I agree; we need an open registry.
I don't understand Keith's concerns that it would "confer legitimacy
on nonstandard field names". We don't have this problem with other
open registries, say, SASL mechanism names.
different protocols need different constraints on their extension
mechanisms, and I don't think the analogy holds in this case.
- SASL names are not as visible as email field names.
- there are not as many SASL implementaitons as there are MUAs.
- offhand, it seems that SASL mechanisms are more-or-less independent
of the protocol being authenticated (so adding another mechanism doesn't
change the protocol),
- and I don't see how different SASL mechanisms can interact with one
another, either.
on the other hand a field like mail-followup-to can change both the
handling of other fields (like reply-to) but also basic UA functionality
(like the reply command).
Keith