[ppml] IPv6 addressing, allocation, and subnets

[Brian Dickson]

michael.dillon at bt.com wrote:
>> It is time to get with the CIDR program, not to advocate the 
>> antiquated class-ful paradigm.
>>     
>
> CIDR is IPv4. We are discussing IPv6 which is different.
No, I'm afraid you're wrong on this point.

CIDR is classless inter-domain routing. It is neither IPv4 nor IPv6.

The basic concept underpinning CIDR is, longest match wins when 
comparing two prefixes.
Without it, CIDR would not have worked, or not have scaled.
With it, covering aggregates and variable-length subnet masks (VLSM) are 
possible, and provide the basic tools for scaling both internal and 
external routing domains.

And, as it happens, IPv6 *is* CIDR-oriented.

The primary RFC for IPv6 specifically excludes any requirement for 
specific structure within IPv6 addresses, when viewed by routers. 
(Currently this is RFC 4291.)
It does anticipate that routers will have knowledge of hierarchical 
routing schemes (examples would include OSPF or ISIS, with stub or 
not-so-stubby areas).

And, while IPv6 supports the notion of fixed-size interface identifier 
(II) (currently fixed at a size of 64 bits for global unicast prefixes), 
the II is used for link-local addressing and node address 
autoconfiguration only.
Static host address configuration, and DHCPv6, support prefixes of 
arbitrary length, and are not restricted to /64 prefixes.

So, your contention that IPv6 is "different", and not CIDR, is patently 
false.

(Link-local addressing is a clever alternative to ARP, and one of the 
better elements of IPv6. However, IPv6 does not require that all IPv6 
addresses on an interface use the EUI-64 value as the host part of the 
address; it does specify that when doing autoconfiguration, the 
interface identifier is used for the host portion.)

Brian