| 1 |
On 30/07/2020 12:13, Remco Rijnders wrote: |
| 2 |
> An IPv6 address is 128 bits in length. Usually an ISP allocates 64 |
| 3 |
> bits to a single customer, allowing the systems on/behind that |
| 4 |
> connection to automatically assign themselves an address based on |
| 5 |
> their MAC address for example. Note that also allocations bigger than |
| 6 |
> 64 bits are common so customers get 70 or 76 bits to use and can use |
| 7 |
> multiple subnets on their home/business networks. |
| 8 |
|
| 9 |
I don't think an ISP is supposed to allocate less ... |
| 10 |
|
| 11 |
As I understood it, the first 64 bits are the "network address", ie |
| 12 |
sort-of assigned to the edge router, and the remaining 64 bits are |
| 13 |
assigned by the network operator. |
| 14 |
|
| 15 |
So in your scenario of customers getting more bits, they are effectively |
| 16 |
being assigned 2^6 or 2^12 network addresses. Exactly the scenario |
| 17 |
planned for high-level ISPs parcelling out address space to low-level ISPs. |
| 18 |
|
| 19 |
And looking at the wikipedia page, it looks like the ISP *must* allocate |
| 20 |
at least a /64, because the spec says each device allocates itself a |
| 21 |
least-significant-64 address at random using a collision-detect |
| 22 |
protocol. Which is why many simplistic algorithms include the MAC |
| 23 |
address to (try to) guarantee a unique address on the first attempt. |
| 24 |
|
| 25 |
Cheers, |
| 26 |
Wol |
Archives