IPv4 Unallocated Addresses Exhausted by 2010

An anonymous reader writes "Ars Technica is reporting on how the unallocated IPv4 address pool could run out as soon as 2010. The IPv4 Address Report gives details on just how fast the available pool of IPv4 addresses is diminishing. Will ISPs be moving towards IPv6 any time soon? Or will IPv4 exhaustion become the next Y2K?"

Comments

Reshuffle existing IPv4 space

[McDutchie]

They could delay the inevitable by reallocating existing IPv4 space more efficiently. Many old/historical allocations are inefficient. Apple Computer, for example, has all of the 17.x.x.x space, comprising 256^3 = more than 16 million addresses, which is just plain absurd in this day and age.

Re:Reshuffle existing IPv4 space

[RzUpAnmsCwrds]

They could delay the inevitable by reallocating existing IPv4 space more efficiently. Many old/historical allocations are inefficient. Apple Computer, for example, has all of the 17.x.x.x space, comprising 256^3 = more than 16 million addresses, which is just plain absurd in this day and age.

Don't complain about Apple. HP has all of 15.x.x.x and all of 16.x.x.x, because they purchased DEC who also had a class-A.

Interestingly, HP is the only company that effectively has a /7 because their block is contiguous.

Re:Reshuffle existing IPv4 space

[TooMuchToDo]

Actually, you don't need an army of lawyers. Those Class A blocks are delegated solely at the whim of ARIN (at least those Class A blocks that fall under ARIN control). If ARIN has a vote, and the majority of stakeholders create a resolution requiring action to be taken to stave off address exhaustion, then anything is possible.

Disclaimer: I've worked with ARIN to get/manage/return blocks of IPs for years.

Re:Reshuffle existing IPv4 space

[Kadin2048]

It only seems ridiculous because of the way we distribute IP addresses today, using CIDR. Prior to 1993 (or whenever CIDR was implemented), if you wanted to run a network with subnets, then you needed at least a Class B allocation, so that your subnets could have Class C blocks (254 hosts each).

This is why MIT, Apple, DEC, IBM, and lots of other big companies were given Class A's. It wasn't just a "thanks for playing" reward, it was because the original design for the IP system required Class A blocks if you wanted to run big networks: if you had a big organization, you needed a Class A, in order to do multiple levels of subnetting.

When you look at the IP allocations and see GE or DEC's Class A blocks, it seems ridiculous. But you have to understand that when those allocations were made, what they were looking at was less the number of actual host IPs in the block (which is what we care about now) but the number of Class B and C subnet blocks that were inside. Put yourself in the shoes of someone at a big company like IBM or GE, with lots of regional offices. Each region/office needs to have a network, with its own subnets (for each department or whatever). That's how they were laying things out. "IBM" as an organization gets a Class A. Each regional office or some other division, Class B. Each network or further subdivision, Class C. Yeah, you end up with a lot of wasted capacity, but this whole scheme was designed back when a "host" was a PDP or VAX; there just weren't enough of them for it to seem like a major issue.

The problem people sometimes refer to when they talk about "the last time we were running out of IPs" (back in the early 90s) wasn't really a shortage of IPs at all (well, at least not immediately, although people were definitely realizing it was going to be a problem), it was a shortage of Class B and C subnet blocks. (Particularly Class B's, since that's what medium-size businesses and .edu's really wanted, and there are only like 16k of them around for direct allocation.)

So that's when CIDR was introduced, and it ended the whole 'Classed Network' concept (A, B, and C classes) and replaced it with the now-familiar bitwise/subnet-mask format. (E.g., IBM's Class A block is 9.0.0.0/8, Apple's is 17.0.0.0/8, etc.) This, along with prefix aggregation, allowed more efficient address allocation, and kept the routing tables from growing out of control. Now that you can subnet at the bit level, rather than at the Class level, those A Blocks seem huge. But keep in mind that before CIDR, each of those A Blocks was looked at, not as 16M hosts, but as 254 subnetworks.

It's only in retrospect, with the help of a bunch of new technologies, that the allocations made back in the Internet's early years look ridiculous.

Easy way to speed IPv6 Adoption

[Anonymous Coward]

Just move slashdot to an IPv6 only address; voilla by monday every corporate will have a functioning IPv6 setup... ;-)