But *why* would an implementation care about the publication
guidelines for the field? That's the *only* thing "X-" tells you.
because any field that doesn't have a published definition probably
isn't something that should be implemented on a widespread basis -
since without a definition you have no reason to expect it to work
predictably from one implementation to another.
admittedly, this inference might be a bit subtle. I see no harm in
making it more explicit.
It's not so much subtle as it is false.
Counterexamples:
X-Face: - there's every expectation of this interoperating just fine.
Note that I said "probably". The fact that there are a small number of
counterexamples doesn't make it false. Yes, it can be made to
interoperate on any platform, but if we were to standardize a field for
the author's face, we might or might not choose this particular field
name, image format, and encoding.
Now IMHO, the use of X- in X-Face shouldn't prevent us from documenting
X-Face in an RFC, nor even from making it a standard - but only if we
really had consensus to choose that name and format.