Speaking of Comcast, the cable giant is offering an interesting proposal to the standards community concerning the long overdue transition from IPv4 to IPv6, using NATs and tunnels:
Comcast is upgrading its networks from IPv4, the Internet’s main communications protocol, to the standard known as IPv6. IPv4 uses 32-bit addresses and can support 4.3 billion devices connected directly to the Internet. IPv6 uses 128-bit addresses and supports an unlimited number of devices.
At issue is how Comcast will support new customers when IPv4 addresses run out, which is expected in 2011. Comcast can give these customers IPv6 addresses, but their home computers, printers, gaming systems and other Internet-connected devices are likely to support only IPv4.
Comcast engineers have come up with a solution to this problem, dubbed Dual-Stack Lite, which it says is backwards compatible with IPv4 and can be deployed incrementally.
Comcast outlined Dual-Stack Lite in a draft document published by the Internet Engineering Task Force on July 7. Dual-Stack Lite will be discussed at an IETF meeting in Dublin scheduled for later this month.
It’s a reasonable approach, putting the onus of dual stacks on the carrier NATs and home gateways where it belongs. It’s fortunate the IETF has companies like Comcast to give it guidance.
H/T CircleID.
UPDATE: Iljitsch van Beijnum has some further illumination on the Ars Technica blog, without using the “C” word; they don’t go for that sort of thing on Ars.
Nodody wants to do dual-stack on the end-points and it defeats the entire purpose of going to a “clean IPv6” stack if you still have to run IPv4 and NAT. Creating a black box that transparently connects IPv4-only and IPv6-only networks makes a whole lot of sense.
Think of IPv4 addresses as “short codes” for the Internet where IPv6 addresses are the long codes. In that sense, IPv4 essentially becomes prime realestate because that’s where all the action is and that’s where the shorter addresses are at.
It sounds to me like the plan is to manage IPv4 addresses the same way that private IPs are managed by NATs today, and when all your devices are upgraded, you can turn the translating NAT into a simple firewall. Decapsulating and encapsulating is something we do today at the boundary between Ethernet and Wi-Fi, and we have chips to make it very simple.
I should also point out that Businesses and Individuals actually LIKE NAT. Businesses like the fact that their internal IP addressing scheme is hidden from the outside world and they like the fact that they have control on the network. Some of them even like to operate at the application layer on their firewall for security reasons.
Home users like the fact that their ISP doesn’t know how many computers are using the Internet and it wasn’t long ago that ISPs charged an extra $5/user/IP. Home users also like the partial anonymity and plausible deniability they have when the RIAA comes a knocking because they’ll just say “but judge, I operate a free hotspot service with my Wi-Fi router and I have no idea who is downloading or uploading this song”.
The downside of NAT is the port-mapping difficulty for internal services, but it’s not insurmountable.
IPv6 uses 128-bit addresses and supports an unlimited number of devices
Not to be a math pedant, but can we please say something like much larger? It may seem unlimited, but I’ve got grandkids.