From: Chris Newman (Chris.Newman@innosoft.com)
Date: Wed Nov 12 1997 - 13:36:01 CST
On Wed, 12 Nov 1997, Charles Lindsey wrote:
> >> 3. Inside of a date-time, though it is actually recommended to include a
> >> comment at the end with the a human-understandable timezone, as in
> >>
> >> Tue, 11 Nov 1997 11:05:02 -0500 (EST)
> >>
> >> Thus, we suggest all the other *CFWS in date-time should become *WS.
>
> >There may be some resistance to this -- mostly because it will make the
> >spec longer if we need two versions of the date/time specification.
>
> Why would you need two versions of the date/time specification? AFAICS
> date-time is used in Date: and Resent-Date:. Unnecessary CFWS is
> undesirable in both cases.
The problem is that 822bis has a strong backwards compatibility
requirement since we're updating a full standard specification. A client
which sticks comments in the middle of the date header is conformant to
RFC 822. Making a previously conformant client non-conformant requires
good justification (e.g., significantly better interoperability with
installed base and few such clients exist). Making a previously
conformant client non-interoperable is simply not an option. Can you
prove that no client exists which generates a comment in the date header
anywhere other than at the end? If not, 822bis has to at least accept
comments in those places.
- Chris