I think this is going to be another y2k thing... Most people won't even notice a difference. Yes some things will be affected, but I doubt the whole world needs to worry about it..
There are some big differences between this situation and y2k, and some similarities at the same time. The y2k thing was a software bug, it was hyped, and no one was quite sure what would happen because of a date rollover. This is being hyped yes, but the difference here is, we know for a fact this is coming.
In 2005 Tony Hain (Cisco Systems) predicted a date around here, while Geoff Huston (
www.potaroo.net) at that point was still talking about 2015 - 2018. Slowly, over the, the figures started to come closer and closer as to when we would run out. Tony seems to have been correct.
The simple fact is, the IPv4 address space *IS* running out, and unlike the Y2K bug where its some bug that may or may not affect various systems, this is something that was a.) predictable b.) something that cannot be denied or ignored.
Less than a month ago the remaining IPv4 address space in the central registry (controlled by IANA) was at around 5%, with the previous 5% having been used up in the space for around 10 months, since then, we've dropped another 3 odd percent. When we hit 5 remaining /8s left in the pool, those 5 blocks are automatically allocated under current policy, with one each of the 5 going to each of the regional registries (afrinic, ripe, lacnic, apnic, apan). At that point IANA runs out. The question as to how long before the regional registries then run out of space to allocate depends on their allocation rates. This could be anywhere from 3 months on the inside to 12 months on the outside.
So what happens when the RIR's can't allocate IPv4 space anymore? One of a couple of possibilities. Either, the companies needing IPv4 space decide to forget IPv4 and go IPv6 only (I'll talk about this in a second), or they attempt to buy IPv4 space through the backdoor (and money is already being offered in vast amounts for space this way, but there are problems with this as well that I will get to), or they attempt to further NAT their space (This isn't practical, and the rates of allocations over the last few years show that the growth of the internet is faster than can be dealt with via NAT, and NAT also breaks many modern day protocols).
So, lets look at the three options mentioned above.
If a site goes IPv6 only, anyone who isn't ready for IPv6 won't be able to reach the site without going through some kind of NAT or translation mechanism, again, this could break many things. This means that IPv4 only providers are going to start getting calls from their customers looking for IPv6, if they aren't ready, their customers will move to someone who is ready. The problem is that for to long, providers have looked at IPv6 as "Revenue Generation", instead of looking at it as "Revenue Retention".
That moves to the next option, the black market sale of IPv4 addresses. There are moves afoot inside the IETF and other places to stop people announcing space that doesn't belong to them. I've been reading some of the documents today on this, and personally, I'm not convinced that in a world of shortage of IPv4 addresses, these mechanisms will actually get adopted. After all, there have been talk about proper filtering of IPv4 space in BGP for years, and while its come some way to getting there, its far far from actually perfect, so many don't implement it at all. I personally believe the black market for IPv4 space is going to arrive, UNLESS the RIR's do something to legitimize the trade in IPv4 addresses (Something I have heard one of the RIR's is looking at, though I can't confirm that, since it is something I only heard the other day and haven't investigated). I do however know of people who have been offered up to a million dollars for a /16 worth of IPv4 space (64k IP addresses).
The next option is to NAT, if you NAT, your end to end connectivity goes for a ball of chalk, and what we're talking about isn't actually NAT that would have to happen, it would be PAT (Port Address translation). If you're doing PAT on a large scale a.) you have scalability issues with translating the number of packets involved, since high bandwidth connections can be moving millions of packets per second b.) Things like torrents and other mechanisms which require some end to end break c.) Other applications that are stateless in nature that require end to end also start to have problems (Good luck doing large scale NAT of VoIP type applications)
So, all in all, if you eliminate options 2 and 3, you're back at option 1, the ISPs need to wake up and do something about IPv6, fast.
Anyway, these are just my thoughts.
Andrew
(Secondary thought, the TENET network is 100% IPv6 ready, the situation with campuses themselves being IPv6 ready however is an entirely other matter. Currently there is active IPv6 work happening on a few of them, but its far from all of them)
