[address-policy-wg] IPv6 allocations for 6RD

Remco van Mook wrote:
> Michael
>
> key words here are 'average' and 'generous'. Perhaps you should try to
> read first. Same goes for your statement of counting addresses. In cases
> like these, I think orders of magnitude are more interesting than
> narrowly defined numbers. 4 million is roughly 0.1% of the IPv4 pool.
> Since I think the AVERAGE ISP is probably smaller than that (either in
> customer count or allocated space), giving all of them 32 bits (or the
> full IPv4 space) PLUS 4-8 bits of network space (not host space!) for
> their 6rd plan sounds wasteful to me. 
>
> The 6rd draft has prefix compression for a reason. Sparse number spaces
> for expansion and automation are fine. Very sparse number spaces you
> know from the outset will never be filled or usable for other purposes,
> are not.
>   
Actually, the main reason we added this is for easy and obvious 
compression for when an SP has decided to run NAT. The RFC1918 space is 
nicely aggregated. Please don't give SPs more reasons to deploy NATted 
IPv4 service.

Most SPs I talk to don't really like the idea of compressing the global 
space, if nothing else because they have no assurance that the global 
IPv4 space they have now will be the same in the future. They'd sooner 
give the customer a /64, which I've already stated why I personally 
think is dangerous.
> That is why deploying multiple instances of 6rd in order to benefit from
> prefix compression to collapse that  required number space without
> losing functionality sounds like the right direction for me.
>   
There is nothing prohibiting multiple instances of 6rd in the current 
version of the draft, however note that this does not come without 
operational complexity - which is exactly what 6rd is trying to allow an 
SP to avoid. So, there's a cost tradeoff here. Increased prudence of 
IPv6 space vs. increased cost for ISPs to deploy IPv6. It's quite easy 
to be penny-wise and pound-foolish here, and given the rate of native 
deployment of IPv6 thus far, I'd err on the side of getting IPv6 out the 
door in the short term until we are sure it no longer needs a boost. 
Consider it similar to stimulus funds to counter a depression if you 
must, but if an ISP which would otherwise qualify for a /30 needs a /28 
or /26 to get IPv6 deployed now vs. later, I'd choose handing out a few 
more bits until we no longer think it necessary for native IPv6 adoption 
to occur.

- Mark