Currently, a typical home Internet user will be assigned one IP address from their ISP, and then use NAT (Network Address Translation) to share the Internet connection among all their computers. The IP address assigned by your ISP is dynamic, and that is not a problem for the average home user, or even your typical power user.
Setting static IPs on computers is not all that uncommon, even among home users, excluding only the very most technically-illiterate ones. For example, your home router might be 10.0.0.1, and the other desktops in your house might be 10.0.0.10, 10.0.0.11, and so on. Then, if somebody drops by and wants to use your WiFi, they might be assigned an address via DHCP, such as 10.0.0.121.
This won’t work in IPv6 if, and only if, ISPs choose to make your subnet allocation dynamic. I urge ISPs to assign static IPv6 subnets to all their customers.
Why? Well, let me give my reasons. In IPv4, all the home machines in the above example are behind a NAT. This means the private IP address (10.0.0.121) gets dynamically translated to your public IP address (184.108.40.206).
Because of the absence of NAT in IPv6, this can’t happen! Your machine’s IPv6 address is tied to the subnet allocated to you by the ISP. And if your ISP changes your subnet every time you connect to the Internet as they currently do with IPv4, your static IPs will break horrendously.
I am aware of site-local and unique local addresses. These addresses are designed to be used only in a local situation, and not routed to the Internet. In theory, this could solve the problem, by allowing you to have a static local address, and a dynamic global address. In practice, this will not work because:
- Site-local addresses have been deprecated by RFC 3879.
- Unique local addresses are considered to be global addresses by current OSes. Wikipedia says that “despite the restricted, local usage of these addresses, they have a global address scope”, which means that your computer will assume either one can get to the Internet.
- Thus, your source IP may be wrong, and your packet may be filtered and rejected by your ISP, or you may never get a reply, as the message won’t be able to get back to you.
- Having both unique local and global addresses are confusing to the end-user, unlike link-local addresses, which are clearly marked as such, and are generally non-routable.
Finally, we must look at the reason why dynamic IPv4 addresses are assigned in the first place. I believe the main reason for this is to conserve space. With most of their address space used up, ISPs would have to count on all of their customers not using their Internet connections at the same time. Dynamic IP addresses means they can effectively over-subscribe their puny IP allocations.
In IPv6, this is not necessary. ISPs typically get a /32 allocation, which if you’re not familiar with CIDR notation, is bleeping huge! With a /32 allocation, an ISP could allocate more than 4 billion /64 subnets (which are suitable for a typical household) to each of their customers. I don’t think any ISP in the world has 4 billion customers, and if they did, they could get a /31 allocation, which would give them about 8 billion /64 subnets. Plenty of space for static allocations for everyone!
In conclusion, I’d like to summarise what I’ve been trying to bring out:
- People that like to set static IPs on their machines will have them break if their subnet changes.
- Site-local and unique local addresses only add to the problem, not solve it.
- There is enough IPv6 address space in a /32 for everybody to have a static subnet.
- There is no business advantage in giving out dynamic subnets. Do the best thing by your customers and go static.
So, dear ISPs of the world, please make static IPv6 subnets a part of your standard offering — not a “paid upgrade” or anything silly like that. It might work in the NAT’ed world of IPv4, but you will do your IPv6 customers a disservice.
Thanks for reading.