"Harald T. Alvestrand" <hta(_at_)dale(_dot_)uninett(_dot_)no> wrote in message
<199512010932(_dot_)KAA17755(_at_)dale(_dot_)uninett(_dot_)no>:
I believe your scheme for header reclassification has two fundamental
flaws:
1) A header that is useful in several contexts (such as Date:) should
not need to have a prefix of "who-invented-the-idea".
I agree with this (and my proposal didn't include a requirement
for a "who-invented-the-idea" part of the header name).
Example: The Date field is, I think, intended to reflect the
time at which the _composition_ of the message was completed.
Say that we would want to standardize a field with a slightly
different meaning, namely the time at which the message was
_sent_ (for a mail message) or _posted_ (for a news message).
This field would be communication-related, not content-related,
and would have the same meaning irrespective of communication
protocol. In that case we should not define different labels,
such as Mail-Sent-Date and News-Posted-Date, but define one
label, which could be for example Message-Sent-Date.
So the category prefix Message- is used for new headers of a
general communication-related nature. When standardizing new
header fields, the authors and reviewers and commenters of the
draft should always investigate if the header is meaningful for
more than one communication protocol and, in that case, try to
write the draft so that it can be used in all of them. THe draft
should then, of course, be presented in working groups or IETF
mailing lists for all affected protocols.
2) Header names that exist can't be changed.
That's true. My proposal to use the category prefixes Content-,
Message-, Mail-, News-, HTTP-, and Local- was only intended for
_new_ headers, headers that aren't already standardized or on
the standards track. It wouldn't hurt, though, if the new IANA
registry included a column showing the category also for those
old fields that don't have the appropriate prefix. The
application form for the registry should encourage the submitter
to state the category of the header to be registered.
In one case I'm not sure what's best to do: Such headers as Fax
and Phone aren't yet defined in an RFC but it would be easy to
put a specification of them and other related headers on the
standards track. If that is done, should they be given new,
more systematic names, such as Message-Originator-Fax,
Message-Originator-Phone, or should the old names be retained?
/Olle