INN commit: trunk/nnrpd (article.c group.c list.c nnrpd.c nnrpd.h)

[Russ Allbery]

Julien ÉLIE <julien at trigofacile.com> writes:

> By the way, is it OK not to strictly parse the range?
>
> I mean:
>
> LISTGROUP newsgroup LETTERS  => 0- (everything)
> LISTGROUP newsgroup 12ABC    => 12-
> LISTGROUP newsgroup 1ABC2    => 1-
> LISTGROUP newsgroup 12-ABC   => 12-0 (thus, nothing)
> LISTGROUP newsgroup -12      => 0-12
> LISTGROUP newsgroup 1-2-3    => 1-2
>
> Are all of these responses OK?  Or must a 501 answer code be returned?
> (it is not specified in the RFC what to do in case the range is wrong...
> yet, I saw that wildmat searches could be interpreted if they are not
> syntactically correct)

This is another one of those bits where it's mostly a question of whether
it's helpful to accept random stuff and make a good guess at what it was
supposed to mean, or whether it's better to be strict and give an error
message.  My inclination would be to keep supporting -12, which is one of
those things that feels like "should" be part of the protocol even if it
isn't.  The rest seem like syntax errors to me, and I'm not sure we're
doing anyone any favors by blindly applying scanf.

> Same question for:
>
>      [C] ARTICLE a.message.id at no.angle.brackets
>      [S] 501 Syntax error
>
> We currently have:
>
>      423 Bad article number
>
> Should it answer 501 instead?

Yeah, that would be better.

> As well as:
>
> ARTICLE <noclosingbracket  (which currently returns "430 No such article")
> ARTICLE 12abc              (not 12?)

Yes, 501 seems more appropriate for both of those to me.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>

    Please send questions to the list rather than mailing me directly.
     <http://www.eyrie.org/~eagle/faqs/questions.html> explains why.