[arin-ppml] A modest proposal for IPv6 address allocations

Milton L Mueller mueller at syr.edu
Sat Jun 6 14:20:21 EDT 2009



> -----Original Message-----
> From: arin-ppml-bounces at arin.net [mailto:arin-ppml-bounces at arin.net] On
> Behalf Of michael.dillon at bt.com
> 
> It never talks about huge numbers of host addresses. And it
> does reference RFC 2374 which explains how interface ids are
> constructed which makes it clear that the minimum subnet
> size is a /64. 

Yes.... The discussion was about "need" if you recall. If the minimum subnet is /64 then an architectural decision has been made that everyone "needs" at least a /64. A very odd definition of need.

> No addresses are wasted. Some bit combinations
> are never used but that is not waste

Ah. I see. No irony detected here. 
I would call unused and structurally unusable bit combinations in the thousands of billions "waste;" perhaps too strong a word but then to say "nothing is wasted" might go a wee bit too far in the other direction.

> neither IPv4 or
> IPv6 is architected in a way that all bit combinations could
> be used for endpoint addresses.

That is true, and it is true of all address spaces even outside of IP. But there are different levels of structural inefficiency; to say that nothing is perfect does not mean that all such differences are irrelevant.

> > Making finer distinctions requires massive investments in
> > organizational overhead: reviews, monitoring, etc.
> 
> Precisely the justification for having standard prefix lengths
> of /64, /48 and /32 in the first place. It massively cuts
> down the overhead of having to make fine grained measurements

Yes, indeed, I am with you here 100%. 

> outgrows its address allocation. IPv4 wastes money, but IPv6
> leverages a much longer bitstring to cut out that waste in
> overhead, monitoring, reviewing and restructuring the network
> to accomodate growth.

Yes, but the unresolved problem/issue -- which is where I had hoped the discussion would begin, not end -- was about how to ration subnet claims. Recall again Leo's post about "what is a subnet," which you dismissed. 



More information about the ARIN-PPML mailing list