ned> Nor did I say you could. The point is that how IP addresses are
ned> used varies.
anthony> My point is that you want to uniquely identify every machine
anthony> in the world using IP, then every machine must have a unique
anthony> IP address. If you are using names instead of IP addresses,
anthony> and you still want to uniquely identify every machine in the
anthony> world, then every machine must have a unique name.
Anthony doesn't seem to get the difference between requirement for
service and implementation, along with not understanding a long list
of other things...
First, assuming that I (as a user of some service) must reach a
particular, unique machine is a geek wish, not a requirement. The
folks monitoring and maintaining the machines need to be able to get
unambiguously to that machine but that is a great use of 1918
space. Wasting public IP addresses one per machine when there is no
requirement seems to violate Anthony's claimed concern about address
space.
Second, there is no reason that, based on service requirements, I
can't choose to either use unique FQDNs *or* unique IP addresses (or
both) to meet the uniqueness requirement, assuming that really is
necessary.
ned> Why is this so hard for you to understand?
anthony> I understand it perfectly. What I am illustrating is how
anthony> poorly people understand the real problems, how careless they
anthony> are when reading, and how readily they confuse one problem
anthony> with another. This is why fixed address spaces will be
anthony> exhausted, no matter how large they are, and this is why TLD
anthony> management will continue to be a mess, no matter what changes
anthony> are made.
Anthony rails here about waste of fixed address space yet advocates
that in order for us to be "pure", we must waste an IP address on
every device that we need to identify as "unique" on the Internet. OK,
which do you want? Better use of address space or a dictated waste of
address space for a dubious technical "need"?
--
Paul