[ppml] Effects of explosive routing table growth on ISP behavior

[Brian Dickson]

Scott Leibrand wrote:
> Brian Johnson wrote:
>   
>> Scott Leibrand wrote:
>>   
>>     
>>>> Question: Does it say you cannot advertise smaller portions as well as the larger block?
>>>>
>>>>       
>>>>         
>>> There's some debate on that, but as I read it, no.  My understanding is that you're free to advertise subnets longer than /32 out of PA space,
>>> but other networks are free to filter them.
>>>
>>> I think it may eventually be necessary to move toward something similar for IPv4, but it's not all that urgent just yet.
>>>
>>>     
>>>       
>> So I can announce subnets longer than /32 to my desire and anyone can
>> filter as they desire.
>>
>> Sounds like the IPv4 status quo to me.
>>     
>
> Almost, but with one important difference: you have to announce your 
> covering aggregate, which makes it safe to filter more-specifics without 
> affecting reachability.
>   
Here's a thought:

For any policy that requires usage of blocks assigned by $RIR, in 
addition to testing reachability of hosts in the block, require that the 
full block itself be seen in the DFZ (i.e. as a covering aggregate).

It puts a rather narrow interpretation of use (announcement) of the 
block, but rather elegantly addresses the covering aggregate issue.

Anyone announcing more specifics but not announcing their covering 
aggregate, would not satisfy the $RIR requirements for their assignment, 
and after some period of time, would be requested to return their 
"unused" block.

(I expect this won't fly, but may give other folks ideas about how to 
approach addressing the issue, policy-wise.)

Brian