2

I am wondering if all the private IPv6 address non overlapping. And if yes, I am wondering if RD is required for IPv6 addresses.

When two customer sites of an ISP are using same overlapping private IPv4 space, RD is used to identify the specific VPN. If an ISP assigns blocks of non overlapping IPv6 to all the customers, is RD required at all since all the addresses will be already unique?

Will MP-BGP multiprotocol Reachable NLRI ever have AFI of 2 (IPv6) and SAFI of 128(NLRI for labeled VPN forwarding)?

2 Answers 2

1

I am wondering if all the private IPv6 address non overlapping.

The closest non-deprecated thing in the IPv6 world to private IPv4 space are "unique local addresses".

The current reccomended scheme for assigning unique local addresses to sites is "probabalisticaly unique" for small groups of sites however.

  1. Some people are sure to ignore the recommendations and pick addresses that look nice or are easy to remember.
  2. Even if people follow the RFC the chance of two sites having colliding addresses increases substantially with the number of sites under consideration.

If an ISP assigns blocks of non overlapping IPv6 to all the customers,

Running your internal network on ISP-assigned addresses is generally a bad idea. The reasonable choices for internal traffic are unique local addresses or provider-independent addresses.

is RD required at all since all the addresses will be already unique?

I'm not at all convinced that there won't be customers with overlapping IPv6 addresses on their internal networks. Furthermore I think even if there are no address overlaps there will still be a desire to keep traffic belonging to different customers separate.

Firstly because it's less error prone to tag everything with a customer than to maintain a massive central list of what addresses belong to what customer.

Secondly because you will get situations where two customers do want to talk to each other but they want to route that traffic through a central location where it can be firewalled/monitored, not just have it hop across by the shortest path.

1
  • That clarifies lots of things for me. Thanks Peter!
    – monica
    Apr 19, 2017 at 0:19
1

IPv6 doesn't have NAT, so ISPs assign Global IPv6 addresses to their customers. That was the whole point of IPv6. IPv4 ran out of public addresses, and NAT was a kludge created as a way to extend IPv4 until IPv6, with its unimaginable number of addresses, becomes ubiquitous. NAT breaks the IP promise of a unique address for every host that allows end-to-end connectivity. IPv6 restores end-to-end connectivity.

IPv6 does have ULA (Unique Local Addressing), but that is for addresses that will never be routed on the public Internet. ULA does have some restrictions in place that make it difficult to duplicate addressing at different sites.

4
  • does that mean IPv6 does not require Route distinguisher?
    – monica
    Apr 17, 2017 at 21:36
  • An RD keeps the network traffic of different customers separate. This will not change with IPv6.
    – Ron Maupin
    Apr 17, 2017 at 21:37
  • 1
    The purists don't like it but that hasn't prevented people from implementing IPv6 NAT. Apr 20, 2017 at 15:07
  • @PeterGreen, right, but there is no RFC standard for NAT on IPv6 like there is for IPv4. NAT actually breaks some IPv6 features. There is an experimental RFC for NAT on IPv6, but it is only a one-to-one NAT, and it specifically disallows NAPT.
    – Ron Maupin
    Apr 20, 2017 at 15:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.