IPv6 Subnet Allocation

[cdnjay]

Haha, you're probably right. It just seems weird to me this notion that a SLAAC address can be anything within a /64 and yet some static addresses are still required (like for the LAN interface on the router) and yet there's no range of addresses in a SLAAC /64 reserved for that. Like you say though, the odds of a collision are pretty much non-existent with there being about 18 Quintillion addresses in a /64.

[mattwilson9090]

Why would it be weird that some devices, especially a router in an IPv6 network need a static address? It's exactly the same in an IPv4 network, the router needs to have a static address.

Why would a range of IPv6 addresses need to be reserved for static addresses? Addresses in IPv4 aren't reserved that way as part of the spec.

Have you ever setup an IPv4 network and handled address assignments through a combination of static and dynamic addresses? The basic concepts aren't really different with IPv6, although with the massively larger numbers of addresses available to work with, a lot more options are opened up.

The biggest difference is that DHCPv6 can no longer provide all of the information that DHCPv4 could, especially the gateway address, means that RA (router advertisement) is going to play a role as well unless you statically assign everything.

[cdnjay]

I find it weird because you don't define a range for SLAAC to use like you do with DHCPv4, it can just use anything in the /64 which makes it difficult to then guarantee that there will never be a conflict with a static IP in the same /64. It would just make more sense to me if there was something saying EUI-64 and the privacy extensions will use something within a specific range and we just need to make sure our static addresses are outside of that. As I mentioned earlier I'm pretty sure something where the first 3 of the last 4 quads are all zeroes will never be used with SLAAC for MAC -> EUI-64 (and if it's used with privacy extensions the system will retry), just can't find anything confirming that.

[mattwilson9090]

Even with DHCPv4 you can still get a conflict with a static IP if someone assigns an addresses that is within the assigned range. There is no technological guarantee that there will never be a conflict. Both IPv4 and IPv6 have mechanisms in place for resolving a conflict like that if it takes place.

It doesn't really matter though. A single /64 has 2^32 more addresses than the entirety of all addresses in all of IPv4. The odds of you choosing a static address that is going to create a conflict with dynamically assigned addresses is so small as to effectively be zero.

Best to just stop overthinking all of this and to design your IPv6 network, including whatever standards you want in place for assigning static addreses.