Re: I-D ACTION:draft-klyne-msghdr-registry-02.txt

[Jacob Palme <[e-mail]>]

At 22.23 -0800 02-02-02, Dave Crocker wrote:
>I'm not thrilled with the vocabulary choice of 
>"provisional" for the description of something registered 
>under #2, since it implies a tentativeness in the 
>allocation and an eventual evaluation by the assignment 
>authority.    But quibbling about specific vocabulary is, 
>well, quibbling.

I agree! I suggest "non-standard" instead. But we have to
define a way for amending non-standard header name
descriptions. I think that some kind of IETF consideration
is wanted also for them. The result of the IETF
consideration may be:

- Use of this name is discouraged
- The creator can have change control (might be automatic
   for names beginning with "vnd-" followed by vendor acronym?)
- Some kind of IETF consensus is needed, even if the name
   is not a standard

>Registering header names that have different associated 
>specifications, depending upon which protocol service is 
>being used, is highly counterproductive.  (I'd say 
>"silly", but the role of inflammatory contributor to this 
>thread is well and truly taken, already.)

But when this is (unfortunately) already the case, such as
with "Newsgroups" in e-mail versus news, and "Date" in
e-mail versus HTTP, the registry should reflect this
reality, even if we do not like it. The registry can
then either warn against some or all use of this header
name, or just describe the different usage in different
standards.

---

A related issue: When a registry is started, it should at
the start have a content something like what I have in RFC
2076 (latest version: draft-palme-mailext-headers-06.txt).
I would suggest that we ask a few experts (Ned? Pete?
Keith?) to read throught it and check that the description
of each header is agreeable and can be moved to the
registry. Headers where these experts disagree could be
discussed in the [e-mail] mailing list. To avoid
the risk of 1000-message flame wars, we could say that
if the discussion in ietf-822 does not seem to lead to
consensus, the registry could just present the different
opinions (example: "Reply-To"), possibly, if we can agree
on that, with a warning.

-- 
Jacob Palme <[e-mail]> (Stockholm University and KTH)
for more info see URL: http://www.dsv.su.se/jpalme/