Who cares? If they gave them back to IANA tomorrow, IANA would be out again in 2 weeks, at the rate of demand they were seeing the last time they had any blocks to assign. It would probably last even less time, now.
Do the class A allocations suck in retrospect? Sure.
Is it worth flushing time, effort, and money on trying to recover them? Hell no.
Whatever amount of money it would cost Haliburton to restructure all of their internal networks and rewrite their applications to compactify their IPv4 usage would be far more efficiently spent if they simply used it to upgrade their infrastructure to IPv6.
Putting any money into IPv4 is just throwing good money after bad at this point.
I really hope people don't shit on the original engineers. A 32-bit address space probably seemed stupidly, uselessly huge at that point in time.
That the decisions they made at the time made sense doesn't make them any less regrettable in retrospect.
Either way, it's a pointless argument anyways. There is no point to fixing those allocations because if you could snap your fingers and efficiently allocate IPv4 address today, we'd still be out and still have to switch to IPv6.
They really couldn't have made any better decision at the time--TCP/IP dates to 1974!
Bits were expensive back then. Backbone links were 56K. The Intel 1103 1K DRAM was the biggest selling semiconductor in 1972 and had just replaced core memory.
The fact that they even allocated 32 bits (the wastrels!) was amazingly forward thinking.
Is there some confusion about what the phrase "in retrospect" means?
Time is linear and we gain perspective as we move forward in it. Often times good, smart seeming decisions, or even the only possible decision that was possible at a given point, become regrettable when we view them with the added perspective of seeing how things ultimately played out.
This isn't a condemnation of anyone; it's a simple, obvious observation: knowing what we know now, some of the decisions we made then have had unfortunate, regrettable consequences, however necessary they were are the time. C'est la vie.
IPv6 is old now. It dates back from 1998. Engineers understood two decades ago that ipv4 would run out and it finally made sense to create something better.
There is some spectacular failure of communication here.
What I am definitely not saying: they should have implemented IPv6 in 1978!
What you are apparently somehow reading: they should have implemented IPv6 in 1978!
Here's what I am saying: Some decisions made in 1978 have, with the perspective of time, had unfortunate, unforeseeable consequences.
This doesn't mean that IPv4 should have had a bloody 128 bit address space in 1978. Rather, it means that, with the perspective of time, we can see that some decisions made have really had dire consequences, and that what would have been counter-intuitive or less than optimal decisions at the time would have worked out better in the long run.
For instance, given the hardware limitations of the time, it would have worked out better in the long run if IPv4 had had a bit smaller address space. At 2^24, it would have lasted well into the early/mid-90s but needed to be replaced right around the huge growth spike, before anything got too entrenched, when there was ample opportunity to roll out a different addressing scheme.
2^32 has left us in an uncomfortable spot where it was big enough to have got us quite far and for us to have become incredibly dependent on it, but that means we also have really, really entrenched a lot of legacy hardware in everyone's homes and offices, and it's not big enough to allow have the world to not live behind carrier-grade NAT.
The usable range is 1.0.0.0 through 126.255.255.255 and 128.0.0.0 through 233.255.255.255. The rest is reserved for various reasons or used for other purposes. That nibbles about 10% of the address space away.
They could've imposed more limits this way, but that might've seemed too heavy handed and would just lead to severe hacks and work-arounds.
It's a Catch-22. Switching to an IPv6-like address space in the mid 1990s would've been easier, but if it caused a lot of turbulence, it would've slowed the adoption of the IP technology in the first place, meaning the conversion was less urgent.
I disagree. We can regret the way something worked out even if we had no other choice to make. That's a pretty normal human thing to do.
But we could also regret that, eg) the technical limitations didn't drive them to choose an even smaller address space, one that would have lasted just long enough to get us to a point where we could move to maybe a 2^64 space in the 90s, instead of a space just big enough to become incredible hard to move out of, but not big enough that half the world isn't now living behind carrier grade NAT.
Sure it can. Just ask everyone in India and China sitting behind three levels of Carrier-Grade NAT how great that experience is.
And hey, working out a rotational schedule as to who gets to have the XBox Live ports to themselves on what night of the month is a great way to meet your neighbours.
Keep in mind that a lot of these allocations were made in the 80s and 90s when nobody worried about running out of IPv4 addresses; in addition I saw what we paid to ARIN every year just to "have" 10.x and it was an insane amount (I remember it being more than my yearly salary).
So it is much better to focus on the implementation of IPv6.
The E block is a lost cause: http://packetlife.net/blog/2010/oct/14/ipv4-exhaustion-what-...
Note that Y in this case is significantly less than 1.
If you were a company with any degree of credibility, it wasn't too hard to petition for a Class A allocation, /8. Apple and MIT managed.
Then CIDR (https://en.wikipedia.org/wiki/Classless_Inter-Domain_Routing) came along and we had more than three sizes to offer, which really eased up on the crazy allocations.
All those /8 allocations created chaos in the routing tables. Now they're usually sub-allocations from larger blocks which makes the routing tables much more orderly. Anything in the 192.x.x.x range was once called "the swamp" since it was full of so many routes.
I agree that they should probably be returned to IANA.
Fact is that if you want a GLOBALLY UNIQUE IP address, then IPv6 is the ONLY way to get one. It has been over 20 years since IPv4 addresses ceased to be globally unique.
Pardon?
Just as any yahoo can choose to manually assign the IPv4 address assigned to my border router to a machine on his internal network, he can choose to do the same thing with its IPv6 address.
My border router has a globally unique IPv4 address and a globally unique IPv6 address, inasmuch as any IP address in globally-routable space can be globally unique.
https://en.wikipedia.org/wiki/List_of_assigned_/8_IPv4_addre...
Possibly 100+ million.
What the frack does Halliburton need with 16 million IP addresses?
On that note, US Department of Defense apparently does not just consume most of the US budget, it also consumes over 210 MILLION ip addresses.
There are also a quarter billion ip addresses in the E block, which is probably easier to reprogram all routers than get all consumer devices to support ipv6 properly.