[ipv6-wg at ripe.net] New Document available: RIPE-343

On 23-feb-05, at 14:14, Gert Doering wrote:

>> 1) Which parts of the community rejected the doc in its previous
>>    incarnation and why?

> It was mostly disliked because of the use of a global common address
> pool, which means that you give up any chance to be able to 
> filter/aggregate
> on region boundaries ("why do I need to know any details about ASes
> located outside my region?") - whether or not someone is doing this 
> today
> doesn't matter, but it was felt that it shouldn't be made impossible
> right from the start.

I'm sorry, but I have to object here.

The notion that divvying up the globe into four or five parts in order 
to save on routing table expenditures in nonsense. Either filtering out 
"far away" information is a good idea, and then we should do it right, 
or it isn't, and then we don't need to do it at all.

By doing it right I mean putting several layers of geographical 
hierarchy into addresses. In Europe, this would mean at least the 
country level, in large countries like the US, China and India the 
state/province level.

>> 2) Why is IANA not a co-author - only the RIRs?

> ICANN/IANA has been very passive and very resistive regarding any 
> attempts
> to make the IANA->RIR allocations more reasonable.  For many years now.

They've been hesitant to change things in the absence of a new policy. 
Now that the IANA is more closely related to the ICANN they're probably 
getting an overdose of lawyer exposure...

About the new document: this is unnecessarily complex, and it suffers 
from terrible end-game properties: when the space starts to run out 
(which isn't as unthinkable as it seems with the H/R thing, especially 
if we end up with several new RIRs and/or a middle layer of NIRs) the 
space will be fragmented very badly.

Just give RIRs tightly stacked /12 allocations, or at most /10s. That's 
2048 - 8192 times more than they were getting until recently. If this 
turns out unworkably small at less than 1000 IPv6 ASes then there is a 
bug in the system somewhere that we'd better fix before the other 17000 
ASes start doing v6 too.