Re: Review of draft-ietf-v6ops-nap-02.txt

[Fred Baker <fred@cisco.com>]

Havard Eidnes read the numbers a little closer than I did: yes, 10^38  
is approximately 2^128.

I think the point remains that the requirement originally stated was  
this:

   CRITERION
      The IPng Protocol must scale to allow the identification and
      addressing of at least 10**12 end systems (and preferably much
      more).  The IPng Protocol, and its associated routing protocols
      and architecture must allow for at least 10**9 individual  
networks
      (and preferably more).  The routing schemes must scale at a rate
      that is less than the square root of the number of constituent
      networks [10].

http://www.ietf.org/rfc/rfc1726.txt
1726 Technical Criteria for Choosing IP The Next Generation (IPng). C.
     Partridge, F. Kastenholz. December 1994. (Format: TXT=74109 bytes)
     (Status: INFORMATIONAL)

Without running through all the history, and the arguments of whether  
in 2006 the numbers selected in 1994 are adequate, the fact is that  
we support a very large number of enumerable networks, on the order  
of the square of the original requirement, and not only do we support  
10^12 hosts in the address space, we support that many and quite a  
few more on each LAN.

Yes, we support about 3 * 10^38 addresses in the address space. That  
is a mathematically accurate statement.

I think the point you're asking is "does that mean we should throw  
the addresses away with both hands?". And I would say, no, that's  
what we did with the IPv4 address space, and we eventually learned  
the error of those ways. It does make sense to think about rational- 
sized address blocks, but frankly I would expect that this is more of  
an RIR issue than an IETF issue. They are currently handing out /32s  
to ISPs, and more if the ISPs can justify it. That means that they  
are allowing for ~4 billion address allocations, each of which is  
presumably enough for 2^16 ISP customers, each of which potentially  
has 2^16 LANs. I do believe that there is a question of market  
dynamics there - if the customer is unlikely to create more than 256  
LANs, selling them a /56 makes sense, and helps with the overall  
address usage density, and if the customer really only needs a  
single /64 or to be a member of a /64, I think those should be  
options the customer can buy. saying "ISPs should only ever under any  
circumstances hand out a /48" seems a trifle over the top.

And I do remember this thing about CIDR, that it removed the classful  
address straight-jacket, which at least at the time some of us  
thought was a good thing, and I wonder why we seem to be re-creating  
it here. But that wasn't your question.