> And in the case of precedence, even if you eliminate the usage that doesn't
> belong in a header field (control over delay notifications), you still
> have two different camps with incompatible views of how the field should
> be used (in X.400 terms, priority versus importance).
hmmm. perhaps this suggests an approach that registers headers known to be
in common use, but lacking in a formal specification, giving all of them a
single generic specification pointer to a document that says:
"This use of this is not standardized.
There is no formal specification for using this.
People use this in different ways."
We do not need to go into value judgements, but simply provide people with
the facts.
This is basically what Jacob Palme's list of headers has already done. And
FWIW, I've noticed neither an increase nor a decrease in usage of such fields
as a result of its publication. Of course it is hard to extrapolate from this
to the effect a registry would have.
Ned