How did we get from routability tto uniqueness?
uniqueness is the RIR goal for allocations today. routability is a new idea that would have to be seriously investigated to make ULA allocations possible.
"uniqueness among their participants" doesn't sound so bad until people start participating in multiple connectivity domains. I don't want to fumble the math so I'll spare you (and myself), but you get in the situation where there is overlap really fast. Also, in computer science there are only three numbers: 0, 1 and many. If we need more than one block of private space to avoid collisions, the obvious choice is to have as many blocks as there are domains.
to me, the obvious choice, since even after wasting half the address space on EUI64, IPv6 is a really large space compared to distance vector convergence limits, is one block, IPv6:0/0. some parts of this will be routed. some will not be routed. some will be routed some days but not others. all allocations from the RIR system will be unique.
which is precisely the point i'm trying to make. the RIR system can guaranty uniqueness among RIR allocations, but can make no assertions either way as to "routability". since any definition of ULA depends on a definition of "routable", i think this slope is too slippery for us.
As I said before, if ARIN feels it should only give out "hopefully routable" address space like it does today, no problem, let other people give out ULA-C space. But ARINs definition issues don't preclude the usefulness of ULA-C.
since ULA is designed to conserve something that we can't use all of (IPv6 address space) because of something else we can't get enough of (routing table size), at the expense of constraining the future usability of the allocations (forcing a renumbering event to go from unroutable to routable) and requiring a new system of monitoring and enforcement, which by example RFC 1918 still lacks), i have to say, ULA looks like all-pain no-gain to me.