From: G. James Berigan (usefor@war-of-the-worlds.org)
Date: Fri Jan 28 2000 - 10:53:45 CST
Russ Allbery <rra@stanford.edu> wrote:
>Charles Lindsey <chl@clw.cs.man.ac.uk> writes:
>> 2. In the References header in MESSFOR (and even in RFC 822) the
>> whitespace between references is optional. In RFC 1036 and in our draft
>> it is obligatory (and software is known to barf if it is omitted).
> In the process of writing my message to the DRUMS people about this, I
> noticed that I don't think our specification requires whitespace either.
> Notice:
>
> References-content = msg-id *( CFWS msg-id )
>
> and from MESSFOR:
>
> CFWS = *([FWS] comment) [FWS]
>
> So CFWS can reduce to the empty string. I believe that in order to
> actually require at least one whitespace character, our specification
> would need to be written as:
>
> References-content = msg-id *( FWS [CFWS] msg-id )
But looking at the production we have in section 2.4 of USEFOR:
CFWS = *([FWS] comment) (([FWS] comment) / FWS )
CFWS can reduce to a comment (which must have at least one pair of
parentheses) without whitespace, but its smallest production does lead to
FWS. Our CFWS has to be optional (i.e. [CFWS]) in order to reduce to the
empty string. In MESSFOR, apparently [CFWS] is the same as CFWS.
Though the above production for References-content does make it clear that
a space will be required between a message-ID and the optional comment and
message-ID following it.
Incidentally, what about the " " convention for noting the position of
trimmed IDs? I see USEFOR and MESSFOR both don't allow for CFWS to have
consecutive whitespace. Should FWS be 1*FWS?