv6 has been around for 10 years, it hasn't been seriously or widely used except for the last 2 or 3 years, so there are still many corner cases around where IPv6 tools or functionality simply doesn't work the way it should, because it's buggy and it hasn't been seriously thrashed.


Eventually once most ISPs have transitioned over to IPv6 (probably by about 2015-2016) this means when you get IP addresses from your ISP, instead of getting a nice IPv4 address like 145.13.7.3, you'll get something that looks like 2001:db8:85a3:8d3:1319:8a2e:370:7348


Note: I work as a software engineer at Allied Telsis so I do this stuff all day. Last couple of weeks I've been fixing bugs in OSPF for IPv4 and IPv6.

Post has been edited 4 time(s), last time on Feb 2 2011, 7:31 am by Lanthanide.

I've heard it's mostly just the US still using IPv4 that's kept the transition to v6 from being fully accomplished. Much of the rest of the world is now using IPv6.

Yeah, well that's not true.

IPv6 has mostly been taken up in Asia (APNIC is the Asia-Pacific RIR) at the moment, particularly China and Japan because of their high population (China) and high penetration of internet access (Japan). But it would make up less than 1% of all globally routable addresses (probably more like 0.1%).

Edit: here's some comparison of the global IPv6 vs IPv4 routing tables: http://bgp.potaroo.net/v6/v6rpt.html
Looks like my guess above of 0.1% was a bit too pessimistic: 0.39% of all globally routable addresses are IPv6.

Post has been edited 1 time(s), last time on Feb 2 2011, 8:26 am by Lanthanide.

It's NAT.

Connect four computers to a router. Router to a modem.

Each computer has a "fake" IP inside the router (e.x. 192.168.0.123). The router has a "real" IP that the world sees. This is called "NAT" -- "Network Address Translation".

(This is related to gaming NAT, i.e. Open, Moderate, Strict. Basically specifies how much IP-related information your router divulges. The more, the merrier.)

NAT means several devices can use one IP, basically. This has slowed the need for IPv6 adoption. Wikipedia said so.

IPv4 addresses ran out in ARIN (North America) this past September. Surprisingly, the sky didn't fall.

http://www.networkworld.com/article/2985340/ipv6/arin-finally-runs-out-of-ipv4-addresses.html

4.43% of all routable addresses are now available via IPv6.

When you get IPv6 from your ISP (or one of the IPv6 tunnel providers), you usually get a whole /64 block or possibly even larger, because of how the protocols for automatic configuration of IPv6 were designed, so you have enough public IPv6 addresses for every device you could possibly have attached.

For reference, a /64 block in IPv6 means that your prefix from the ISP is only the first 4 of 8 groups of numbers (64 bits), and you get to define the entire last half of the address. In other words, you have 18,446,744,073,709,551,616 IPv6 addresses available for your own personal use.

A way to put that in perspective: when residential customers get a fixed IPv4 address from their ISP, they get a single IP. Some ISPs charge extra for this.

When you get an IPv6 address, *everyone* gets 18,446,744,073,709,551,616 addresses by default. Which is 4.3 billion times larger than the entire IPv4 address space.

It's a bit stupid IMO; they should dish out /112s to residential customers as the default prefix; that still gives you 65,536 addresses to play with.

The auto-configured link-local addresses that use MAC addresses are link-local and private by default; they aren't routable to the internet. So they aren't a reason for allocating /64 to end-users.

The most common way to assign public IPv6 addresses to individual devices, which is the most widely supported, actually uses the same last 64 bits of those link-local addresses (everything after the fe80:: ) and tacks it onto the 64-bit prefix for the public IP. This is what I was referring to in my last post and is the reason they assign /64's minimum.

Post has been edited 1 time(s), last time on Nov 7 2015, 10:07 pm by ShadowFlare.

Doesn't that make your individual device personally identifiable if my MAC address is embedded in my IP? If I were to choose a random subnet different from my MAC address, would I start breaking things?

Correct. That's why a feature named Privacy Extensions were added. They randomize the host part of your IPv6 address.

That said, no. You can also choose your own addressing scheme. ::1 for the host part is often used in manual configurations. Admins might add more significant bits to differentiate teams: ::1:1