From: Bruce Lilly (blilly@erols.com)
Date: Wed Jun 30 2004 - 12:35:42 CDT
Henry Spencer wrote:
> On Wed, 30 Jun 2004, Russ Allbery wrote:
>
>>Maybe we should just ban cmsg for non-control messages? (That might have
>>even been what you meant.)
>
>
> It wasn't what I was thinking of, but you're right, that might be the best
> thing to do.
I don't think a broad-brush ban is reasonable; in any event it has been
discussed before.
>>Regardless of what we do, we should clearly mark this as transitional...
>
>
> Agreed.
Yes, and I think Mark Crispin had some excellent suggestions for dealing
with this sort of situation in
http://www.landfield.com/usefor/2003/Feb/0493.html
and
http://www.landfield.com/usefor/2003/Feb/0554.html
So we could
1. define syntax & semantics by reference to RFC 2822 in USEFOR (no syntax
or semantics issues).
2. define protocol in USEPRO: no protocol use of Subject, defined use for
Control. Deal with the transition issue from RFC 1036's "should be
interpreted as a control message".
3. handle the backwards compatibility issues in USEAGE, e.g. "Old
implementations may interpret 'Subject: cmsg BLURDYBOOP' as a
BLURDYBOOP control message. Therefore it is recommended to avoid
Subjects which might be so interpreted by old implementations.
Injection agents and/or gateways to news from other applications
may accomplish this by suitable mechanisms, e.g. by encoding "cmsg "
via the method detailed in [RFC2047] and [Errata] (taking care to
ensure that the line length restrictions in [RFC2047] are met)."
That way we can produce documents #1 and #2 on schedule, and they need
not change. Document #3 can be revised if and when the affected old
implementations are retired from service, without adversely affecting
the Standards Track documents. Document #3 includes what Mark referred
to as an Implementation Recommendations section.