[address-policy-wg] 2023-04 Review Phase (Add AGGREGATED-BY-LIR status for IPv4 PA assignments)

On Wed, 2023-11-29 at 12:28 +0100, Peter Hessler wrote:

> I mentioned this during the WG session but want to bring it up on the
> mailing list, what is the definition of assignment-size.
> 
> In the IPv6 implementation of AGGREGATED-BY-LIR there is an
> assignment-size attribute, which is proposed to be optional for the IPv4
> version.
> 
> From the inet6num documentation:
>  "assignment-size:" - This specifies the common size of all
>  individual assignments aggregated into one block with the status
>  'AGGREGATED-BY-LIR'. This attribute is required to be present if the
>  inet6num object has this status. The individual assignments do not need
>  to be represented in the RIPE Database. But one or more assignments may
>  be included if the member wishes to specify them for any reason.
> 
> While there is no definition of what "size" means, my understanding is
> that the technical implementation follows the implemention requirements
> of inet6num objects, which is the CIDR size.
> 
> However, in IPv4 it is allowed to do CIDR _or_ an arbitrary range of
> addresses.  So if one were to create this inetnum object:
> 
>   inetnum: 192.0.2.0 - 192.0.2.255
>   netname: customers of example
>   status: AGGREGATED-BY-LIR
>   assignment-size: 32
>   ...
> 
> Here the assignment size is ambiguous.  Is it 32 addresses aka a /27, or
> is it a /32 aka 1 address.
> 
> I propose that we define this to be a CIDR assignment and they put in
> the number of bits of the netmask, so the above example would be
> assignment-size of /32.

Hi Peter,

Agreed 100%, that is the intuitive understanding, especially
considering that it is already in use like that for inet6num.

I propose that we simply ask the RIPE NCC (hello Angela!) to confirm
that their intended implementation of assignment-size for IPv4 inetnum
is a CIDR prefix length. If it is, and there are no opposing voices
against defining it in that way, I believe there should be no need to
do a v3 of the proposal just to state this explicitly.

(I also suggest that while the NCC is at it, they should document
assignment-size as being a CIDR prefix length for inet6num as well. As
you point the formal definition is not crystal clear there either,
which was a bit of a surprise to me, honestly. But all current inet6num
objects have an assignment-size within the range 31-128, so I guess
LIRs simply get it - unless the database software enforces it by
rejecting updates with values above 128.)

Tore