Kari Hurtta wrote:
[2822]
obs-in-reply-to = "In-Reply-To" *WSP ":" *(phrase / msg-id) CRLF
phrase = 1*word / obs-phrase
[I-D.ietf-eai-utf8headers-04]
word = utf8-atom / utf8-quoted-string
Introducing UTF-8 in any <obs-cenities> is a dubious plan. Here's
another example in the same direction:
| obs-bcc = "Bcc" *WSP ":" (address-list / [CFWS]) CRLF
The "C" in CFWS again. Getting a clean split between terms used (also)
in <obs-cenities> and terms used (only) in no-nonsense sections (ch. 3)
in ABNF would be a big effort.
We better solve it once and for all in prose, something like this:
| The revised ABNF specified here is not intended to be used in obsolete
| forms of the relevant header fields or terms defined in the Internet
| Message Format chapter 4 [RFC 2822]. As stated in [RFC 2822] these
| obsolete forms MUST NOT be generated.
|
| Implementations supporting message/utf-8 MAY accept UTF-8 in obsolete
| forms of header fields and terms where the corresponding revised ABNF
| specified here suggests it, but this is not required. The statement
| in [RFC 2822] that all implementations MUST accept the obsolete forms
| does not affect the use of UTF-8 in these obsolete forms.
