David A. Black wrote:
>
> > On 2004-06-03, daz <dooby / d10.karoo.co.uk> wrote:
> > ><extract>
> > > 2.2. Optional Headers
> > >      ^^^^^^^^
> > >  2.2.5. References
> > >    [...] It is required for all follow-up messages, [...]
> > ></extract>
> >
> > I still don't think that the `References' header should be a problem,
> > no NNTP software I know declines mails on the base of a lacking references
> > header as it's truely optional (every new thread does not have any references).
> > Perhaps feeding the articles could help. I think currently they're submitted
> > via NNRP (the client<->server protocol).
>
> But doesn't the above indicate that it's "required for all follow-up
> messages"?  The problems we've been having are just that: follow-up
> messages (indicated either by an In-Reply-To: header or Re: in the
> subject) which do not have a References: header.
>
>

I don't try to confuse :-(

Section 2.2. Optional Headers (Optional underlined)
 then SUB-Section 2.2.5. References:

Yes, I see the word 'required' but my interpretation is matched
by most, if not all of Usenet's.  It's *not* mandatory.
*Iff* you use this header, this is the official definition.

Bad RFC wording may be the reason that TU-Berlin's NNTP hosting
software is wrongly preventing valid messages going out to Usenet
or it may be deliberate because the group of which they are part
run a public service which needs to enforce the group's rules ??

IOW-1, my news service and others, probably make no checks on
headers at all;  it only needs to check that I'm a subscriber,
then it injects my message into a Usenet server.  They don't
care if it's malformed, junk isn't going to crash Usenet.
But *none* of _our_ messages are malformed.

IOW-2, any massaging of headers would be to make ML -> NG work on
TU-Berlin's machine -- *not* to make them "correct" news messages.


daz