Ca By wrote:
The proper number to be considered should be percentage of IPv6 hosts which can not communicate with IPv4 only hosts.
Isn't it 0%?
I think you agree with me, here.
For those of us running networks, especially growing networks, uniquely numbering hosts is our goal and ipv6 fits that task.
Then, you should be running some isolated network. In this thread, we, except you, are discussing how to uniquely identify customers, not hosts, without (much) logging.
For many networks, rfc1918 space is not sufficiently large to number end-points. Around the world, there are many networks that fit this.
The global address space of IPv4 with NAT is combination of IPv4 address and part of port number spaces, which should be enough to identify customers and, maybe, hosts. and is much larger than private space of rfc1918.
So far, i just talked about why eyeball networks deploy ipv6 — which is basic and sensible engineering and economics. A similar set of forces are at work on the content / cloud / iot side.
Perfect argument for OSI. Masataka Ohta