ietf-nntp Article concepts

Charles Lindsey chl at clw.cs.man.ac.uk
Tue Mar 25 04:54:25 PST 2003


In <ylfzpcr4bp.fsf at windlord.stanford.edu> Russ Allbery <rra at stanford.edu> writes:

>Charles Lindsey <chl at clw.cs.man.ac.uk> writes:

>> Be careful about that 'space' after the colon. It is not obligatory in
>> RFC 2822, and one day our arms may be twisted to allow it to be omitted
>> in Usefor (I would rather it were made obligatory in 2822bis, but that
>> is beyond our control).

>I included the space in my proposed wording intentionally, since if you
>read through the descriptions of various commands, you'll find that our
>current standard assumes that it will always be there.  There are various
>references to it in, for example, the OVER and HDR descriptions.

That is fine by me. But it means that Usefor will have to insist on that
SP in the syntax of all headers. I shall quote "existing NNTP standards
and usage" as an added reason when striving to preserve that state of
affairs. But the Dan Kohns of this world are not going to like it :-( .

>Those descriptions could potentially be rewritten to not make that
>assumption, but I'm not sure if it's worth it.

I agree. It crops up in all sorts of places, such as the definition of
overview, as Clive points out.


>> 1. Any article stored by the server MUST have exactly one Message-ID
>> header.

>Too much article format stuff.  I like the way that Clive handled this,
>and I think that all we additionally have to say is that any article
>stored by the server MUST be associated with a message ID and then
>eliminate the <0> stuff.

I agree.

>> 2. If an article arrives without one by POST, the server MUST generate
>> one before storing it.

>I would dearly like to not get involved in the responsibilities of an
>injecting agent and instead defer them to USEFOR.  Can we avoid getting
>into this?

I think that if you say "any article ... MUST be associated with a message
ID" as above, and get rid of that <0>, then you have to say _something_
about POSTed articles that arrive without one; i.e. at least that one must
be generated on the spot (which clearly implies creating a Message-ID
header, whether you say so or not). I agree that we say the minimum we can
about the details of the injecting process, but I think inventing a
message-id has to be a part of it.

>> But now you have to worry about comments and WSP both before and after
>> the <...> (Usefor currently disallows these, but RFC 2822 does not).

>No, you don't; parsing of the header field is outside the scope of the
>NNTP protocol.  From the NNTP perspective, the server takes the news
>article, consults an oracle, and discovers its message ID.  The details of
>how that happens can be deferred to the relevant article standard.

The problem that arises with Clive's new text (and that didn't arise
before) is that he is getting dangerously close to describing parseable
headers. If he steers clear of that, then the existence of that oracle will
have to be implicit, if not explicit. Tricky stuff.

-- 
Charles H. Lindsey ---------At Home, doing my own thing------------------------
Tel: +44 161 436 6131 Fax: +44 161 436 6133   Web: http://www.cs.man.ac.uk/~chl
Email: chl at clw.cs.man.ac.uk      Snail: 5 Clerewood Ave, CHEADLE, SK8 3JU, U.K.
PGP: 2C15F1A9      Fingerprint: 73 6D C2 51 93 A0 01 E7 65 E8 64 7E 14 A4 AB A5



More information about the ietf-nntp mailing list