You have thus far given no demonstration that a registry would in any
way increase the number of UAs using bad fields. I have given (IMHO)
a cogent argument, based on historical practice, that a registry
would *decrease* such occurrences.
I agree that a registry could serve to decrease use of bad fields.
But I don't think a registry is likely to decrease the use of bad fields
unless review happens before registration (else, how do we determine
that they are bad?). I don't think that we're likely to get good
documentation either unless review is a part of the process.
- When many current UAs get charsets they don't know about, they fall
flat on their faces and show garbage to the user.
- When some current UAs get content-types they don't know about
(especially subtypes), they fall flat on their faces and show garbage
to the user, or fail to show the user anything.
- When all of the current UAs that I know of get a header field that
they don't know about, they ignore it.
The first two cases are damaging to the current interoperation of
Internet mail. They deserve the heightened scrutiny that they get.
The present case is different and cannot in a reasonable way be
compared to the first two. It deserves a far lower bar.
I'm not concerned about UA behavior in the presence of header fields
that the UA doesn't know about - as you say, the UA will generally
ignore such fields. I'm far more concerned about UA implementation of
bad header fields because there was no review of those fields before
deployment.
Keith