I agree with the folks who have been saying that the spec should not even
list algorithms for which there isn't a solid reference about exactly how
to do the algorithm, including the exact parameters. For example, saying
"18 - elliptic curve" is sure to get you into much more trouble with
interoperability than not saying anything about elliptic curve. Company A
will choose one flavor of elliptic curve, Company B will choose another,
and they will publicly berate each other for doing the wrong thing.
This spec is about interoperability, and keeping focused on that now will
prevent a lot of debate and bad will in the future. I suggest that the
editors add an unambiguous reference for every algorithm listed in section
9, and remove all the ones for which no such reference exists.
--Paul Hoffman, Director
--Internet Mail Consortium