From what I have seen, those who think "local scope" is harmful, are
concerned about the ambuity of addresses, as Keith says here again.
They are NOT concerned about the fact that a given address may not be
reachable from some places, or may be reachable via different routes
from different places.
There are multiple problems.
One problem is simply that ambiguous addresses are a pain; to use
them in any kind of referral (including DNS), you have to keep track of
the context in which they're valid, and without a global scheme for
naming those contexts there's no way to do that. This is a primary
reason why we haven't been able to identify any good workarounds for
NATs in IPv4 - the very use of ambiguous addresses by those NATs removes
the one way we had of uniquely identifying not just hosts, but
also networks or realms.
(it's scary to think about, but we might actually be able to develop a
reasonable NAT-tolerant architecture if we still had globally unique
names for the realms)
Another problem is that the existence of addresses that have limited
reachability, _coupled with_ the expectation of multiple prefixes per
host, immediately begs the problem of how the host is going to choose
between those prefixes, especially if there's no obligation of the
network to make something like longest prefix match work if the hosts
are allowed to talk to one another. (it also begs the problem of how
how apps can do referrals in that environment - yes, they can pass
around all of the addresses but that's a pain).
Both problems exist, and both are serious. Only the first problem is
specific to SL (I don't consider v6 LL an issue). We don't pretend that
getting rid of SL will address the second problem, but it helps because
the absence of SL provides less incentive for networks to use addressing
tricks as a means of filtering traffic.
Keith