The question here is simple: Can someone reserve and use a field name
for a protocol extension that Keith doesn't like?
This is a yes-no question. Describing the two possible answers as
``extreme positions'' doesn't contribute to the discussion.
Nor does your attempt to cast the question in those terms.
But fwiw, for at least one other protocol that uses an extension
mechanism similar to the one I proposed for header fields, there is
an existence proof that a registration "that Keith doesn't like"
can be accepted.
(the specific example is the PIN NID for URNs, but there are probably
others)
Keith