> I would merge the two ideas:  `write-unreadable` outputs "#<", a datum
> as if by `write`, and ">" in that order, and #< is required to signal an
> error satisfying read-error?.  That's enough for a SRFI.

Marc N-W has made me see the light on #<...> so I can no longer
recommend it. I can only accept it as an alternative for the sake of
compatibility.

IMHO lexical syntax SRFIs should be standalone anyway; procedures and
syntax are orthogonal concerns and mixing them causes problems for the
community. But I don't expect to change anyone's mind.