Moving to IPv6
According to Pawlik, that day is going to arrive very soon. 'For the Asia and Pacific region it has already started,'
he says, 'and for our region – that's Europe and certain parts of Asia - it's probably going to happen in the second half of this year. It's difficult to forecast, because the incoming requests are likely to accelerate a bit. We can't really say what date that will be, but it's likely to be in the second half of this year. Then people will come to us and say they'd like IPv4 addresses, and we'll have to say: "Sorry, you're too late – there are none left for you."'
So what is IPv6, and what will be involved in implementing it? 'The basic ideas is to pre-empt this problem coming up again,'
says Pawlik, 'we're seeing lots of mobile devices on the Internet now, and it's just been growing so fast that we need to be prepared.'
With a 128-bit system, we're talking about a ridiculous number of addresses – around 340 undecillion in fact – a number so large that it wasn't even recognised by the Word spell-check.
Pawlik likens IPv6 addresses to grains of sand - there might be undecillions of
them, but that doesn't mean you can dish out whole beaches to everyone
Even so, Pawlik says that RIPE NCC isn't just going to hand out IP addresses on platters. 'There is no immediate replacement for IPv6 on the horizon,'
he says, 'so we need to be careful with how we allocate IPv6 addresses.'
He likens an IPv6 address to a grain of sand. 'We can't just assume that we can give out whole beaches at a time,'
he points out, 'just because there are so many sand grains on it – then we will run out as well.'
The development process of IPv6 apparently hasn't been completely smooth, either. 'A few of the engineers with the biggest mouths were going for a "beautiful design" - meaning a pure design,'
says Pawlik. 'This is where the address field is always the same length, which is good for the routers - we can build that stuff relatively efficiently in hardware, and move routing packets around.
'However, another option would have been to make IPv6 addresses of variable length, and then the IPv4 addresses could just be a subset of the IPv6 addresses, so they would be compatible. But that would be a bit more complicated with the routers and the computations, so they said we don't want to do this.
'In the end they all said we'll run IPv4 and IPv6 in parallel – it's a beautiful solution with a happy ending. Of course, we know that never happened, because nobody wanted to invest much in IPv6 – in terms of manpower or machinery – and the customers weren't asking for it. So we have a situation now where we should have migrated to IPv6 a long time ago, but we're still busy with it.'
Routers like address fields that are all the same length
Sorting out the shortage of IP addresses is an all-or-nothing scenario for IPv6 according to Pawlik – we basically need to forget about IPv4 as soon as possible, and move everything over to IPv6. 'There are now engineering solutions that translate between IPv4 and IPv6,'
he notes, 'and they work relatively well, but they add more complexity in the network – engineers have to look after it, keep it running and maintain it.'
He also points out that other possible solutions to the problem have been suggested, but that these are impractical to implement in the long-term for similar reasons. 'Two years ago people were saying IPv6 will never really happen,'
he recalls, 'they said we'll go on with Carrier-grade NAT and all those things, and yes that's possible, but it's also expensive and at some point it will fall over. IPv6 brings back the idea of interconnection and openness – anybody can speak with anybody – if you devise a cool new service want to use to communicate with someone over the Internet, you can do it because it's open – we'll have public addresses and it should be fine.'