IPv6 Prefix Length

[taichimaster]

I have a question about IPv6 address and its prefix length that I don't quite fully understand.

As far as I understand, a typical IPv6 node address consist of a prefix and an interface identifier (total 128 bits).  The prefix is the part of the address where the bits have fixed values or are the bits that define a subnet.  RFC 4291 (Section 2.5.1) also says that all IPv6 unicast addresses are required to have a 64-bit interface identifier, so that means subnet prefix (including the global routing prefix) is going to be 64-bit wide.

Any prefix that is less than 64 bits refers to a summarized route or an address range that is summarizing a portion of the IPv6 address space but not an actual node address.

If all of the above is correct understanding, then how come when we assign an IPv6 address to an interface/adapter (via ifconfig or netsh), we could specify a prefix length?  Isn't it always going to be 64?  When comparing if 2 IPv6 addresses are in the same network, do we always match the first 64-bit of the 2 addresses to see if they have the same values?

What's the meaning when someone assigns a node address with a non-64 prefix?

Thanks.

[cholzhauer]

You can assign prefix lengths other than /64.

Like you said, /64 is what most OS's expect and the length you should use, but you do have the ability to change it.

I've seen some people use a /126 for router intefaces