[afripv6-discuss] Are AfriNic's /48 being filtered?

[Leo Vegoda]

Mark,

On 27 Aug 2007, at 11:52, Mark J Elkins wrote:

[...]

>>> This is the reason why I always insisted that PI should be /32.
>>>
>>> Many people will IGNORE what AfriNIC or other RIRs tell them and  
>>> will
>>> keep
>>> filtering anything longer than /32.
>>>
>>> Nice to have non-reachable critical infrastructures !
>>>
>>> In my opinion we should still consider a change in the policy to  
>>> allow
>>> AfriNIC to assign a /32 in case is proven, as in this case, than the
>>> /48 is
>>> getting filtered.
>>
>> You raise an interesting point. Is it your opinion that AfriNIC (or
>> any RIR) should guarantee routability? If so, to what proportion of
>> inter-connected networks should it guarantee routability and how
>> should such routability be measured? How would you document such a
>> guarantee in a policy?
> No one can guarantee routability.

Except over their own network.

> The question should be - can an RIR guarantee routability/propagation
> of a /48 to be the same as that of a /32.
> The answer is No. What then is needed is more education or awareness?
> Perhaps this needs to be the topic of a talk at RIR meetings? Perhaps
> the blocks need to be spoken about on RIR mailing lists - with  
> suggested
> (generic/cisco) configs for allowing them through.

I think it is more complicated than that.

There are two issues:

- Do providers want to allow /48 PI prefixes?
- How long is the filter-update cycle?

I can't speak on the first point and I believe it is worth comparing  
with new /8 IPv4 allocations for the second. It looks like  
reachability issues for addresses in 69.0.0.0/8, which was assigned  
by IANA in 2002, continue into 2007. See:

http://69box.atlantic.net/

I am hoping that the RIPE NCC may be able to present some research on  
this at RIPE 55. In the meantime, the graph shown at:

http://www.ris.ripe.net/debogon/

is informative.

Regards,

Leo Vegoda