In an article in the Hindustan Times, The American Registry is quoted as telling us that they are running out of IPV4 addresses.
On Wednesday July 1, the ARIN - in charge of North America - was forced to turn down a request for a block of IP addresses for the first time in history. The CIO Richard Jimmerson told CBS news "We are weeks away from having zero left."
On the same subject, Arstechnica details the emerging IPv4 address trading market.
We spoke to Janine Goodman, vice president of Avenue4, a broker of IPv4 addresses, about what to expect in the short term.
"IPv6 is going to happen, that's the direction it's going," she said. "But it's going to take a while. Organizations are not ready to turn to IPv6 tomorrow; this will take a few years. A transfer market allows for the transition from IPv4 to IPv6 in a responsible way, not a panicked way."
"The price for blocks of IPv4 addresses of 65,536 addresses (a /16) or smaller is about $7 to $8 per address in the ARIN region. In other regions, which have fewer addresses out there, the price tends to be a little higher," Goodman said. "We expect the IPv4 market to be around for at least three to five years. During that time, the price per address will likely go up and then finally come back down as IPv6 is being widely deployed."
(Score: 2) by TheRaven on Thursday July 09 2015, @09:18AM
If you're going to make the addresses bigger, then you need a new IP header and you're going to break absolutely everything that relies on the old address format or the old header format. Allowing the kind of bridging that you suggest would have required the same level of complexity as 6to4 and ended up with some IPs being reachable and others not, with the same protocol stack, which just gets messy.
If you're going to make the address field bigger, then you may as well solve some other problems. One of the big scalability issues with the Internet currently is the size of routing tables. If you're a handing out class Cs (which we are, because address exhaustion is not a binary thing) then every backbone router needs to know where every class C is, which means that the fast path needs 2^24 entries in the routing table. Companies that build big TCAM chips love this, but no one else does. With IPv6, the address space is sparse enough that you can assign addresses more hierarchically, so most of a continent will have the same prefix (meaning the decision whether to route something across a transatlantic cable or onto the local network is very easy, for example), then ISPs within a country can each have a prefix within that, and you can leave enough gaps that as new ISPs emerge this structure can be maintained.
Sparse addressing is also useful at the endpoints: it's required for the stateless autoconfiguration to work. IPv6 devices can discover the network prefix and just pick a local address, ask if anyone else is using it, and retry if they're not. With a 64-bit subnet, even with a lot of computers there's a tiny chance of collision and a minuscule chance of more than one collision, so this works very well.
So who the hell wants to use it? People who run large networks, people who build routers, and people who want to use peer-to-peer protocols. Basically, everyone except ISPs who like being able to sell static IP addresses for a lot of money.
sudo mod me up