On May 31, 2006, at 7:08 AM, Thomas Narten wrote:
That is, given /48s to end sites, stateless address autoconfiguration and 64-bit interface identifiers, do we undermine our own credibility by claiming IPv6 supports 10^38 addresses?
as opposed to 10^128?Stateless autoconfiguration provides a way that any system can get an address. It reduces the count of LANs to 10^64, and the business about /48s means we reduce the number of sites on the backbone to 10^32 ISPs each supporting 2^16 customers. I would prefer that we talked about /48s for campuses and /56s for SOHO environments, and I thought that was on the table at least at one point (wasn't that draft-narten-ipv6-3177bis-48boundary?).
Frankly, the requirement stated that we needed to support 10^12 LANs and 10^15 hosts. We passed that a long time back.