[IPv6]Re: Analysis of Ungleich ULA Registry

[Kyle Rose <[e-mail]>]

FWIW, to try being constructive for a change, I might recommend the ULA-R
spec be updated to support prefixes as short as (say) /40, but to start
with a /48 and start shaving off bits from the end when more space is
needed, and to always start with 40 bits of randomness. And the update
should indicate in a table how much more likely collisions are with each
fewer bit.

If we publish such an update, we can regard ULA-R prefixes as short as /40
within RIO or PIO fields as known-local. (My preference would be that any
prefix within fc00::/7 should be regarded as known-local for simplicity and
flexibility for operators, but evidently there's been push-back against
that proposal.)

We could also loosen the guidance to allow anywhere between /40 and /64,
i.e., 32-56 bits of randomness, but I haven't really thought through the
implications of longer prefixes.

Kyle