PD configuration proposal

[Tomek Mrugalski]

The purpose of this thread is to have a quick discussion about PD
configuration. Here's how you configure IPv6 addresses:

> config add Dhcp6/subnet6
> config set Dhcp6/subnet6[0]/subnet "2001:db8:1::/48"
> config set Dhcp6/subnet6[0]/pool [ "2001:db8:1::/64" ]
> config commit

Please note that for addresses, it is ok to have the range specified as
either min-max or prefix/len. For prefixes it must start and end exactly
on prefix boundaries, so only prefix/len notation is valid.

PD configuration should look as similar as possible. There is one extra
parameter needed for PD: delegated prefix length. For example the
following configuration:

2001:db8:1::/56 delegated-length=64 means that the pool (/56) is to be
split into /64 prefixes (there will be 256 of them).

Initially I thought that we will support only one specific prefix length
for a given subnet, but the way I implemented CfgMgr allows specifying
several PD pools for a given subnet, with each pool having different length.

Here's the proposal:

The aforementioned example can be expanded to also have support for PD.
It may look like this:

> config set Dhcp6/subnet6[0]/pd-pool ["2001:db8:2::/56 64"]

Please note that pool (and pd-pool) is a coma separated list of pools,
so there may be more than one.

I'm not sure how to define the extra piece of information (delegated
prefix length), but adding it as a text after the pool seems like the
most similar way to what we already have.

So the PD only configuration would look like this:

> config add Dhcp6/subnet6
> config set Dhcp6/subnet6[0]/subnet "2001:db8:1::/48"
> config set Dhcp6/subnet6[0]/pd-pool [ "2001:db8:2::/56 64" ]
> config commit

And address + prefix configuration would look like this:

> config add Dhcp6/subnet6
> config set Dhcp6/subnet6[0]/subnet "2001:db8:1::/48"
> config set Dhcp6/subnet6[0]/pool [ "2001:db8:1::/64" ]
> config set Dhcp6/subnet6[0]/pd-pool [ "2001:db8:2::/56 64" ]
> config commit

This approach has a benefit of being simple and reasonably quick to
implement. The minor drawback of this approach is that t1, t2, preferred
and valid lifetimes specified for a subnet will be shared for both
address and prefix pool. This is not a big deal, especially that
draft-ietf-dhc-dhcpv6-stateful-issues-04 in section 4.3 recommends that
all T1, T2 values must be set to the same value.

What do you think? Does this seems reasonable?

Tomek