On Fri, Nov 18, 2005 at 09:16:42AM -0500, Thomas Narten wrote:
The original (and still compelling, IMO) rational behind the current policy is to ensure good aggregation for the long term, in order to keep the number of distinct prefixes in the DFZ manageable. Hence, the current policy is oriented not towards _all_ ISPs, but those who will (hopefully) have _many_ customers that they assign addresses too. That is, they aggregate addresses for _many_ customers.
And then we can discuss the definition of "many" for hours. I "do" the backbone at the danish NREN. We have no upstream provider from whom to ask for IP addresses and we almost completely employ two /16s and one /17 of IPv4 space. There's a certain pressure on NRENs to get going on IPv6, we were involved in the EU 6net project and now run IPv6 on the backbone, exerting what little pressure we can to get services running around the country. We have a bit more than 100 institutions. About six of those are major universities with a large number of institutes. We planned (and still plan) to get everybody running IPv6 and by setting a policy that every institute should have a /48, we were able to claim without exactly lying that we were planning at least 200 /48s. Of course, what makes sense is for a university to get a /48 and then let them subdivide that for institutes, but that would effectively make Forskningsnettet (the danish research network) incapable of obtaining IPv6 at all. Executive summary: I believe that the 200 customers demand halts IPv6 development rather than help it. If we really wanted to be cautious, we should forget about /32s and go /36 or /40 instead. A /40 would be plenty for Forskningsnettet. Now we have /32 and a completely absurd subassignment policy instead. Peter B. Juul, Uni·C (PBJ255-RIPE)